BodoiaPuranik-ApplyingReinforcementLearningToCompetitiveTetris
- 格式:pdf
- 大小:161.72 KB
- 文档页数:5
名称英译资产assets流动资产current assets现金及约当现金cash and cash equivalents库存现金cash on hand零用金/周转金petty cash/revolving funds银行存款cash in banks在途现金cash in transit约当现金cash equivalents其它现金及约当现金other cash and cash equivalents短期投资short-term investment短期投资-股票short-term investments-stock短期投资-短期票券short-term investments-short-term notesand bills短期投资-政府债券short-term investments-government bonds短期投资-受益凭证short-term investments-beneficiarycertificates短期投资-公司债short-term investments-corporate bonds 短期投资-其它short-term investments-other备抵短期投资跌价损失allowance for reduction of short-terminvestment to market应收票据notes receivable应收票据notes receivable应收票据贴现discounted notes receivable应收票据-关系人notes receivable-related parties其它应收票据other notes receivable备抵呆帐-应收票据allowance for uncollec-tible accounts-notesreceivable应收帐款accounts receivable应收帐款accounts receivable应收分期帐款installment accounts receivable应收帐款-关系人accounts receivable-related parties备抵呆帐-应收帐款allowance for uncollec-tible accounts-accounts receivable其它应收款other Receivables应收出售远汇款forward exchange contract receivable应收远汇款-外币forward exchange contract receivable-foreign currencies买卖远汇折价discount on forward ex-change contract应收收益earned revenue receivable应收退税款income tax refund receivable其它应收款-关系人other receivables-related parties其它应收款-其它other receivables-other备抵呆帐-其它应收款allowance for uncollec-tible accounts-otherreceivables存货inventories商品存货merchandise inventory寄销商品consigned goods在途商品goods in transit备抵存货跌价损失allowance for reduction of inventory tomarket制成品finished goods寄销制成品consigned finished goods副产品by-products在制品work in process委外加工work in process-Outsourced原料raw materials物料supplies在途原物料materials and supplies in transit备抵存货跌价损失allowance for reduction of inventory tomarket预付费用prepaid expenses预付薪资prepaid payroll预付租金prepaid rents预付保险费prepaid insurance用品盘存office supplies预付所得税prepaid income tax其它预付费用other prepaid expenses预付款项prepayments预付货款prepayment for purchases其它预付款项other prepayments其它流动资产other current assets进项税额VAT paid ( or input tax)留抵税额excess VAT paid (or overpaid VAT)暂付款temporary payments代付款payment on behalf of others员工借支advances to employees存出保证金refundable deposits受限制存款certificate of deposit-restricted递延所得税资产deferred income tax assets递延兑换损失deferred Foreign Exchange losses业主往来(股东往来)owners'(stockholders') current account 同业往来current account with others其它流动资产-其它other current assets-other基金及长期投资funds and long-term investments基金funds偿债基金redemption fund (or sinking fund)改良及扩充基金fund for improvement and expansion 意外损失准备基金contingency fund退休基金pension fund其它基金other funds长期投资long-term investments长期股权投资long-term equity investments长期债券投资long-term bond investments长期不动产投资long-term real estate in-vestments人寿保险现金解约价值cash Surrender value of life insurance 其它长期投资other long-term investments备抵长期投资跌价损失allowance for excess of cost over marketvalue of long-term investments固定资产property , plant, and equipment土地land土地land土地-重估增值land-revaluation increments土地改良物land improvements土地改良物land improvements土地改良物-重估增值land improvements-revaluation increments累积折旧-土地改良物accumulated depreciation-landimprovements房屋及建物buildings房屋及建物buildings房屋及建物-重估增值buildings-revaluation increments累积折旧-房屋及建物accumulated depreciation-buildings机(器)具及设备machinery and equipment机(器)具machinery机(器)具-重估增值machinery-revaluation increments累积折旧-机(器)具accumulated depreciation-machinery租赁资产leased assets租赁资产leased assets累积折旧-租赁资产accumulated depreciation-leased assets 租赁权益改良leasehold improvements租赁权益改良leasehold improvements累积折旧-租赁权益改良accumulated depreciation-leaseholdimprovements未完工程及预付购置设备款construction in progress and prepaymentsfor equipment未完工程construction in progress预付购置设备款prepayment for equipment杂项固定资产miscellaneous property, plant, andequipment杂项固定资产miscellaneous property, plant, andequipment杂项固定资产-重估增值miscellaneous property, plant, andequipment-revaluation increments累积折旧-杂项固定资产accumulated depreciation-miscellaneousproperty, plant, and equipment递耗资产depletable assets递耗资产depletable assets天然资源natural resources重估增值-天然资源natural resources-revaluation increments 累积折耗-天然资源accumulated depletion-natural resources 无形资产intangible assets商标权trademarks商标权trademarks专利权patents专利权patents特许权franchise特许权franchise著作权copyright著作权copyright计算机软件Computer Software计算机软件computer software cost商誉goodwill商誉goodwill开办费organization costs开办费organization costs其它无形资产other intangibles递延退休金成本deferred pension costs租赁权益改良leasehold improvements其它无形资产-其它other intangible assets-other其它资产other assets递延资产deferred assets债券发行成本deferred bond issuance costs长期预付租金long-term prepaid rent长期预付保险费long-term prepaid insurance递延所得税资产deferred income tax assets预付退休金prepaid pension cost其它递延资产other deferred assets闲置资产idle assets闲置资产idle assets长期应收票据及款项与催收帐款long-term notes , accounts and overduereceivables长期应收票据long-term notes receivable长期应收帐款long-term accounts receivable催收帐款overdue receivables长期应收票据及款项与催收帐款-关系人long-term notes, accounts and overdue receivables-related parties其它长期应收款项other long-term receivables备抵呆帐-长期应收票据及款项与催收帐款allowance for uncollectible accounts-long-term notes, accounts and overdue receivables出租资产assets leased to others出租资产assets leased to others出租资产-重估增值assets leased to others-incremental valuefrom revaluation累积折旧-出租资产accumulated depreciation-assets leased toothers存出保证金refundable deposit存出保证金refundable deposits杂项资产miscellaneous assets受限制存款certificate of deposit-restricted杂项资产-其它miscellaneous assets-other负债liabilities流动负债current liabilities短期借款short-term borrowings(debt)银行透支bank overdraft银行借款bank loan短期借款-业主short-term borrowings-owners短期借款-员工short-term borrowings-employees短期借款-关系人short-term borrowings-related parties短期借款-其它short-term borrowings-other应付短期票券short-term notes and bills payable应付商业本票commercial paper payable银行承兑汇票bank acceptance其它应付短期票券other short-term notes and bills payable 应付短期票券折价discount on short-term notes and billspayable应付票据notes payable应付票据notes payable应付票据-关系人notes payable-related parties其它应付票据other notes payable应付帐款accounts pay able应付帐款accounts payable应付帐款-关系人accounts payable-related parties应付所得税income taxes payable应付所得税income tax payable应付费用accrued expenses应付薪工accrued payroll应付租金accrued rent payable应付利息accrued interest payable应付营业税accrued VAT payable应付税捐-其它accrued taxes payable-other其它应付费用other accrued expenses payable其它应付款other payables应付购入远汇款forward exchange contract payable应付远汇款-外币forward exchange contract payable-foreigncurrencies买卖远汇溢价premium on forward exchange contract应付土地房屋款payables on land and building purchased 应付设备款Payables on equipment其它应付款-关系人other payables-related parties应付股利dividend payable应付红利bonus payable应付董监事酬劳compensation payable to directors andsupervisors其它应付款-其它other payables-other预收款项advance receipts预收货款sales revenue received in advance预收收入revenue received in advance其它预收款other advance receipts一年或一营业周期内到期长期负债long-term liabilities-current portion一年或一营业周期内到期公司债corporate bonds payable-current portion一年或一营业周期内到期长期借款long-term loans payable-current portion一年或一营业周期内到期长期应付票据及款项long-term notes and accounts payable due within one year or one operating cycle一年或一营业周期内到期长期应付票据及款项-关系人long-term notes and accounts payables to related parties-current portion其它一年或一营业周期内到期长期负债other long-term lia-bilities-current portion 其它流动负债other current liabilities销项税额VAT received(or output tax)暂收款temporary receipts代收款receipts under custody估计售后服务/保固负债estimated warranty liabilities递延所得税负债deferred income tax liabilities递延兑换利益deferred foreign exchange gain业主(股东)往来owners' current account同业往来current account with others其它流动负债-其它other current liabilities-others长期负债long-term liabilities应付公司债corporate bonds payable应付公司债corporate bonds payable应付公司债溢价(折价)premium(discount) on corporate bondspayable长期借款long-term loans payable长期银行借款long-term loans payable-bank长期借款-业主long-term loans payable-owners长期借款-员工long-term loans payable-employees长期借款-关系人long-term loans payable-related parties 长期借款-其它long-term loans payable-other长期应付票据及款项long-term notes and accounts payable长期应付票据long-term notes payable长期应付帐款long-term accounts pay-able长期应付租赁负债long-term capital lease liabilities长期应付票据及款项-关系人Long-term notes and accounts payable-related parties其它长期应付款项other long-term payables估计应付土地增值税accrued liabilities for land value incrementtax估计应付土地增值税estimated accrued land value incrementaltax pay-able应计退休金负债accrued pension liabilities应计退休金负债accrued pension liabilities其它长期负债other long-term liabilities其它长期负债-其它other long-term liabilities-other其它负债other liabilities递延负债deferred liabilities递延收入deferred revenue递延所得税负债deferred income tax liabilities其它递延负债other deferred liabilities存入保证金deposits received存入保证金guarantee deposit received杂项负债miscellaneous liabilities杂项负债-其它miscellaneous liabilities-other业主权益owners' equity资本capital资本(或股本)capital普通股股本capital-Common stock特别股股本capital-preferred stock预收股本capital collected in advance待分配股票股利stock dividends to be distributed资本capital资本公积additional paid-in capital股票溢价paid-in capital in excess of par普通股股票溢价paid-in capital in excess of par-commonstock特别股股票溢价paid-in capital in excess of par-preferredstock资产重估增值准备capital surplus from assets revaluation资产重估增值准备capital surplus from assets revaluation处分资产溢价公积capital surplus from gain on disposal ofassets处分资产溢价公积capital surplus from gain on disposal ofassets合并公积capital surplus from business combination 合并公积capital surplus from business combination 受赠公积donated surplus受赠公积donated surplus其它资本公积other additional paid-in capital权益法长期股权投资资本公积additional paid-in capital from investeeunder equity method资本公积-库藏股票交易additional paid-in capital-treasury stocktrans-actions保留盈余(或累积亏损)retained earnings (accumulated deficit)法定盈余公积legal reserve法定盈余公积legal reserve特别盈余公积special reserve意外损失准备contingency reserve改良扩充准备improvement and expansion reserve偿债准备special reserve for redemption of liabilities 其它特别盈余公积other special reserve未分配盈余(或累积亏损)retained earnings-unappropriated (oraccumulated deficit)累积盈亏accumulated profit or loss前期损益调整prior period adjustments本期损益net income or loss for current period权益调整equity adjustments长期股权投资未实现跌价损失unrealized loss on market value decline oflong-term equity investments长期股权投资未实现跌价损失unrealized loss on market value decline oflong-term equity investments累积换算调整数cumulative translation adjustment累积换算调整数cumulative translation adjustments未认列为退休金成本之净损失net loss not recognized as pension cost 未认列为退休金成本之净损失net loss not recognized as pension costs 库藏股treasury stock库藏股treasury stock库藏股treasury stock少数股权minority interest少数股权minority interest少数股权minority interest营业收入operating revenue销货收入sales revenue销货收入sales revenue销货收入sales revenue分期付款销货收入installment sales revenue销货退回sales return销货退回sales return销货折让sales allowances销货折让sales discounts and allowances劳务收入service revenue劳务收入service revenue劳务收入service revenue业务收入agency revenue业务收入agency revenue业务收入agency revenue其它营业收入other operating revenue其它营业收入-其它other operating revenue其它营业收入-其它other operating revenue-other营业成本operating costs销货成本cost of goods sold销货成本cost of goods sold销货成本cost of goods sold分期付款销货成本installment cost of goods sold进货purchases进货purchases进货费用purchase expenses进货退出purchase returns进货折让charges on purchased merchandise进料materials purchased进料material purchased进料费用charges on purchased material进料退出material purchase returns进料折让material purchase allowances直接人工direct labor直接人工direct labor制造费用manufacturing overhead间接人工indirect labor租金支出rent expense, rent文具用品office supplies (expense)旅费travelling expense, travel运费shipping expenses, freight邮电费postage (expenses)修缮费repair(s) and maintenance (expense )包装费packing expenses水电瓦斯费Utilities (expense)保险费insurance (expense)加工费manufacturing overhead-outsourced 税捐taxes折旧depreciation expense各项耗竭及摊提various amortization伙食费meal (expenses)职工福利employee benefits/welfare训练费training (expense)间接材料indirect materials其它制造费用other manufacturing expenses劳务成本制ervice costs劳务成本service costs劳务成本service costs业务成本gency costs业务成本agency costs业务成本agency costs其它营业成本other operating costs其它营业成本-其它other operating costs-other其它营业成本-其它other operating costs-other营业费用operating expenses推销费用selling expenses推销费用selling expenses薪资支出payroll expense租金支出rent expense, rent文具用品office supplies (expense)旅费travelling expense, travel运费shipping expenses, freight邮电费postage (expenses)修缮费repair(s) and maintenance (expense)广告费advertisement expense, advertisement 水电瓦斯费utilities (expense)保险费insurance (expense)交际费entertainment (expense)捐赠donation (expense)税捐taxes呆帐损失loss on uncollectible accounts折旧depreciation expense各项耗竭及摊提various amortization伙食费meal (expenses)职工福利employee benefits/welfare佣金支出commission (expense)训练费training (expense)其它推销费用other selling expenses管理及总务费用general & administrative expenses管理及总务费用general & administrative expenses薪资支出payroll expense租金支出rent expense, rent文具用品office supplies旅费travelling expense, travel运费shipping expenses,freight邮电费postage (expenses)修缮费repair(s) and maintenance (expense)广告费advertisement expense, advertisement水电瓦斯费utilities (expense)保险费insurance (expense)交际费entertainment (expense)捐赠donation (expense)税捐taxes呆帐损失loss on uncollectible accounts折旧depreciation expense a各项耗竭及摊提various amortization外销损失loss on export sales伙食费meal (expenses)职工福利employee benefits/welfare研究发展费用research and development expense佣金支出commission (expense)训练费training (expense)劳务费professional service fees其它管理及总务费用other general and administrative expenses 研究发展费用research and development expenses研究发展费用research and development expenses薪资支出payroll expense租金支出rent expense, rent文具用品office supplies旅费travelling expense, travel运费shipping expenses, freight邮电费postage (expenses)修缮费repair(s) and maintenance (expense)水电瓦斯费utilities (expense)保险费insurance (expense)交际费entertainment (expense)税捐taxes折旧depreciation expense各项耗竭及摊提various amortization伙食费meal (expenses)职工福利employee benefits/welfare训练费training (expense)其它研究发展费用other research and development expenses营业外收入及费用non-operating revenue and expenses,other income(expense)营业外收入non-operating revenue利息收入interest revenue利息收入interest revenue/income投资收益investment income权益法认列之投资收益investment income recognized under equitymethod股利收入dividends income短期投资市价回升利益gain on market price recovery of short-terminvestment兑换利益foreign exchange gain兑换利益foreign exchange gain处分投资收益gain on disposal of investments处分投资收益gain on disposal of investments处分资产溢价收入gain on disposal of assets处分资产溢价收入gain on disposal of assets其它营业外收入other non-operating revenue捐赠收入donation income租金收入rent revenue/income佣金收入commission revenue/income出售下脚及废料收入revenue from sale of scraps存货盘盈gain on physical inventory存货跌价回升利益gain from price recovery of inventory坏帐转回利益gain on reversal of bad debts其它营业外收入-其它other non-operating revenue-other items 营业外费用non-operating expenses利息费用interest expense利息费用interest expense投资损失investment loss权益法认列之投资损失investment loss recog-nized under equitymethod短期投资未实现跌价损失unrealized loss on reduction of short-terminvestments to market兑换损失foreign exchange loss兑换损失foreign exchange loss处分投资损失loss on disposal of investments处分投资损失loss on disposal of investments处分资产损失loss on disposal of assets处分资产损失loss on disposal of assets其它营业外费用other non-operating expenses停工损失loss on work stoppages灾害损失casualty loss存货盘损loss on physical inventory存货跌价及呆滞损失loss for market price decline and obsoleteand slow-moving inventories其它营业外费用-其它other non-operating expenses-other所得税费用(或利益)income tax expense (or benefit)所得税费用(或利益)income tax expense (or benefit)所得税费用(或利益)income tax expense (or benefit)所得税费用(或利益)income tax expense ( or benefit)非经常营业损益nonrecurring gain or loss停业部门损益gain(loss) from discontinued operations停业部门损益-停业前营业损益income(loss) from operations ofdiscontinued segments停业部门损益-停业前营业损益income(loss) from operations ofdiscontinued segment停业部门损益-处分损益gain(loss) from disposal of discontinuedsegments停业部门损益-处分损益gain(loss) from disposal of discontinuedsegment非常损益extraordinary gain or loss非常损益extraordinary gain or loss非常损益extraordinary gain or loss会计原则变动累积影响数cumulative effect of changes in accountingprinciples会计原则变动累积影响数cumulative effect of changes in accountingprinciples会计原则变动累积影响数cumulative effect of changes in accountingprinciples少数股权净利minority interest income少数股权净利minority interest income少数股权净利minority interest incom。
吸附试验英文专业词组In the realm of scientific experimentation, the term "adsorption test" refers to a method used to determine the capacity of a substance to attract and hold other molecules on its surface.This process is pivotal in various fields, including chemistry and environmental science, where it helps in understanding the interaction between different materials and pollutants.The efficiency of an adsorbent is often measured by its "adsorption capacity," which is the amount of substance that can be held on the surface per unit weight of the adsorbent.Experiments may involve the use of "adsorbate," the substance being attracted, and "adsorbent," the material doing the attracting. The interaction is typically studied under controlled conditions to ensure accurate results.One common technique for conducting an adsorption test is "batch adsorption," where a known quantity of adsorbate is introduced into a solution containing the adsorbent.Monitoring the "adsorption kinetics" is essential, as it provides insights into the rate at which the adsorption process occurs and helps in determining the optimal contact time.Another important aspect is the "adsorption isotherm," which graphically represents the relationship between the amount of adsorbate on the surface and its concentration in the solution at equilibrium.The "adsorption equilibrium constant" is a measure of the affinity between the adsorbate and the adsorbent, indicating the strength of the interaction.Finally, understanding the "adsorption mechanism" is crucial for interpreting the results and predicting the behavior of different materials under various conditions. This knowledge is invaluable for developing new materials and technologies.。
A constitutive model of frozen saline sandy soil based onenergy dissipation theoryYuanming Lai a ,b ,*,Mengke Liao a ,c ,Kai Hu aa State Key Laboratory Frozen Soil Engineering,Cold and Arid Region Environmental and Engineering Institute,Chinese Academy ofSciences,Lanzhou,Gansu 730000,Chinab School of Civil Engineering,Lanzhou Jiaotong University,Lanzhou,Gansu 730070,Chinac University of Chinese Academy of Sciences,Beijing,100049,Chinaa r t i c l e i n f oArticle history:Received 3February 2015Received in revised form 10October 2015Available online 2November 2015Keywords:B:AnisotropyB:Mechanical propertiesB:Plastic deformationB:ThermomechanicalFrozen saline soil a b s t r a c t A series of triaxial compression tests are carried out for frozen saline sandy soil with Na 2SO 4contents 0.0,0.5,1.5,and 2.5%under con fining pressures from 0MPa to 16MPa at À6 C,respectively.The test results indicate that,the Critical State Line (CSL)of frozen saline sandy soil is curve and is not through the origin in (p ,q )plane,and the soil particles have the properties of initial anisotropic rotational angle and loaded anisotropy in process of loading.In order to describe the deformation properties of frozen saline sandy soil,a new double yield surface constitutive model is proposed based on the triaxial compression tests in this paper.The proposed model contains the in fluence of salt contents on me-chanical characteristics,so it is suitable for describing the stress e strain relation of frozen saline soil.The proposed model has the following properties:(1)By de fining a modi fiedeffective stress p *,a critical state strength envelope function is established according to theModi fied Cam Clay model.The envelope approximates a straight line under low con finingpressures,but it is curve downward under high con fining pressures due to pressuremelting.(2)The effect of the initial anisotropic rotational angle and loaded anisotropy,during process of loading under plastic volumetric compression mechanism,on yieldsurface of rotational hardening is taken into account.(3)A paraboloid yield surfacefunction,including the rotational hardening law induced by loading,is proposed underplastic shear mechanism.The universality of the proposed model is veri fied by the testresults of frozen saline sandy soil under different stress paths.Finally,in order to furtherstudy the applicability of the proposed model in this paper,the stress e strain relation ofthe cemented clay is simulated by it.And the in fluences of the pressure melting phe-nomenon and the rotational hardening rule on the calculated results of the proposedmodel are investigated.The research results indicate that the proposed model can simulatenot only the mechanics properties of materials whose CSL is straight but also those ofmaterials whose CSL is curved,other than predict the deformation regularity of frozensaline sandy soil well.©2015The Authors.Published by Elsevier Ltd.This is an open access article under the CCBY-NC-ND license (/licenses/by-nc-nd/4.0/).*Corresponding author.State Key Laboratory Frozen Soil Engineering,Cold and Arid Region Environmental and Engineering Institute,Chinese Academy of Sciences,Lanzhou,Gansu 730000,China.Tel.:þ869314967288.E-mail address:ymlai@ (Y.Lai).Contents lists available at ScienceDirectInternational Journal of Plasticityjournal homepa ge:www.elsevie/locate/ijplas/10.1016/j.ijplas.2015.10.0080749-6419/©2015The Authors.Published by Elsevier Ltd.This is an open access article under the CC BY-NC-ND license (/licenses/by-nc-nd/4.0/).International Journal of Plasticity 78(2016)84e 113i et al./International Journal of Plasticity78(2016)84e11385 1.IntroductionFrozen soil consists of mineral particles,ice inclusions,liquid water and gaseous inclusions(Tsytovich,1985).And the frozen saline soil is a type of frozen soil which contains solution of soluble salt and salt crystal.Test results show that the precipitation of salt crystals at negative temperature is strongly sensitive to temperature(Wan and Lai,2013).Compared with general frozen soil,the structure of the frozen saline sandy soil is changed with the precipitation of ice crystal and salt crystal in the freezing process,so the theoretical modeling of the constitutive relation of frozen saline frozen soil is more complex and difficult than that of other soils.With the development of society and economics,many engineering projects,such as highways,railways,houses,airports and oil pipelines,will be constructed in cold saline soil regions.At present,the study on mechanical property for frozen saline soil is rarely performed,so it is significant and essential to investigate the mechanical behaviors of frozen saline soil for engineering design.Based on the theory of critical state soil mechanics,Roscoe et al.(1963)proposed Cambridge Clay model for soft soil. Afterward Roscoe and Burland(1968)presented Modified Cam Clay de and Duncan(1978)formulated a dilatancy model for sandy soil,which is called Lade-Duncan model.Since then,the constitutive model research of geotechnical ma-terials becomes a hot issue.In the case of Rock Mechanics,Khan and co-workers have developed constitutive models to capture elasto-plastic behavior of Berea sandstone under a large range of confining pressures(Khan et al.,1991,1992).In damage constitutive model study on the brittle material,Shojaei et al.(2013)built a damage constitutive theory based on the dynamic dissipation work method for polycrystalline materials under dynamic loading,and developed a continuum damage mechanics(CDM)constitutive model to describe elastic,plastic and damage behavior of porous rocks(Shojaei et al.,2014).For unfrozen geomaterials,some researchers proposed a lot of strength criteria and constitutive models to solve different en-gineering problems(Li and Dafalias,2000;Altenbach et al.,2001;Hashiguchi and Tsutsumi,2003,2007;Hashiguchi,2005; Vorobiev,2008;Steinhauser et al.,2009;Shen et al.,2012;Xie and Shao,2012;Yao and Wang,2014;Mortara,2015).The deformation behaviors of soils are usually identified by two basic plasticflow mechanisms:one is related to plastic shear,and the other to plastic volumetric compression(Xie and Shao,2006).In order to describe the two plasticflow mechanisms well,a double yield surface model has been established by combining a cap surface for plastic compression and a cone surface for plastic shear(Peric and Ayari,2002;Lai et al.,2010).Because of the structural complexity of frozen saline sandy soil,it is difficult to describe the mechanical properties by a single yield surface,so the mechanical characteristics of plastic volumetric compression mechanism and plastic shear mechanism are used to investigate the constitutive relation of frozen saline sandy soil in this paper.Many researchers have made a lot of researches to describe the mechanical characteristics of general frozen soil (Chamberlain et al.,1972;Bragg and Andersland,1981;Fish,1991;Hashiguchi,2005;Zhang et al.,2007;Lai et al.,2009;Yang et al.,2010;De and Pereira,2013;Painter and Karra,2014).By the test and numerical methods and selecting different test materials and theory,the various affecting factors on mechanical behaviors for frozen soil have been studied and different constitutive models have been established up to the present.The more detailed frozen soil test methods and research pro-cesses were introduced in the authors’previous works(Lai et al.,2009,2014).However,it should be pointed out that stress e strain relationship of frozen saline sandy soil is more complex than that of unfrozen soil or general frozen soil.During the freezing process,ice crystals,salt crystals,and soil particles are cemented together to make the frozen saline sandy soil have tensile capacity.The characteristics of frozen soil are similar to that of cemented clay.In the study on mechanical behavior of cemented clay,Suebsuk et al.(2010)proposed a constitutive model,based on the Structure Cam Clay(SCC)model proposed by Liu and Carter(2002),to simulate the mechanical behaviors of cemented clay better.Gao and Zhao(2012)and Nguyen et al.(2014)proposed a modified SCC model and its simulated results agree well with the test results of cemented clay with different cement contents.However,with the increase of confining pressure,the frozen saline sandy soil has a char-acteristic of pressure melting but the cemented clay does not have it.The pressure melting of frozen soils results in that with the increase of confining pressure,the critical state line(CSL)of frozen saline sandy soil will gradually bend downward,while the critical state line of cemented clay is linear.The critical state line of the frozen saline sandy soil will degenerate as that of cemented clay if the influence of pressure melting is not considered.During the process of shearing loading or isotropic compression,the unfrozen geotechnical granular materials appear different degree of anisotropic properties(Anandarajah,2008;Yin and Chang,2010;Rowshanzamir and Askari,2010;Liu et al.,2013;Fonseca et al.,2013).Based on theory and tests,many researchers have made a lot of studies on anisotropic properties for granular material and soft soil.Voyiadjis et al.(1995)proposed a granular material constitutive relation based on anisotropic rotating yield surface,and studied the deformation of anisotropic elasto-plastic damage for concrete(Voyiadjis et al.,2008).Cleja-Tigoiu(2000)established an anisotropicfinite elasto-plastic constitutive equation by combining the ki-nematic hardening law of material.Based on test data of soft soils,Wheeler et al.(2003)proposed an anisotropic elasto-plastic constitutive model called as S-CLAY1,in which a rotational yield surface and a rotational hardening law were used, and the sensitivity of rotational hardening parameters was analyzed.In the previous study,it is found that the plasticflow directions does not coincide with the hydrostatic axis(Lai et al.,2009)when the isotropic consolidation was completed and the shearing began,and the initial anisotropy is produced in the process of consolidation.From the triaxial tests of frozen saline sandy soil,it is found that the soil particles generate initial anisotropic angle before shearing loading due to the consolidation pressure.With the increasing of shearing loading,the anisotropic rotational angle of frozen saline sandy soil decreases from initial rotation angle below horizontal axis to zero,and then increases to the maximum rotational angle above horizontal axis.In the geotechnical engineering theoretical analysis and study,the thermodynamics theory is introduced to solve the problem of related theory.Based on thermodynamic framework,Collins and Hilder (2002)built a theory foundation of anisotropic elasto-plastic rotational hardening model by triaxial tests,and studied the forms of Helmholtz free energy rate and the parameter's choice of dissipation rate.Voyiadjis et al.(2011)proposed a thermodynamic damage and healing process constitutive model for viscoplastic materials,which contains two new yield surfaces for the damage and healing processes that take into account the isotropic hardening effect.Xiao (2014)proposed a rate-independent finite elastoplastic equation with thermo-coupled effects to bypass the yield condition and loading-unloading conditions.The purpose of this paper is to propose a constitutive model which can describe the mechanical behaviors and the particle rotational rule of frozen saline sandy soil.Based on the thermodynamic theory,we proposed a new rotational hardening double yield surface constitutive model of frozen saline sandy soil considering initial anisotropy and loaded anisotropy.By employing double yield surface functions and non-associated plastic potential functions,the proposed model contains plastic compression mechanism and plastic shear mechanism.Additionally,the in fluences of rotational hardening rule on the calculated results are investigated to validate the advantages of the proposed model of frozen saline sandy soil.It is found that the calculated results of the proposed model are agreement well with the test results.2.Test conditions and results2.1.Description of the testThe tested Na 2SO 4saline sandy soil was collected from permafrost regions in Qinghai-Tibet plateau and its grain size composition is given in Table 1.Firstly,in order to desalinate the soil,salt in saline sandy soil was removed ten times with distilled water.The desalinated soil sample was dried at 105 C for 24h,and then pulverized.Secondly,the dry preparation soil samples are selected to make up the saline sandy soil whose salt contents are 0,0.5,1.5,and 2.5%(The salt is anhydrous Na 2SO 4fine particles),respectively.The freezing temperatures of saline sandy soil with salt contents 0.0,0.5,1.5,and 2.5%are À0.19 C,À1.47 C,À1.54 C and À1.58 C,respectively.To ensure the uniformity of the samples,the saline sandy soil was mixed with water at moisture content of 13%,and then kept for 12h without evaporation to allow the water to be uniformly distributed in the soil.When the water was uniformly distributed in the soil,the prepared saline sandy soil was filled in a cylindrical mold to make cylindrical soil specimens with target dry density 1.89g/cm 3under certain compression rate.The specimens were prepared as cylinders with 6.18cm in diameter and 12.5cm high.Then,the specimens in the molds were saturated under a vacuum.In order to avoid large frost heaving and prevent moisture migration,the soil specimens were placed in a refrig-eration unit and frozen quickly at a temperature below À30 C.After 48h of freezing,the molds were removed and the specimens were mounted with epoxy resin plates on both ends and covered with a rubber sleeve to avoid moisture evap-oration.Finally,the specimens were kept in an incubator for over 24h at the test target temperature of À6 C,such that the specimens reached a uniform temperature.The con fining pressures were from 0to 16MPa,and the strain rate of axial loading _εa ¼1:67Â10À4s À1was employed in the conventional triaxial test according to the GB/T 50123-1999(People's Republic of China National Standard,1999).Each sample was loaded until failure and the test duration was about 20min,so the conventional triaxial test is fast shear test,and the deformation of frozen soil is mainly elastic e plastic deformation,and the in fluence of loading rate on mechanical properties can be neglected.Therefore,during so short test time,the visco-plastic deformation is not taken into account in the analyses of test results.Based on the conventional triaxial fast shear test results,the rate-independent mechanical properties of frozen soil are investigated in this paper.2.2.Test results and analysesThe present study used a cryogenic triaxial apparatus improved from the MTS-810material test machine which was illustrated by Lai et al.(2014).After the specimens mentioned above were prepared,they were placed into the pressure cell of the MTS-810material test machine and a series of triaxial compression tests were performed.Firstly,the temperature in the pressure chamber was set at À6 C with a precision of ±0.1 C;then,the prepared specimen was consolidated under a preset pressure for 2h prior to the axial loading.In the axisymmetric case of specimen,the principal stress s 2¼s 3,and the effective principal stress of axial load is s 1.We suppose that the stress and strain be positive under compression condition like the assumption in geotechnical engineering.The stress and strain for conventional triaxial test are expressed as:Table 1Physical parameters of saline sandy soil (unit:%).Composition of particle diameters>0.50mm0.50e 0.25mm 0.25e 0.075mm 0.075e 0.005mm <0.005mm 3.72710.57344.36035.920 5.420i et al./International Journal of Plasticity 78(2016)84e 11386p ¼13ðs 1þ2s 3Þ;q ¼s 1Às 3(1a)εv ¼ε1þ2ε3;εg ¼23ðε1Àε3Þ(1b)where p ,q ,εv ,and εg are hydrostatic pressure,deviator stress,volumetric strain,and shear strain,respectively.The normal components of the stress and strain tensors are assumed to be positive under compression condition like the general assumption in geotechnical engineering.A series of triaxial compression tests were carried out at À6 C at the salt contents of 0.0,0.5,1.5,and 2.5%,respectively.The test results of frozen saline sandy soil with salt content 2.5%are shown in Fig.1.3.Thermodynamics basic frameworkThe theory of thermodynamics has better been developed and applied to research of geotechnical constitutive relation (Collins and Kelly,2002).Based on the general thermodynamic method,many new critical state constitutive models can be developed (Collins and Kelly,2002;Collins and Hilder,2002).3.1.The first and second laws of thermodynamicsThe first law of thermodynamics states that there exists a state function called the internal energy (Ziegler,1977;Collins and Houlsby,1997),suchthat:Fig.1.Test result curves of frozen saline sandy soil with salt content 2.5%under various con fining pressures.i et al./International Journal of Plasticity 78(2016)84e 11387_U ¼s ij _εij Àq k ;k (2)where U ,q i ,s ij ,and εij are internal energy in unit volume,heat flow vector,stress tensor,and strain tensor,respectively.When the deformation of material is irreversible,the internal dissipation rate _fsatis fies the Clausius e Duham inequality of the second law of thermodynamics,which is expressed as:_f ¼T _s i ¼T ð_s À_s r Þ¼T _s À Àq k ;k þq k T ;k T !0(3)where _si is internal entropy rate when the material deformation is irreversible;_s r is external entropy rate and _s r ¼Àgrad q k T ;_sis system entropy generation rate;T is absolute temperature.Substituting Eq.(2)into Eq.(3),we obtain:_f ¼s ij _εij À _U ÀT _s Àq k T ;k T !0(4)Supposing that j ¼U ÀTs is the (Helmholtz)free energy of unit mass,then,we have _j¼_U ÀT _s Às _T .Based on Eq.(4),the following formula can be obtained:_f ¼s ij _εij À _j þs _T Àq k T ;k !0(5)3.2.Thermodynamics basic equationsIf the continuum element undergoes isothermal condition and irreversible deformation process,the in fluence of tem-perature on the dissipation function can be omitted,i.e.,_T¼0and T ,k ¼0in Eq.(5).Based on the conventional triaxial fast shear test,the deformation of frozen soil is mainly elastic e plastic deformation,and the in fluence of loading rate on me-chanical properties cannot be taken into account.Therefore,the visco-plastic characteristics are not considered in the deformation of the material in this study.For the elasto-plastic problem of rate-independent materials in isothermal con-dition,the elasto-plastic constitutive relation can be expressed by incremental forms.The incremental constitutive laws can be established based on the thermodynamics theory (Collins and Houlsby,1997;Collins and Kelly,2002).The incremental work done by the applied stresses is the sum of the increment in the free energy function,j ,and the increment in the dissipation function,f .Both functions are de fined in per unit volume.Here,based on the above conditions and simplifying Eq.(5),the incremental work-energy function can be expressed as (Collins and Hidder,2002;Collins and Kelly,2002):s ij d εij ¼d j þdf ;where df !0(6)where s ij d εij ,d j and df are incremental work,differential of the free energy and dissipation potential increment in unit volume,respectively.The total strain tensor εij can be expressed as εij ¼εe ij þεp ij ,so we have d εij ¼d εe ij þd εp ij .For the non-coupled elastic material,the instantaneous elastic modulus is independent on the internal variables.Based on the energy decomposition principle,the free energy in unit volume can be expressed as (Collins and Houlsby,1997,Collins and Kelly,2002;Ulm and Coussy,2003):j εe ij ;εp ij ¼j e εe ij þj p εp ij (7)where j e ðεe ij Þand j p ðεp ij Þare elastic free energy and plastic free energy in unit volume,respectively.The differential of the freeenergy is given by:d j ¼v je v εe ij d εe ij þv j p v εp ij d εp ij (8)From Eqs.(6)and (8),the following formula can be obtained:s ij Àv j e v εe ij !d εe ij þs ij d εp ij Àv j p v εp ij d εp ij Àdf ¼0(9)The dissipation incremental function df is not a state function,because it only depends on the plastic strain increment.It cannot depend on the total strain increment since a purely elastic deformation would not produce dissipation.For energy dissipation of elasto-plastic deformation process in the rate independent case,the increment of dissipation function df isi et al./International Journal of Plasticity 78(2016)84e 11388first-order homogeneous function of generalized increment d εp ij (Collins and Kelly,2002).According to the Euler's theorem of homogeneous function,the following formula can be obtained:df ¼v ðdf Þv d εpij d εp ij (10)It is supposed that the internal variables εp ijare “generalized displacements ”in thermodynamics,and the “generalized force ”,conjugated work of “generalized displacements ”,is v ðdf Þv ðd εp ijÞ,which can be called dissipative stress (Ziegler,1983;Ziegler and Wehrli,1987).In dissipation stress space,the dissipation yield condition can be obtained by eliminating the component d εp ij .Substituting Eq.(10)into Eq.(9),we have:s ij Àv j ev εe ij !d εe ij þ24s ij Àv j pv εp ij Àv ðdf Þv d εp ij 35d εp ij ¼0(11)In any process of loading under isothermal condition,for any d εe ij and d εp ij ,whilst Eq.(11)is satis fied if:s ij ¼v j e v εe ij(12a)and s ij ¼v j p v εp ij þv ðdf Þv d εpij ;where df !0(12b)For isothermal condition,the relationships between the stress components and the free energy as well as dissipation potential,Eqs.(12a)and (12b),are deduced from the first law and the second law of thermodynamics.If the soil does not yield,only the elastic deformation appears,then df ¼0,the relationship between stress components and free energy can be expressed by Eq.(12a).In contrast,if the frozen soil generates plastic (or irreversible)deformation,then df >0,Eqs.(12a)and (12b)must be satis fied simultaneously during the process of loading.Eq.(12a)denotes the elastic part of constitutive law.Andin Eq.(12b),v j pv εp ij is called as shift stress,and v ðdf Þv ðd εp ijÞdissipation stress.Eq.(12b)demonstrates that the true stress is equal to shift stress plus dissipation stress (Collins and Hilder,2002),and it illustrates that the relationship between stress components and the plastic free energy as well as dissipation potential under plastic state.According to the work-energy theorem,the external work increment of geomaterial under the isothermal condition is expressed as d U ¼s ij d εij .Thus,whilst the plastic work increment is the product of the true stress with the plastic strain increment,the plastic dissipation increment is the product of the dissipative stress with the plastic strain increment.And the total plastic work increment contains the plastic dissipation increment and plastic free energy increment.Thus,the external work increment is also expressed as:d U ¼s ij d εij ¼v je v εe ij d εe ij þv j p v εp ij d εp ij þv ðdf Þv d εp ij d εp ij (13)The second law of thermodynamics requires the dissipation increment df !0,but it does not restrict the increment of shift work v j pv εp ij d εpij ,which can be positive or negative.During the process of loading,if the stress state is only elastic,or the current stress state is in elastic unloading,the deformation is elastic and there is no energy dissipation.Then,the free energy is elastic free energy,and its relation with the stress components is described by Eq.(12a).If an irreversible deformation occurs in the current stress state,the relationships between the stress components and free energy as well as dissipation potential should be described by Eqs.(12a)and (12b)simultaneously,and then the dissipation increment df is always larger than or equal to zero.For multi-yield-surface elasto-plastic problem,the total strain and total strain increment are given by:εij ¼εe ij þXn k ¼1εp k ij ;d εij ¼d εe ij þX n k ¼1d εp kij (14)where εe ij and d εe ij are elastic strain and elastic strain increment,respectively;εp k ij and d εp k ij (k ¼1,2,…,n representing the k -thyield surface)are inelastic strain and inelastic strain increment,respectively.i et al./International Journal of Plasticity 78(2016)84e 113894.Elasto-plastic constitutive model of frozen saline sandy soilFor the elasto-plastic constitutive relation of frozen saline soil,according to the continuum mechanics theory,the total strain of frozen saline soil can be decomposed into elastic strain and plastic strain,such as Eq.(14).For the deformation of geomaterial,the volumetric and shear strain components are considered.The thermodynamics results in the principal stress space in Section 3can be deduced to those in the p e q plane.In Section 4.1,the determination methods of elastic deformation law and elastic parameters are investigated by isotropic compression loading-unloading-reloading tests and triaxial shear loading-unloading-reloading tests for frozen saline sandy soil.According to the concept of critical state line,the characteristic of critical state of frozen saline sandy soil is studied in the Section 4.2.The deformation behaviors of soils are usually identi fied by two basic plastic flow mechanisms:one is related to plastic shear,and the other to plastic volumetric compression (Xie and Shao,2006).In the Section 4.3,the double yield surface constitutive model is used to study the plastic volumetric compression and shear mechanisms.In the following section,the stress state and deformation law of the frozen soil will be described by the variables in the p e q plane,i.e.,the generalized stresses p and q and the corresponding generalized strains εv ;εg ;εe v ;εe g ;εp v and εp g .4.1.Elastic deformationIn the sample preparation,water and salt were uniformly distributed in the soils,so the samples of salt soil are uniform.The soil specimens were placed in a refrigeration unit and frozen quickly at a temperature below À30 C,to ensure the uniform distribution of ice crystals and salt in the frozen soils.Under the condition of small elastic deformation,in this paper,in order to reduce the model parameters,the isotropic elastic stress e strain relationship is used in the proposed model by referencing the general practice of the anisotropic elastic e plastic constitutive model for rotational yield surface in the present studies (Wheeler et al.,2003;Yin and Chang,2010;Liu et al.,2013).Based on the test results of isotropic compression loading-unloading-reloading tests,triaxial shear loading-unloading-reloading tests and conventional triaxial tests,we studied elasto-plastic deformation mechanism of the frozen saline sandy soil under the loading condition of constant axial strain rate.In the part of elastic deformation,there is only elastic free energy j e ðεe Þproduced within materials,so that the dissipation function does not exist in the elastic deformation,which has been described in the Section 3.By using the related test results and combining with elastic free energy function,the elastic strain increment and the elastic strain can be calculated.For the elasto-plastic decoupling material,a purely elastic deformation would not produce dissipation.The elastic part of constitutive law Eq.(12a)only relates to elastic free energy,and the following differential expression can be given:d s ij ¼v 2je v εe ij v εe kl d εe kl (15a)Under the hypothesis that the elastic behavior is isotropic and independent on the third invariant,in p e q plane of general stress space,Eq.(15a)can be rewritten as:dp ¼v 2j ev Àεe v Ád εe v and dq ¼v 2j e v εe g d εe g (15b)As long as the formulation of elastic free energy in unit volume expression j e (εe )is obtained,based on Eq.(15b),the stress e strain relationship of material can be calculated from elastic free energy.In other words,the free energy also char-acterizes the elastic parameters.In p e q plane,the two main elastic parameters,bulk modulus K and shear modulus G ,can be obtained by test methods.In general,the bulk modulus K can be determined by the isotropic loading-unloading-reloading compression test,and the shear modulus G by the loading-unloading -reloading triaxial shear test.(1)The determination of bulk modulus KIn the study of nonlinear elastic model,according to the Domaschuk e Villiappan method of K e G model (Li,2004),the bulk modulus can be obtained by the isotropic compression test (q ¼0),the bulk modulus is obtained by K ¼dp /d εv ,then K ¼p a d (p /p a )/d εv ¼pd [ln (p /p a )]/d εv .p a is atmospheric pressure.In the isotropic compression tests with loading e unloading e reloading cycles,five different stress levels were used to investigate the in fluence of con fining pressures (or stress levels)on bulk modulus of frozen sandy soil.Based on the research by Lai et al.(2014),in isotropic compression tests the loading and unloading rate of 0.1MPa/min was used.When the con fining pressure reaches the preset value,it begins to unloading to 0.5MPa,and reloading at the next set con fining pressure value at the same loading rate after unloading.Based on the results of loading-unloading-reloading isotropic compressive tests under different pressures,the ln (p /p a )Àεv curve can be described in Fig.2.From Fig.2,the closed-hysteresis loops can be seen clearly.Supposing k v ¼d [ln (p /p a )]/d εv ,its physical meaning is that k v is chosen as the slope of the unloading curve or the connecting line of the hysteretic loop ends,that is to say,it is the ratio K /i et al./International Journal of Plasticity 78(2016)84e 11390。
提纯硅藻土对废水中磷的吸附张开永;黄玲【摘要】In this work ,the effects of diatomite dosage ,solutionpH ,adsorption time and temperature on the adsorption process were investigated through studying the adsorption of phosphorus by purification of diatomite in the simulated wastewater .The results showed that the optimal dosage of diatomite was 3 .5g per 0 .1 mg phosphorus and the adsorption time was about 20min .The optimal solution pH was around 3 .0 and the adsorption rate decreased with the increase of pH value ,especially in alkaline solution .The best adsorption temperature was about 25℃ .Diatomite on phosphorus removal in wastewater adsorption process is in line with the Freundlich isothermal adsorption equation .%通过提纯后的硅藻土对模拟废水中磷素的吸附研究,考察了硅藻土的用量、吸附时间、溶液pH值、吸附温度对吸附过程的影响。
研究结果表明:硅藻土的最佳用量为3.5g 硅藻土/0.1mg磷左右;最佳震荡吸附时间为20min左右;溶液的最佳pH值为3.0左右,且随着pH值的增大吸附率减小,特别在溶液呈碱性的时候吸附效果明显变差;最佳吸附温度约为25℃;硅藻土对废水中磷素的吸附过程符合Freundlich等温吸附方程。
膨润土水力分级法提纯英文回答:The hydraulic classification method for purifying bentonite involves the separation of particles based on their settling velocity in a liquid medium. This method is commonly used in the mining and construction industries to remove impurities and improve the quality of bentonite. The process involves the use of different size screens to separate the particles into different size fractions. The larger particles settle faster and are collected at the bottom, while the smaller particles remain in suspension and are collected at the top. This allows for the purification of bentonite by removing unwanted particles and achieving the desired particle size distribution.One of the advantages of the hydraulic classification method is its simplicity. It does not require complex equipment or extensive chemical treatments. The process is relatively inexpensive and can be easily scaled up forindustrial production. Additionally, the method is highly efficient in removing impurities, as it relies on the settling velocity of particles, which is influenced bytheir size and density. This allows for the removal of both larger and smaller impurities, resulting in a purified bentonite with a more uniform particle size distribution.Another advantage of the hydraulic classification method is its versatility. It can be used to purify bentonite with different initial particle sizedistributions and impurity contents. The process can be adjusted by changing the screen sizes and the liquid medium properties to achieve the desired purification level. This flexibility makes the method suitable for a wide range of applications and allows for the production of bentonite with specific properties tailored to different uses.However, there are also some limitations to the hydraulic classification method. One limitation is the potential loss of fine particles during the process. Since the smaller particles remain in suspension and arecollected at the top, there is a possibility of losing someof the finer particles, which may be desirable for certain applications. Another limitation is the requirement for a liquid medium, which adds to the overall cost and complexity of the process. Additionally, the method may not be suitable for purifying bentonite with extremely high impurity contents, as the settling velocity may not be sufficient to effectively separate all the impurities.In conclusion, the hydraulic classification method is a simple and efficient way to purify bentonite. It offers advantages such as low cost, versatility, and high impurity removal efficiency. However, it also has limitations, including the potential loss of fine particles and the requirement for a liquid medium. Overall, the method provides a reliable and effective means of purifying bentonite for various industrial applications.中文回答:膨润土水力分级法是一种基于颗粒在液体介质中的沉降速度进行分离的方法,用于提纯膨润土。
Design Research in the Netherlands75 7. Developing NPD-Process KnowledgeJan BuijsDepartment of Product Innovation & ManagementSub-Faculty of Industrial Design EngineeringDelft University of Technology7.1 IntroductionThis conference on Design Research in the Netherlands 2000 gives us a nice opportunity to show the results of design research which is being carried out at the Delft School for Product Design (officially the Sub-faculty of Industrial Design Engineering at the Delft University of Technology). Since the 1995 conference a lot has happened. In those days the Delft School of Product Design was the independent Faculty of Industrial Design Engineering. Now we have merged with the Schools of Mechanical Engineering and Naval Architecture into the new Faculty of Design, Construction and Production (DCP). The number of students and staff for product design stayed constant for all those years (ca. 100 fte staff and 1600 students). Originally we had five organisational units: four “Vakgroepen” responsible for teaching and research in the fields of respectively Construction, Ergonomics, Formgiving and Management Sciences, with one shared “Werkgroep” responsible for teaching design.Now we have three departments (“Afdelingen”), responsible only for research: Industrial Design (ID), Design Engineering (DE) and Product Innovation & Management (PI&M). All education is separately organised, headed by the Director of Education. Design teaching is an integral part of this organisation (although it is separately organised as the Institute for Design Teaching (= IvOO = Instituut voor het Ontwerp Onderwijs) and has the same budgetary status as the three research departments ID, DE and PI&M. The Department of Industrial Design is the combination of the former Ergonomics and Formgiving groups, Design Engineering comes from the former Construction group and Product Innovation & Management comes from the Management Sciences group. Design Methodology was part of the Management Sciences group and is now part of PI&M.7.2 Design researchIt could be argued that all research carried out within a school of product design is a form of design research, but that would be much too pretentious. For instance within the Department of Design Engineering research is done in the field of material sciences on plastics, and within Industrial Design researchers look at the physical limitations of elderly people in order to design better suited products for them. Within the Department of Product Innovation & Management research has been done on market introduction strategies for new products. These and other research projects are not considered as design research projects though.It would be difficult to make a sharp distinction between what is design research and what is not, especially considering the multi-disciplinary character of design itself. I will limit design research to only those research subjects that are aimed at the development of process knowledge of the New Product Development (NPD) process and not covered by other76Developing NPD-Process Knowledge traditional mono-disciplinary domains. This gap partly exists because the other disciplines are not interested in them (i.e. intuition and creativity by psychologists) or because they are unable to do it within a mono-discipline (i.e. real protocol analysis of product design projects needs both designers or engineers and psychologists).I will also limit myself to the research work of the Department of Product Innovation & Management. Others at this conference will take care of the research work that is being done in the other departments.By doing so I will not go into the research carried out within the Marketing group (a sub group of PI&M), because their research is part of the mono-discipline of marketing. Even though they have, besides marketeers, economists, psychologists, communication scientists and even product designers in their staff. I will only report about the developments within the two other groups of PI&M, the Design Methodology Group and the Management & Organisation Group.7.2.1The Design Methodology Group(Permanent research staff per May 1st 2000: ir. Norbert Roozenburg, dr. Peter Lloyd and 2 vacancies. Temporary research staff: 2 vacancies).This has been the core design research group at our school, right from its beginning in 1964. Design Methodology is one of the key elements in the curriculum of Delft School of Product Design. According to the research of Hanny de Wilde (1997), about the history and development of this school, explicit attention to design methodology was one of the key elements to start the first product design school in the Netherlands at a university level. The founder of our school, an architect called Joost van der Grinten, borrowed the ideas about design methodology both from the Royal College of Art in the UK and from the Hochschule für Gestaltung in Ulm, Germany. The work of Bruce Archer was quite influential.The graduation work of our first graduate (Norbert Roozenburg in December 1971) was about the application of a specific design method in product design. He still works at the school and is, not only, very active in the design methodology and design research field, but is also the Director of the School’s Institute for Design Teaching. He is unable to be here because he is currently guest professor at the Danish University of Technology in Copenhagen. So I will be his humble representative.The first professor in Design Methodology was Johannes Eekels (he became emeritus in 1987). Together with Norbert Roozenburg he produced numerous books and articles. The latest Dutch version of their book was published in 1998 (Roozenburg and Eekels 1998). An English version was published in 1995 (Rozenburg and Eekels 1975).Besides this traditional emphasis on the prescriptive and normative ways of designing, which is still of concern, the research in this field now also embraces empirical studies.The publication of the book on the Delft workshop on protocol analysis is a landmark in this respect (Cross, Christiaans and Dorst 1996). The workshop was organised to discuss, among leading scholars in design research, the results of different analyses from shared data.The shared data consisted of a protocol study on both individual and group design work. It was based on the same design brief. The experiment itself has taken place at Xerox PARC in California. The experimenters were Nigel Cross (at that time part-time professor in Design Methodology in Delft), Anita Cross, Henri Christiaans and Kees Dorst; the participating designers came from IDEO, the leading product design firm in the US.The workshop offered a great deal of insight into how designers actually work. At the workshop invited scholars shared their results, ideas, objections and doubts. It was interesting to watch the discussion because every attendant of the workshop had used the same originalBuijs77 data. It proved to be a very effective way of having detailed discussions about both the content of a design process as well as the way of doing protocol studies.Another interesting project of this group has been the research of Kees Dorst. This empirically based study proved that the use of different paradigms within the design research field could be used to study different aspects of design. Traditionally within the design research domain the rational problem solving paradigm, based on Herbert Simon’s ideas, is dominant (Simon 1967). Kees showed that this paradigm has its limitations, and looked for another paradigm. Donald Schön’s idea of “design as a reflective practice” proved to be this interesting other paradigm (Schön 1983). Kees showed that using both paradigms to interpret the same empirical data leads to different views and different conclusions about how designers are really working (Dorst 1997). It is my opinion that this multi-paradigmatic analysis of product design will produce more interesting results.The arrival, last year, of Peter Lloyd from the UK, an ethnographic oriented design researcher, is the next step to continue the current new stream of conducting further empirical studies.The teaching of this group is focused on a fourth year course in Design Theory and Design Methods for all our design students. Of course the group is very active in the design studio work within the “IvOO”.7.2.2The Management & Organisation Group(Permanent research staff per May 1st 2000: prof. dr. ir. Jan Buijs, ir. Frido Smulders, ir. Rianne Valkenburg, dr. Hanny de Wilde, and 2 vacancies. Temporary research staff: ir. Danielle Hendriks, ir. Remko van der Lugt, and 2 vacancies).The main objective of the Management & Organisation Group, the group I am responsible for, is to study product design processes in their natural environment, that is in the competitive situation of design projects, within companies, working together with suppliers and customers. Its focus is on design as a business activity. We usually refer to it as “design in context” or “design in business”We are looking into product design as the result of teamwork. We are interested in both the communication within the team, as well as the influence of the project leader on team behaviour. This approach looks at team behaviour not in terms of group dynamics, but in terms of design work. Of course design work and group behaviour are intertwined, but we are primarily interested in the content of the product design work.This shift from individual designers towards design groups has been caused by the very practice of industrial product development. Few product designs are the work of just one lonely designer. Nowadays complicated consumer- and industrial products are always the results of multi-disciplinary design teams.However we are not only interested in the teamwork itself, but also in the interfaces between those design teams and the rest of the organisation.We are continually conducting case studies of product development in real corporate situations. This allows us to compare empirical studies with theories of product development and has resulted in two books on Integrated New Product Development and a new course for our first year product design students (Buijs and Valkenburg 1996 and 2000).During the discussions of the aforementioned Delft workshop on analysing design activities we discovered big differences in the ways psychologists and design researchers were looking at design behaviour. For example two researchers were looking at the same type of a group design activity. Both looked at a specific action on the videotape. However the psychologist looked at body language and group dynamics, while the design researcher looked at the78Developing NPD-Process Knowledge content of the discussions within the design team. So for both there was something interesting to see, but the results were completely different. More surprisingly, some times the conclusions were completely different or even opposing.This has led to some very intriguing research projects. Helga Hohn, a psychologist, started to look at the behaviour of team leaders in helping teams with innovative tasks. She questioned more than 75 international working professionals on how they inspire their (design) teams, how they keep them on track, and how they deal with the company pressure to perform better, quicker or cheaper. Once again process and content were very closely related with “playing”proving to be very important in keeping teams alive and kicking (Hohn 1999).Rianne Valkenburg, a design researcher, is looking at team design work on the content level. She is comparing two teams of students designing during the Philips Design Competition, and two professional design teams, which took part in the earlier Delft experiment at Xerox PARC. Inspired by Kees Dorst’s work she is using Donald Schön’s paradigm to compare these different design teams. She has operationalised Schön’s theory and is heading towards some interesting conclusions about shared understanding and team communication based on the content of the design project (Valkenburg and Dorst 1998). Her thesis will be published at the end of this summer.Within this team-based research Danielle Hendriks and Hanny de Wilde are doing research about the role and influence of project leaders on the results of the product design team. Besides interviewing project leaders in Dutch design consultancies, they were also allowed to study the archives of one of the leading Dutch design firms. From a knowledge management perspective these archives have not proved useful. However, they have shown that if designers want to learn from their past they have to be more accurate in what and how to file their actual design work. Recently, an e-mail-based way of making weekly diaries has been developed. In analysing these diaries they hope to find some of the heuristics, project leaders use to solve their professional problems (Hendriks and De Wilde 1999). They are helped in this by a research student, Sjors Witjes, who is doing empirical research in cooperation with Stanford University. He is observing and interviewing project leaders of product development teams in the US high tech industry. Hopefully we can compare the results from the Netherlands with those from the US. These results will be integrated in our recently developed fourth year course on Product Development Management.In our attempts to study the real life of designers we have discovered that most designers talk about intuition as an important element in their work. Although intuition is difficult to study within the traditional way of doing scientific research, we have taken up the challenge. Robin Groeneveld has interviewed about twenty professional designers. Most of them are very explicit about the influence of intuition and about the way they can rely on it. Hopefully his PhD thesis will be published the end of this year.Finally within the Management & Organisation Group we are interested in stimulating creativity in product design. Not only have we developed a fourth year course on Creative Problem Solving (CPS), we have also started a research project in this field. Creative Problem Solving (i.e. brainstorming or synectics) is usually verbally based, while product designers tend to be visually oriented. The research project of Remko van der Lugt is trying to bridge the gap between the original CPS-rules and the more visual attitudes of product designers. The first results are promising (Van der Lugt 1998). An extended version of braindrawing, as opposed to brainstorming, seems to be an effective tool for product designers. His PhD thesis is scheduled for early next year.Beside the already mentioned courses we are also teaching a third year course on Strategy and Organisation (Frido Smulders is responsible) and we all participate in the design studio work.Buijs79 7.3 Final remarksThe research in both the Design Methodology Group and in the Management & Organisation Group is aimed at getting better insights into the process of New Product Development (NPD). With this insight we hope to improve the quality of product design work.By sharing a selection of our work with other design researchers in the Netherlands we hope to get enough energy not only to continue, but also to improve.7.4 ReferencesBuijs, J. A., and Valkenburg,R. (1996).Integrale Produktontwikkeling, Lemma,Utrecht.Buijs, J. A., and Valkenburg,R. (2000).Integrale Productontwikkeling - Tweede Geheel Herziene Druk, Lemma, Utrecht.Cross, N. G.and Christiaans,H.and Dorst, K. (1996).Analyzing Design Activities, Wiley, Chichester.Dorst, C. H.(1997).Describing Design: A Comparison of Paradigms, PhD thesis, Delft University of Technology.Hohn, H.(1999).Playing, Leadership and Team Development in Innovative Teams, PhD thesis,Delft University of Technology.Hendriks,D.and Wilde, H. de (1999).Project Management for New Product Development Projects: An Empirical Study, in: Proceedings ICED‘99,München.Lugt, R. van der and Buijs, J.A.(1998). Creative Problem Solving in Product Development: An Exploration Into the Use of CPS in Design Practice, in: Dingli, S., Creative Thinking, Towards Broader Horizons, Malta University Press.Roozenburg N. F. M. and Eekels, J. (1995).Product Design: Fundamentals and Methods, Wiley,Chichester. Roozenburg,N. F. M. and Eekels, J.(1998).Produktontwerpen, tweede druk, Lemma, Utrecht.Schön, D. A.(1983).The Reflective Practitioner, Basic Books,New York, 1983.Simon, H.A. (1967).Sciences of the Artificial, The MIT Press, Cambridge MA.Valkenburg, R. and Dorst,K.(1998).The Reflective Pratice of Design Teams, in:Design Studies,19, pp.249-271.Valkenburg, R. (2000).The Reflective Pratice of Product Design Teams. PhD thesis Delft University of Technology, forthcoming in2000.Wilde,H. de (1997).Passie Voor Productontwikkeling, Lemma,Utrecht.。
Increase your BIM workflow productivity and quality using Vault custom applicationsDarius ŠimkūnasBIM & Application strategist –collaboration platformsMagnus SvenssonBIM & Application strategist –Autodesk CivilAbout the speakerDarius SimkunasDarius Šimkūnas is the BIM & Application strategist at Tyréns AB. He focuses on collaboration platforms and Autodesk Civil products. Darius has over 15 years of experience with Autodesk, Inc. products, consulting and training employees. He was an Autodesk University speaker in 2013, 2014, 2015 and 2016. He has been working with AutoCAD Civil 3D software since 2005 and has a portfolio of functionalities for specific regional and company needs. He is an AutoCAD Civil 3D Certified Professional, has been a member of the Autodesk Developer Network (ADN) since 2008, and he has been working on infrastructure projects using new technology and methodology. Since 2012 Darius has been responsible for implementation of Vault software. As a project manager, he directly participates in the stages of implementation, resolution of strategic and technical issues, and product functionality enhancement according to the company's needs.About the speakerMagnus SvenssonMagnus Svensson is BIM and application strategist with focus on Autodesk civil products. He has been working on the development of functionalities and working methods in AutoCAD Civil 3D and Infraworks, holding educations and acting as BIM and application strategist and advisor in large infrastructure projects. Magnus is engaged in Swedish and international network for developing Civil 3D and Infraworks. He has special skills in parametric sections for corridor design in Civil 3D and 25 years of experience from different areas of civil design and IT management in large organizations.Agenda•About Tyréns•Introduction to a future standard of a Vault project•VDC/ICE/BIM workflow optimization –real project example •Sharing Vault outside the company network•Expanding Vault possibilities with API•Integrating Vault with other systems, ensuring project quality •Automating processes using Job servers•Automate automated processes using schedulers •QuestionsGeneral information on Tyréns Founded in 1940.One of Sweden's leading multi-disciplinary consultanciesSpecialists in urban planning and infrastructure2,700employees across Sweden, Denmark, the UK, Lithuania and Estonia Offering custom solutions for sustainable developmentPart of the international network SEEN with 6,000 expertsMost frequently used design software:•Autodesk AutoCAD, Civil 3D, MEP•Autodesk Revit•Autodesk Navisworks•Tekla•Bentley Microstation,Power Civil / Rail Track, PW•Bentley Navigator•Esri ArcGIS•Autodesk Vault Pro•…How Tyréns AB optimizes VDC/ICE/BIM processes with Vault API and integrations with other applications –a real project example!Bypass Södertörn•Vault use since early-2016•Stage–preliminary design•New road length –20 km (12.5 miles)•3+3 lanes highway• 3 main tunnels, length –over 5 km (3 miles)•Up to 100 new bridges and overpasses•About 25different disciplines involved in the project•About 400engineers from different disciplines working on the project •About 15different subcontractors using Vault live from Sweden and abroad •Customers use Vault for documents and model review.•…Bypass Södertörn statistic in Vault•~17,000 files.•~120 GB, latest versions only.•~4,000 DWG files including 2,500models.•~4,000Civil Data references includingover 2,000shared surfaces•VDC/ICE working process •ICE meetings and coordination models update once a week•Model base delivery with drawings •Classification system demands•Full team/project transparency for the client:o access to the live modelo time reportso progress stageo meeting notes, etc.•The client is part of the team with a special role.Project demandsOnce a week!Once a week!Improving BIM workflow. How to make a BIM model live?What we need to get a live BIM model? Data/Modelmanagement•Single source of data24/7Cross-discipline collaboration •Effective collaboration between disciplinesClassification system •Clear and easy to use automated controlProcess automation •Automate processes, avoid errors, ensure data relevancyCODESData management –the entire project in a single place*BIM model location –VaultBIM model*ProjectOffice 1Office 2Office 3CustomerDMZPartnersDMZSharing data with external partners and the client•ADMS server on DMZ allows sharing Vault using the https protocol for external users in a secure way•External partners work with Vault the same way as internal users•We are using Vault to share templates, configurations files, etc.•The Client can access the live model 24/7.Coordination model –Navis worksSource of informationCoordinationmodelLink to live modelAutoCovertManualCovertCoordination model is not the final result, but a live source of information to make decisions and progress•Main source of information for liveCivil 3D surfaces, alignments,corridors, pipe networks.•CAD 3D objects•CAD 2D objectsWhat we need to have a live BIM model?Datamanagement •Single source of data24/7Cross-discipline collaboration •Effective collaboration between disciplinesClassification system •Clear and easy to use automated controlProcess automation •Automate processes, avoid errors, ensure data relevancyCODESCross-discipline collaboration –Level 1 RoadsBridges Pipes Landscapes TunnelsCivil 3Dmodel filesEmpty file withSurface DrefsEmpty file withCorridor Drefs Surface DREFAlignment DREFPipe network DREFCorridor DREFBIM workflowExisting GroundsurfaceRock surfacePipe networkRoad AlignmentProfileRoad Corridor,Top, BottomsurfacesBridge Corridor,Top, Bottom surfacesRoad Corridors DREF’s fileBridge Corridors DREF’s file Road Surfaces DREF’s fileBridge SurfacesDREF’s fileCoordination modelDrawings, PDFDREFDREF DREFDREFDREFDREFDREFDREFDREF DREFDREFExport DataLive linkMain alignment file is used in 4,747files/objectsExisting Ground or Rock surface input dataLaser scanMeasuredinformationDrill, scaninformationAutomatic Civil 3Dsurface creationCivil 3D Surface Live Model collaborationCivil 3D modelsOthersoftwareprovidermodels2D CAD objects(lines, polylines)3D CAD objects(3D polylines,Mesh, solids,surfaces)XML, IFCAutomatic Civil3D models exportWhat we need to have a live BIM model?Datamanagement •Single source of data24/7Cross-discipline collaboration •Effective collaboration between disciplinesClassification system •Clear and easy to use automated controlProcess automation •Automate processes, avoid errors, ensure data relevancyCODESClassification system –CodesSQL database API engines allow integrations from different softwareAPIInput information fromdifferent sources (.csv)Classification system and ICE meetingsK-32CD--EZN1--14G-01How to explain what it is to everybody in the meeting room?Classification system and TyrEngineTyrEngine is a unique platform forvisualizing sustainable spatialplanning. By combining advancedgraphics techniques from theworld of gaming, a 3Denvironment is produced thatdemonstrates the results andimpacts of projects in spatialplanning, property and industrialdevelopments.What we need to have a live BIM model?Datamanagement •Single source of data24/7Cross-discipline collaboration •Effective collaboration between disciplinesClassification system •Clear and easy to use, automated controlProcess automation •Automate processes, avoid errors, ensure data relevancyCODESExpand Vault possibilities with APIVault API possibilities1. Vault client functionality •Additional functionalities of Vault client (properties tab, copy attribute values, etc.)2. Event handler •Automatically run functionalities on specific events(Check in, Move, delete, etc.)3.Vault + Civil3D,Excel, etc.•Vault initiates third applications and runs commands4. Job server•Run user initiated functions on a Job server(publish to PDF, Export C3D file to AutoCAD, etc.)5. Scheduledjobs •Scheduler runs jobs automatically without user inputVault client functionality•Smart properties tabo File name was encoded inCodes and values checkedo Using a special option valueswill be updated by CodesstandardVault client functionality•Update multiple file propertieso You can select one file as a baseand copy values to other fileso If the file name conforms to theClassification system standards,encoded attribute values will beupdated by CodesVault client functionality•Deleting multiple folderso You can delete more than onefolder at onceo Important! Do not give regularusers access to this function.Administrators only!o Ask for approval to deletedisplayed folders.Vault client functionality•Visual life cycle•Special thanks to DOUG REDMOND•Code example can befound athttp://justonesandzeros•New file from a templateo Select a template stored in theVaulto Create a file name by selectingvalues from the list (values from Codes database)o Attributes will by filled inautomatically•Safe deletion of files/folderso Regular users do not have the right todelete fileso Using a dedicated command files arerenamed to _delete by an admin usero If the file is used, an error messageinformation will be displayed.o_delete files are downloaded and backed up before automatically deleted fromVault.•Copying files with attribute values (taking values from a source file or a destination folder)•Getting user information about who is using your file even if you do not have access to that file •Write an e-mail directly•Creating users in Vault even you are not an administratoro Select an internal user or create an external oneo Select groups to addo Select projects to addo No possibilities to add to Admin groups and other restrictionsVault API possibilities1. Vault client functionality •Additional functions to Vault client (properties tab, copy attribute values, etc.)2. Event handler •Automatically run functions on specific event (Check in, Move, delete, etc.)3.Vault + Civil3D,Excel, etc.•Vault initiates third applications and runs commands.4. Job server•Run user initiated functions on a Job server(publish to PDF, Export C3D file to AutoCAD, etc.)5. Scheduledjobs •Scheduler runs jobs automatically without user input.•Folder creation controlo Fixed/controlled folderstructureo Regular users can createfolders only in _Tempo Illegal actions will beforbidden•Adding new fileso Set a parent folder category toa file automaticallyo Copy parent folder propertyvalues to a file automatically Folder category Properties values•Checking-in fileso Inform users that CAD filesmust be checked-in with CAD application integration.o Store illegal check-ins indatabase for administratoranalysis•There is a list of events to catch•What you need to Automate?Process automatization -Vault API1. Vault client functionality •Additional functions to Vault client (properties tab, copy attribute values, etc.)2. Event handler •Automatically run functions on specific event (Check in, Move, delete, etc.)3.Vault + Civil3D,Excel, etc.•Vault initiates third applications and runs commands.4. Job server•Run user initiated functions on a Job server(publish to PDF, Export C3D file to AutoCAD, etc.)5. Scheduledjobs •Scheduler runs jobs automatically without user input.Download thefile with all dependencies to local cache Open the filewithapplicationRuncommandAdd/update thefile in the Vault1.Make sure you can run your commands without any user input2.If the function does not exist or needs user input, write it using the software’s API3.Run function manually first4.Run function in the Vault workflow•Publish dwg to nwc.1.Download files withdependencies to C drive2.Check out nwc files3.Open files with AutoCADAccoreconsole4.Run NWCOUT5.Check in nwc files•Create Civil 3D surfaces from 3D faces •Surveyors create surface in other applications and export 3D faces.•One region can have about 500,000 3D faces. If you try to select all those 3D faces in Civil 3D it can take a very long time of crash software.•This custom application adds 3D faces in portions (100,000) using Accoreconsole and a Job server.•It takes a couple of minutes.Vault integrations with other applicationsCreate Civil 3D surface from 3D faces (514828)Create surface On Job server -7 minutes Create surface On Job server -50 minutes•Publishing Civil 3D to AutoCAD•Publishing DWG to PDF •Cleaning DWG files (Audit, Purge, Purge reg. apps., detach not found Xrefs)•Other functions depending of project needs Run Accoreconsole from Vault APIRun script -Load .dll and run command Write command with Civil 3D APIAutomated process Manual processVault integrations with other applications•Publishing nwf to nwd1.Download nwf with all dependencies to alocal drive2.Open the nwf file with Navis works3.Save the file as nwd4.Add/Check in the file to Vault nwfnwd。
第3期蒙脱石在胡敏酸存在时去除土壤中重金属的研究综述29李丽,刘中,宁阳,等.不同类型粘土矿物对镉吸附与解吸行为的研究山西农业大学学报(自然科学版)"017"7(01):60—66.王耀龙,邓红梅,吴海宏,等.钠基蒙脱石吸附TI(I)的性能研究[C).//中国矿物岩石地球化学学会.中国矿物岩石地球化学学会第14届学术年会论文摘要专辑.贵阳市:中国矿物岩石地球化学学会"013:2.韩桂洪.腐植酸基铁矿球团粘结剂的构效关系研究[D).长沙:中学2013谢志磊.乌梁素海沉积物中胡敏酸和胡敏素与重金属相互作用机理研究[D).呼和浩特:内蒙古大学"015.[10)Maryam Haghighi,Mohsen Kafi,Amir Khoshgoftarmanesh. EFFECT OF HUMIC ACID APPLICATION ON CADMIUM ACCUMULATION BY LETTUCE LEAVES:J).Journal of Plant Nu-trition201336:10[11)Aiguo Liu,Richard D Gonzalez.Adsorption/Desorption in a System Consisting of Humic Acid"Heavy Metals"andClay Miner-als[J).Journal of Colloid And Interface Science,1999,218:1.(2)滕飞,李福春,吴志强,等.高岭石和蒙脱石吸附胡敏酸的对比研究[J).中国地质,2009,36(04):892—89&(3)何为红.重金属离子在粘土矿物-胡敏酸复合体上的吸附研究[D)南京:南京农业大学,2007.(4)吴宏海"长秋云,方建章,等.高岭石和硅/铝-氧化物对腐殖酸的吸附实验研究岩石矿物学杂志200302:173—176.[15)Brindley,G.Formation and properties of clay-polymer comple-xes[J).Earth-Science Reviews,1979,16(15):295―296.[16)LinDaohui"Tianxiaoli"LiTingtingetal Surface-boundhumic acid increased Pb2+sorption on carbon nanotubes[J)Environmen-talpo l ution2012167:138—147A Review on the Removal of Heavy Metals fromSoil by Montmorillonite in the Presence of Humic AcidXue Zhe,Wan Juan,Chen Wenqing{College of Architecture and Environment,Sichuan University,Chengdu610065,Sichuan,Chian) Abs"rac":Usingadsorbentstocontrolthe migrationandtransformationofheavy metalsincontaminated soilisacommonlyused method"whichhashighe f iciency"lowcostandsimpleoperation Asakindof claymineral"montmori l onitehasgoodadsorptioncapacityandstablephysicalandchemicalproperties It isthemostwidelyusedheavy metalfixationadsorbentinsoilenvironment However"thehumicacidin thesoilwi l a f ecttheadsorptionperformanceofthemontmori l oniteadsorbentInthispaper"theresearch and application of montmori l onite and humic acid adsorbing heavy metals are introduced"which can pro-videreferenceforremediationofheavymetalpo l utionKey words:montmorillonite;humic acid;heavy metal;adsorbent中化涪陵化工建设20万吨/年合成氨等环保搬迁项目中化涪陵化工有限公司环保搬迁项目位于重庆涪陵白涛化工园区哨楼村一组。
越南汪秘电厂单位工程ITP检测与计划(中英文对照)SubjectCast-In-Place Concrete Piling (Installation)Sub contrac tor CHD UPC/UPC's Consultant 1.0Setting out/surveySurvey/drawings Survey recordP S R 2.0Piling equipment preparation VisualInspection form (Form No. 631-24-0001-01)P R S 3.0BoringWork method statementP SW R Position and alignment checkInspection form (Form No. 631-24-0001-01)P SW SW Underground servicesP W W Surface casing size & condition P W W First casing installationP W W Correct borehole size & side P W W VerticalityP W W Length of boreholeP W W Cleanliness of bottom boreholeP W W Collection for soil sample (if required)P W W T emporary casing size & condition (if any) P W W Bored Soil RemovalP W W Equipment Specification Requirement P W W Final Site ClearingP W W 4.0ReinforcementWork method statement Inspection form (Form No. 631-24-0001-02)4.1MaterialStandard Lab tests Material certificateH R W 4.2P W S 4.3P W S 4.4P W W 4.5P W S 4.6P W S 4.7PW S 4.8P W R 4.9P W R P W W 4.10P W W 4.11P W W 4.12P W W 5.0RecordInspection form (Form No. 631-24-0001-03)Date P W S TimeP W S Bored Depth P W W Soil Description P W R Bentonite checkedPW R 6.0Concrete6.1Material Standard Lab testsMaterial certificateH R W 6.2SupplyWork method statementInspection form (Form No. 631-24-0001-04)P SW SW Concete Class / Grade P W R Mix Design No.P W R Attached conc. delivery slips P W S Discharge Time P W S Concrete Quantity P W S Placing Method P W S Batch Time PWSProcess Description or QC activityRef. No.Number & length tube testing (if any)The top elevation of the re-bar cage Embeded method_______________Quality embeddingInspection/TestMethodRemarksRecordApproval/Inspection by Correct re-bar size, type & length Number of re-bars usedCorrect diameter, length & number of cage Re-bar spacing Correct spacer for cover & sufficient numbers Overhang of cage______cm Embeded length______cmCleanliness of re-barSubjectCast-In-Place Concrete Piling (Installation)Sub contractor CHD UPC/UPC's Consultant Process Description or QC activityRef. No.Inspection/TestMethodRemarksRecordApproval/Inspection by。
Catalytic decomposition of biomass tars:useof dolomite and untreated olivineLopamudra Devi a,*,Krzysztof J.Ptasinski a ,Frans J.J.G.Janssen a ,Sander V.B.van Paasen b ,Patrick C.A.Bergman b ,Jacob H.A.Kiel baEnvironmental Engineering Group,Faculty of Chemical Engineering and Chemistry,Eindhoven University of Technology,Den Dolech 2,P.O.Box 513,5600MB Eindhoven,The Netherlands b Energy research Centre of The Netherlands (ECN),P.O.Box 1,1755ZG Petten,The NetherlandsReceived 21January 2004;accepted 12July 2004Available online 16September 2004AbstractAlthough biomass is getting increased attention as a renewable energy source,one of the remaining problems still to be solved is the reduction of the high level of tar present in the product gas from gasification of biomass.The purpose of the present work is to study the activity of olivine and dolomite for tar destruction.Some researchers investigated olivine as bed material for biomass gasification.But it is not yet known how tars behave in the presence of olivine and whether olivine has some activity towards tar destruction.A slipstream from a lab-scale atmospheric bubbling-fluidised-bed gasifier (located at ECN)is passed through a secondary fixed-bed reactor where the additives are placed.For easy understanding,the results are represented in terms of the following tar classes;GC-undetectable tars (class 1),heterocyclic compounds (class 2),aromatic compounds (class 3),light polyaromatic compounds (class 4),heavy polyaromatic compounds (class 5).The general observation is that the conversion of all tar classes increases as the temperature was raised from 800to 9008C for both additives.The water-soluble heterocyclic compounds can be easily converted by thermal treatment.At the temperature of 9008C,the water-soluble heterocyclic compounds are completely converted.A 48%decrease in heavy PAHs is observed with pure sand.Addition of 17wt%olivine to the sand leads to a 71%decrease of PAHs at 9008C,whereas addition of 17wt%(pre-calcined)dolomite converted 90%.Also improvement in conversion of other tar classes is observed when olivine and dolomite are added during hot gas cleaning.A total tar amount of 4.0g m 0K 3could be reduced to 1.5and 2.2g m 0K 3using dolomite and olivine,0960-1481/$-see front matter q 2004Elsevier Ltd.All rights reserved.doi:10.1016/j.renene.2004.07.014Renewable Energy 30(2005)565–/locate/renene*Corresponding author.Tel.:C 31-40-2473734;fax:C 31-40-2446653.E-mail address:l.devi@tue.nl (L.Devi).respectively,at a temperature of 9008C.Inspite of this reduction in total tar concentration,a limited impact on the tar dewpoint is observed.q 2004Elsevier Ltd.All rights reserved.Keywords:Biomass tar;Tar classification;Dolomite;Olivine1.IntroductionIn view of the worldwide concern about the depletion of fossil fuels and environmental problems associated with the use of these sources,renewable energy sources are getting increased attention.Biomass is considered as a potential source of renewable energy.Among all biomass conversion processes,gasification is one of the promising ones.One of the remaining problems still to be solved is the reduction of the tar present in the product gas.Tar is undesirable because of various problems associated with condensation,formation of tar aerosols and polymerisation to form more complex structures,which cause problems in the process equipment as well as the engines and turbines used in application of the producer gas.However,the minimum allowable limit for tar is highly dependent on the kind of process and the end user application.Considerable efforts have been directed towards tar removal from syn-gas.Catalytic tar removal from the syngas is extensively reported in the literature.These catalysts include Ni-based catalysts,calcined dolomites and magnesites,zeolites and iron catalysts.The presence of additives not only influences the gas composition,but also the heating value of the product gas.The use of catalytically active materials during biomass gasification promotes the char gasification,changes the product gas composition and reduces the tar yield.Besides these,addition of active bed materials also prevents agglomeration tendencies and subsequent choking of the bed.Dolomite and Ni-based steam reforming catalysts have been proven to be very active in terms of tar reduction.A lot is known on the behaviour of dolomite with respect to tar cracking.It has been tested as a primary [1–3]as well as secondary [4,5]measure,and it has been proven to be active in reducing tars produced in biomass gasifiers.However,besides good activity in terms of tar reduction,it has some critical limitations if used inside the gasifier.Dolomite is softer and thus gets eroded.Also,some dolomite particles break during the calcination and give rise to a large production of fines,leading to an increased carry over of solids from the bed.Several nickel-based catalysts have been investigated and found to be very effective in terms of tar removal [3,6,7].Ni-based catalysts are also very effective for NH 3removal.Wang et al.[8]reported that NH 3can be decomposed effectively over a Ni-based catalyst above 8008C.Wang and his coworkers [9]reported 95%conversion of NH 3along with 89%conversion of light hydrocarbons,which they defined as C 2H 6,C 6H 6,C 7H 8and C 9H 8,over Ni-based catalyst in a secondary reactor at a temperature of 8748C,pressure of 12bar and a space time of 3s.The main limitation of using Ni-based catalysts is severe deactivation of the catalyst.This deactivation occurs mainly when the catalyst is placed right after the gasifier;the high tar concentration has a devastating effect on catalyst activity.Steam reforming catalysts can however be usedL.Devi et al./Renewable Energy 30(2005)565–587566for polishing purposes,when very clean gas is needed in,for example,Fischer–Tropsch processes.An alternative of dolomite can be naturally occurring particles of olivine,which is a mineral containing magnesium oxide,iron oxide and silica.Olivine is advantageous in termsof its attrition resistance over that of dolomite.Rapagna`et al.[10]investigated the catalytic activity of olivine and observed that it has a good performance in terms of tar reduction and the activity is comparable to calcined dolomite.They reported more than 90%reduction inaverage tar content,the tar amounted 2.4g m 0K 3compared to 43g m 0K 3with only sand.Theauthors also performed experiments with olivine as the bed material and a La–Ni–Fe trimetallic perovskite catalyst in a secondary reactor.The combined action of the materials was very promising;a gas with around 0.3g m 0K 3of tar was produced [11].Rose´n et al.[12]reported the use of olivine as a bed material for pressurised gasification (0.4–1.0MPa)of birch.Looking into the initial investigations on olivine,olivine can be a prospective candidate as an in-bed additive.However,Abu El-Rub et al.[13]used olivine to decompose naphthalene as a model tar compound and observed no significant activity of olivine.Therefore,there is still ambiguity on the prospective use of olivine as a tar decomposing catalyst.It is not yet well known how tars behave in the presence of olivine and hence more attention should be given to find out whether olivine could produce a clean gas with very low tar content.The focus of this paper is to determine the influence of olivine and to predict the decomposition behaviour of tar.Although a lot is known from the literature about dolomite,experiments are also performed with calcined dolomite for the purpose of comparison.A new approach of tar classification is described and followed throughout this paper.Besides the amount of tars present in the syn-gas,the composition of tar is very crucial as it determines the condensation behaviour.Since the dewpointof L.Devi et al./Renewable Energy 30(2005)565–587567the tars present in the syn-gas is very important for many end user applications,tar dewpoints are calculated after catalytic treatment.2.Tar classificationTar is a complex mixture of condensable hydrocarbons,which includes single ring to multiple ring aromatic compounds along with other oxygen containing hydrocarbons and complex polycyclic aromatic hydrocarbons.Tar is normally considered as a single lump of hydrocarbons.Considerable efforts have been directed towards identifying all the constituent components of tar and the inter-connection between them.Several researchers tried to put tars in different classes and study the behaviour among these classes.Different approaches of tar classification are reported in the ne et al.[14]classified tars in four different groups depending on the reaction regimes as shown in Fig.1.These four groups are:‘primary products’which are characterised by cellulose-derived,hemicellulose-derived and lignin-derived products;‘secondary products’which are characterised by phenolics and olefins;‘alkyl tertiary products’which are mainly methyl derivatives of aromatic compounds;‘condensed tertiary products’which are PAHs without substituent groups.Primary products are destroyed before the tertiary products appear [14].Corella et al.[15]grouped all the tar speciesinFig.1.Tar classes as reported by Milne et al.[14].L.Devi et al./Renewable Energy 30(2005)565–587568six lumps,namely benzene,1-ring compound,naphthalene,2-ring compounds,3-&4-ring compounds and phenolic compounds.The authors also proposed a reaction network suggesting the inter-dependency of each lump[15].In another approach Perez et al.and Corella et al.[4,16]mentioned tars in two groups,namely‘easy to destroy’which are characterised by more reactive tar species and‘hard to destroy’which are mainly less reactive tar species.In this present work,tar is considered as all organic contaminants with a molecular weight larger than benzene.For easy understanding,the results are represented in terms of the tar classes presented in Table1.This tar classification system has been developed in cooperation with Energy research Center of The Netherlands(ECN), Toegepast Natuurwetenschappelijk Onderzoek(TNO)and University of Twente(UT) within the framework of the project‘Primary measures for the inhibition/reduction of tars in biomass fuelledfluidised-bed gasifiers’,funded by the Dutch Agency for Research in Sustainable Energy(SDE).This classification is mainly based on solubility and condensability of different tar compounds,rather than reactivity of the compounds. At this point,it should be clearly mentioned that the objective of this present paper is tofind out decomposition of different tar classes.Inter-relation and inter-dependency between the tar classes is not taken into account and this is out of the scope of the present paper.Table1List of tar compounds that are considered for different tar classesTar class Class name Property Representative compounds Compounds considered inthis study1GC-unde-tactable Very heavy tars,cannot bedetected by GCNone None2Heterocyclic Tars containing heteroatoms;highly water sol-uble compounds Pyridine,phenol,cresols,quinoline,isoquinoline,dibenzophenolBenzonitril,phenol,quinoline3Lightaromatic Usually light hydrocar-bons with single ring;donot pose a problemregarding condensabilityand solubilityToluene,ethylbenzene,xylenes,styreneStyrene4Light poly-aromatic Two and three ring com-pounds;condense at lowtemperature even at verylow concentrationIndene,naphthalene,methylnaphthalene,biphenyl,acenaphthalene,fluorene,phenanthrene,anthraceneIndene,naphthalene,1-methyl naphthalene,2-methyl naphthalene,bi-phenyl,acenaphthalene,fluorene,phenanthrene,anthracene5Heavy poly-aromatic Larger than three-rings,these components con-dense at high temperaturesat low concentrationsFluoranthene,pyrene,chrysene,perylene,coroneneFluoranthene,pyrene,triphenylene,benzo(a)anthracene,benzo(c)phenanthrene,benzo(e)pyrene,benzo(j)fluoranthene,benzo(k)fluoranthene L.Devi et al./Renewable Energy30(2005)565–5875693.Experimental3.1.SetupThe experiments have been conducted at the facility of ECN,Petten,The Netherlands.A slipstream of the biomass gasification gas from the lab-scale atmospheric bubbling fluidised bed gasifier was passed through a secondary fixed bed reactor where the additives were placed.The gasifier operates with a capacity of 1kg biomass h K 1.The schematic representation of the experimental setup is given in Fig.2.The stainless steel secondary reactor is of 3cm internal diameter and 50cm length.For preheating the gasifier gas,a standard tube with a 6mm outside and 4mm inside diameter is wounded around the reactor.Temperature sensors are installed at the gas entrance and exit points and in the middle of the catalyst bed to ensure that the reactions occurred at the desired temperatures.Experiments have been performed in a temperature range of 800–9008C.The temperature across the catalyst bed can be considered constant during each experiment.The position of the bed was chosen from the temperature profile over the reactor,which was determined with the oven temperature set to 10108C and a gas flow of 1l min K 1.The catalysts were mixed with sand and the mixture was placed on top of a bed of coarse sand.Temperature readings were taken from the bottom of the sand bed to top.Fig.3shows the profile of the 10cm above the sand bed.Prior to experiments dolomite was calcined at 9008C for 1h under nitrogen flow.No pre-treatment was carried out for the olivine except sieving for the right size fraction.Typical experimental conditions are given in Table 2.3.2.Sampling and analysisThe inlet and outlet tar compositions were sampled using the SPA (solid phase adsorption)method developed by KTH,Sweden [17].The SPA samples were takenatFig.2.Schematic representation of the experimental setup for tar decomposition.L.Devi et al./Renewable Energy 30(2005)565–587570Fig.3.Temperature profile over the bed.Table 2Experimental conditionsConditionParameter Gasification conditionsBiomass feedBeech Gasification temperature (8C)850Bed materialSand Feed rate (kg biomass h K 1)1Gas composition of the syn-gas (vol%)H 210.9CO15.7CO 216.5CH 44.7Conditions of the secondary reactorTemperature (8C)800–900Catalyst particle size (m m)250–355Mass fraction of catalysts mixed with sand0.17Bed material (catalysts)Dolomite/sand Olivine/sand Amount of catalysts (g)4.88 4.75Average gas flow through the reactor (l min K 1)1.1 1.2Residence time (s)0.300.26Space time (kg cat h m 0K 3)7.4!10K 2 6.6!10K 2Pre-treatment Calcination NoneL.Devi et al./Renewable Energy 30(2005)565–587571a temperature of 3008C.The SPA samples were analysed by a SHIMADZU QP5000GCMS with a WCOT fused silica column.A typical chromatogram of the inlet tar is shown in Fig.4.Several samples were taken at the same operating condition and the average value is presented here.Concentrations of individual tar compounds were calculated in mg m 0K 3and then added to get the concentration of a particular tar class.The tar compounds that were considered for a particular class are tabulated in Table 1.Concentrations of compounds with a higher boiling point than pyrene were determined using the calibration data of pyrene.The heaviest compound identified was benzofluoranthene.Benzene is not considered as tar.Light tar,e.g.toluene,was identified,but could not be measured quantitatively accurately,especially at lower temperature.So toluene was not taken into account during total tar calculation,but presented separately.Very high molecular weight (class 1)tars and final gas composition could not be measured due to experimental limitations.3.3.Catalyst characterisationTwo types of additives,olivine and dolomite were tested during these experiments.Calcined dolomite is a porous catalyst;its large (internal)surface area and the presence of oxides in its matrix (CaO,MgO)make it an active catalyst with respect to tar reduction.Olivineisanaturallyoccurringsilicatemineralinwhichmagnesiumandironareembeddedin the silicate tetrahedral [10].The properties of both the additives are tabulated in Table 3.The BET surface area has been measured by chemisorption with ASAP.Olivine is a nonporous material as it has an extremely low surface area.Results of mercury porosimetry for calcined dolomite are included in Table 3as well,for olivine mercury porosimetry could not bedoneFig.4.Typical chromatogram found with GC/MS analysis (inlet tar concentration).L.Devi et al./Renewable Energy 30(2005)565–587572because of negligible pore volume.The Scanning Electron Microscopy scans for calcined dolomite and olivine are shown in Fig.5.X-ray diffraction (XRD)reveals that olivine has an orthorhombic structure whereas calcined dolomite has a simple cubic structure.Figs.6and 7give the typical X-ray diffractogram for calcined dolomite and olivine.4.Results and discussionTar decomposition occurs due to a series of complex,multiple and simultaneous reactions.The main reactions during catalytic treatment are shown below.(a)Thermal cracking:p C n H x /q C m H y C r H 2(b)Steam reforming:C n H x C n H 2O /(n C x /2)H 2C n COTable 3Catalyst propertiesProperties Dolomite Olivine Chemical composition (wt%)aCaO 31.5–MgO 20.349SiO 20.341Fe 2O 30.47Al 2O 30.10.5Cr 2O 3–0.3NiO –0.3Physical propertiesColour Greyish GreenishDensity a – 3.3g cm K 3bHardness a Soft 6.5–7.0Moh’s scale bMelting point a na Approx.17608C bThermal expansion a na Linear;approx.1.1%upto 12008C b Mercury porosimetry (with Micromeritics instrument,Auto pore IV 9500)Total pore area (m 2g K 1) 5.048ndTotal pore volume (ml g K 1)0.0788nd Average pore diameter (4V/A)(nm)62.4ndBulk density 2.5454ndApparent (skeletal)density 5.3543ndNitrogen adsorption (with Micrometitics ASAP 2000instrument)BET surface area (m 2g K 1)9.11780.42X-ray diffraction (determined withRigaku Geigerflex D/MAX-B XRDequipment)Structure Cubic OrthorhombicState of material Calcium oxide (CaO),magnesium oxide (MgO)Ferroan-forsterite (Mg,Fe)2SiO 4,syn-forsterite (Mg 2SiO 4)na:not available,nd:could not be determined due to extremely low porosity.a As indicated by the supplier.b ;July,2003.L.Devi et al./Renewable Energy 30(2005)565–587573Fig.5.Scanning Electron Microscope scans for olivine and calcined dolomite.L.Devi et al./Renewable Energy 30(2005)565–587574Fig.6.X-ray diffraction pattern forolivine.Fig.7.X-ray diffraction pattern for calcined dolomite.L.Devi et al./Renewable Energy 30(2005)565–587575(c)Dry reforming:C n H x C n CO 2/(x /2)H 2C 2n CO (d)Carbon formation:C n H x /n C C x /2H 2(e)Water gas shift:CO 2C H 2/CO C H 2O(f)Boudouard reaction:C C CO 2/2COC n H x represents tar,which can be a mixture of several individual tar compounds.Besides these reactions,different tars can also react with each other by severalpossibilities.Due to these complexities,it becomes difficult to understand the exactbehaviour of the catalyst towards tar removal.The simplest way to measure activity is interms of conversion of tar in a particular condition.The results are expressed in terms ofconversion of individual tar classes defined as:X ð%ÞZ ðC t ;in K C t ;out ÞC t ;in !100%(1)Once the concentration of each class is known before and after the catalytic reactor,conversion for each class is calculated using Eq.(1).The conversions of different tarclasses using calcined dolomite and olivine as catalysts as a function of temperature areshown in Figs.8and 9.The results include the thermal effect besides the catalytic effect ontar decomposition.Temperature plays a very crucial role in decomposing tars.Conversionof each tar class increases when the temperature is raised from 800to 9008C.Attemperature of 8008C,production of class 3tars is observed leading to anegativeFig.8.Conversion of different tar classes with dolomite as catalyst.L.Devi et al./Renewable Energy 30(2005)565–587576conversion over dolomite.Probably,this is mainly due to the fact that higher hydrocarbonsare breaking down forming class 3tars.At 850and 9008C,also decomposition of class 3tars is observed.Olivine shows a moderate activity in terms of tar reduction and onlybecomes active at higher temperature.At 9008C,olivine shows considerable activitytowards decomposition of all the four tar classes.Heterocyclic (class 2)tars are very easilyconverted by both dolomite and olivine.The comparison of dolomite and olivine is evident from Fig.10,which represents theconversion of different tar classes at 9008C.In view of better comparison,an experimentwith sand only has been conducted under the same operating conditions.It is observed thatat a temperature of 9008C,the water-soluble heterocyclic compounds are completelyconverted for all the three cases.A 48%decrease in heavy PAHs (class 5)is observed withsand.The addition of olivine leads to a 71%decrease of the total heavy PAHs,which is animprovement of around 50%.The addition of calcined dolomite causes a decrease ofalmost 90%at the same operating conditions.The light PAHs (class 4)also show aconsiderable increase in decomposition when olivine is added.The decrease in tar whenonly sand is used amounts to 25%whereas the decrease is 54%when olivine is added.Again adding calcined dolomite increases the conversion the most;the decrease in lightPAHs amounts to 56%.In case of light aromatic compounds (class 3),olivine showsa slight increase in conversion when compared to sand.The conversion when dolomite isused is 71%,whereas the sand-only case yields 48%;olivine shows a conversion of 57%.Apparently,calcined dolomite is more reactive than olivine with respect totarFig.9.Conversion of different tar classes with olivine as catalyst.L.Devi et al./Renewable Energy 30(2005)565–587577decomposition.As seen from Fig.10,class 4tars are the least reactive and need severetreatment.Adding up the concentration of the different classes gives the amount of totalconsidered tar and with it the total conversion.Fig.11shows the conversion for the bulktar,and Fig.12shows the exiting concentrations of the combined considered tars.It isclear that calcined dolomite is the more active additive.At 8508C,the conversion forcalcined dolomite amounts to 56%,whereas the olivine shows a conversion of about 16%.When the temperature is raised to 9008C,a slight increase of total tar conversion isobserved for dolomite.But for olivine,this increase in temperature shows a bigger effect,atotal tar conversion of 46%is observed,which is a sharp increment.The entering tar concentrations for the dolomite and olivine experiment are approximately 4.0g m 0K 3at 9008C.After catalytic treatment with dolomite,the total tar concentration dropped to 1.5g m 0K 3,and with olivine the concentration dropped to 2.2g m 0K 3.It has to be kept inmind that this is only a part of the tar present in the system.The unidentified heavy andvery light tars are not included,because they cannot be captured by the SPA analyses method.Rapagna `et al.[10]observed tar concentration of approx.2.4g m 0K 3at the exit ofthe gasifier,even at a temperature of 7708C using olivine as bed material.It must be madeclear that Rapagna`and his coworkers used olivine inside the gasifier and used steam as a gasifier medium which favours more steam reforming reactions in the gasifier,thus lesstar in the exiting gas.Fig.12gives the concentration of total considered tars remaining after catalytictreatment along with the individual contribution of each class of tar.At lower temperature,the total amount of considered tar is lower for olivine than for dolomite.This is mainly dueto a higher naphthalene concentration.It is obvious from Fig.12that class 4contributesFig.10.Conversion of tars with different additives (T r Z 9008C).L.Devi et al./Renewable Energy 30(2005)565–587578the major part of the total tar for both catalysts.As the temperature of the catalytic reactorincreases,its percentage in the remaining tar increases,although the amount of class 4tardecreases.At the lower temperature range investigated,class 4tar contributes around 80%of the total tar,whereas with increasing temperature the contribution of class 4tarbecomes around 90%.Simell [18]also reported the effect of temperature on tar removalusing dolomite as a catalyst.They reported similar values of tar content as in this presentpaper for the investigated temperature range except 9008C,where they found less than100mg m 0K 3of tars.It should be kept in mind,however,that the properties of the dolomiteinvestigated in the present study are quite different from that used by Simell [18].For instance,the surface area of the dolomite used in this study is lower than that of thedolomite investigated by Simell.Among all the compounds identified in class 4tars,naphthalene contributes to morethan 50%at the lower temperature range as shown in Fig.13.At 9008C,the contributionof naphthalene is more than 70%,thus more than 60%contribution in total tar for bothdolomite and olivine.Also,the conversion of naphthalene is observed to be lower as givenin Fig.14.This low conversion might be due to the fact that either naphthalene is a verystable compound,or decomposition of higher tars (higher than naphthalene)leads tothe formation of naphthalene.At 8508C,formation of naphthalene and thus class 4tar isobserved over olivine.Due to this formation,the total tar conversion decreases as thereaction temperature was raised from 800to 8508C as shown in Fig.11over olivine.Although benzene is not considered as a tar,it is formed during the decompositionreactions.Benzene normally do not pose problem of condensation.The amountofFig.11.Conversion of total considered tars with additives.L.Devi et al./Renewable Energy 30(2005)565–587579benzene was observed to increase up to a maximum value by Simell [18]withincreasing temperature.At 8008C,benzene contributes almost 80%of the total tar.Asmentioned earlier,in this study benzene and toluene have not been taken into accountwhile calculating the total tar.The approximate concentration of these two compoundsbefore entering the reactor and after catalytic treatment at 850and 9008C is given inFig.15.The formation of benzene is observed to be very high,especially when olivineis used as catalyst.Also formation of toluene is found for both dolomite and olivine at8508C.Dolomite and olivine show different catalytic behaviour for each tar class.Their activityis very much dependent on temperature.At a temperature of 9008C,a total tar concentration at the inlet of around 4.0g m 0K 3could be decreased to 1.5and 2.2g m 0K 3at the outlet withdolomite and olivine,respectively.It is anticipated,that the difference in activity is mainlydue to structural differences and differences in the composition of both catalysts,as shown inFigs.6and 7.The moderate activity of calcined dolomite (compared to literature data)should probably be attributed to its moderate surface area.These results also revealthatFig.12.Contribution of individual tar classes on the amount of total tar.L.Devi et al./Renewable Energy 30(2005)565–587580。
Optimizing Semiconductor Fluid Handling Solutions with 3M Science.3M™ Dyneon™ FluoroplasticsHigh purity material solutionsto help protect chip purityand increase yieldAdvancing semicon.Enhancing lives.From the Internet of Things and trends in 5G, to virtual reality and self-driving vehicles, the growing demand for ever smaller, faster, and more powerful devices has never been greater. 3M brings decades of science-based expertise to a range of raw materials and fluid-handling solutions for chip manufacturing, helping innovators across the industry meet their demand and shape the future of the semiconductor industry.3M ™ Fluoroplastics for fluid handling. Purity. Precision. Performance.As our lives and our devices become increasinglyinterconnected, advances in technology are pushing the limits of integrated circuits and chip design. At 3M, we produce ultra high purity PFA and PTFE materials that help enable the parts, equipment, and critical fluid delivery systems manufacturers need to maximize performance and yield.At our fingertips and all around us, smart devices have revolutionized how the world connects—and how we connect with our world. 3M science helps the products, equipment and fluid-handling systems the semiconductor industry needs today to create the chips of tomorrow.Semicon chip fabrication presents unique fluid-handling challenges, such as harsh cleaning and etching chemicals, high heat, and the need for handling ultra-pure water. Process contamination reduces reliability. As requirements get stricter, chip fabricators rely more than ever on the right materials. 3M ™ Fluoroplastics helpmanufacturers protect against contamination, corrosion and leaching—minimizing downtime and enhancing equipment performance.Challenges around optimizingfluid handling systems3M is a trusted partner serving industries around the world. Our decades of experience in fluoroplastic design andmanufacturing, backed by extensive purity and application knowledge, make 3M ultra high purity PFA and PTFE an excellent choice for semicon fluid handling applications.Our customers across the entire semicon value chain benefit from bench-to-bench 3M support,including material testing expertise and application insights.Reliable solutions backed by 3M scienceOur semicon material solutions start with 3M ™ Fluoroplastics, but they’re built on industry and application expertise. 3M raw materials,including ultra high purity PFA and PTFE, are engineered to maintain purity and integrity in some of the toughest conditions and manydemanding chemical delivery system environments. Our semicon experts help innovators across the value chain tailor 3M material solutions to better serve their customers’ needs and testing requirements.Semicon solutions from end to endChemicalManufacturersTubing & Component Manufacturers A handle on high-purity fluid handlingMax performance and chip yield come from minimum contamination and corrosion. Throughout the fabrication process, 3M ™ Fluoroplastics deliver critical fluids like etching chemicals and ultra pure water so fabs can push the limits of their semiconductor chip and integrated device design. • PFA tubing and piping • PFA valves and fittings • PTFE gaskets and seals • Anti-static PFA• PTFE and PFA sheets • PTFE and PFA films• PFA drums and containers• PFA and PTFE tanks and vessel liners • PFA dip tubes and dispense systemsTanks and Containment LinersChemical Distribution SystemsFour Value Chain Sectors Semifinished GoodsUsed primarily for its extremeresistance to chemical attack, broad temperature range, and transparency relative to PTFE, PFA, is meltprocessable through extrusion and injection molding. Its thermoplastic behavior allows vastly higherproductivity when manufacturing injection-molded parts.3M ™Dyneon ™Ultra High Purity PFA is engineered to withstand some of the toughest conditions while maintaining purity and integrity. Every lot is tested to meet the SEMI C90 standard, guaranteeing quality and consistency that enable ultra high-purity tubing and components needed in the semicon industry.PFAPFA ApplicationsTubing Wafer Carriers Pumps FittingsValvesLinings BasinsWafer CarriersLiningsPumpsGasketsFiltersA fully fluorinated plastic withapplications in nearly every industry, PTFE is an indispensable problem solver. Its low permeability and coefficient of friction, wide service temperature range, universal chemical resistance, and long service life make it ideal for manufacturing compression molded semifinished parts.3M ™ Dyneon ™ PTFE is engineered for technical excellence and reliability in low volume end useapplications such as liners and wafer baths.3M ™ Dyneon ™ TFM ™ Modified PTFE offers a modified PTFE polymer structure that provides enhanced properties including low surfaceroughness and improved mechanical properties.PTFEPTFE ApplicationsWhich raw material is right for your application?At 3M, we apply science in collaborative ways to enhance lives daily. Our skilled researchers at service labs around the globe continue to develop and improvefluoroplastics for semicon. Our applications specialists are here to help you select and handle materials to ensure the smoothest transition into your manufacturing process.While we produce premier raw materials, 3M expertise is deeply involved throughout the semiconvalue chain, including SEMI standards testing and troubleshooting every step of the way. As a global company, we have application engineers and testing capabilities around the world. Our material experts are available to apply extensive research and development catered towards the semiconductor industry—delivering solutions to the toughest challenges.3M ™ Fluoroplasticsare only the beginningA global partner for the world of semiconInvested in the advancement of semiconWarranty, Limited Remedy, and Disclaimer: Many factors beyond 3M’s control and uniquely within user’s knowledge and control can affect the use and performance of a 3M product in a particular application. User is solely responsible for evaluating the 3M product and determining whether it is fit for a particular purpose and suitable for user’s method of application. User is solely responsible for evaluating third party intellectual property rights and for ensuring that user’s use of 3M product does not violate any third party intellectual property rights. Unless a different warranty is specifically stated in the applicable product literature or packaging insert, 3M warrants that each 3M product meets the applicable 3M product specification at the time 3M ships the product. 3M MAKES NO OTHER WARRANTIES OR CONDITIONS, EXPRESS OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, ANY IMPLIED WARRANTY ORCONDITION OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE OR ANY IMPLIED WARRANTY OF NON-INFRINGEMENT OR ANY IMPLIED WARRANTY OR CONDITION ARISING OUT OF A COURSE OF DEALING, CUSTOM OR USAGE OF TRADE. If the 3M product does not conform to this warranty, then the sole and exclusive remedy is, at 3M’s option, replacement of the 3M product or refund of the purchase price.Limitation of Liability: Except where prohibited by law, 3M will not be liable for any loss or damages arising from the 3M product, whether direct, indirect, special, incidental or consequential, regardless of the legal theory asserted, including warranty, contract, negligence or strict liability.Technical Information: Technical information, recommendations, and other statements contained in this document or provided by 3M personnel are based on tests or experience that 3M believes are reliable, but the accuracy or completeness of such information is not guaranteed. Such information is intended for persons with knowledge and technical skills sufficient to assess and apply their own informed judgment to the information. No license under any 3M or third party intellectual property rights is granted or implied with this information.3M Dyneon and TFM are trademarks of 3M Company. Used under license.Please Recycle. Printed in USA. © 3M 2022. All rights reserved. Issued: 01/20223M Advanced Materials Division 3M CenterSt. Paul, MN 55144 USAP hone 1-800-367-8905Web /SemiFluidHandling。
专利名称:A PACKAGE CAPABLE OF RELEASING PRESSURE发明人:ARDPRU, Ittiporn,MAHAJAROENSIRI, Juthamas,NGAMNAWAKUL,Buncherd,VORAKUNPINIJ, Adisak申请号:TH2019/000023申请日:20190719公开号:WO2020/022965A1公开日:20200130专利内容由知识产权出版社提供专利附图:摘要:The present invention relates to a package capable of releasing pressureformed therein. Said packaging comprises a container portion which is formed for accommodating product, and a flexible film part comprising at least one pressure-releasing design on an external surface of said part, characterized in that said pressure-releasing design comprises a score of at least two identical or different patterns combined together such that no less than three intersections of the patterns are fanned and at least one said pattern is a closed-shape pattern. Moreover, the present invention relates to the pressure-releasing design for a flexible film, characterized in that said pressure-releasing design comprises a mark of at least two identical or different patterns combined together such that no less than three intersections of the patterns are formed and at least one said pattern is a closed-shape pattern.申请人:SCG PACKAGING PUBLIC COMPANY LIMITED地址:1 Siam Cement Road Bangsue Sub-District, Bangsue District, Bangkok, 10800 TH 国籍:TH代理人:SUKSANKRAISORN, Usacha更多信息请下载全文后查看。
Applying Reinforcement Learning to Competitive Tetris Max Bodoia(mbodoia@)1618Sand Hill RdPalo Alto,CA98498Arjun Puranik(apuranik@)1618Sand Hill RdPalo Alto,CA98498IntroductionFor our project,we attempt to apply reinforcement learning to the game of Tetris.The game is played on a board with 20rows and10columns,and each turn a player drops one of seven pieces(the seven tetrominoes)onto thefield.Points are gained by dropping pieces in such a way that all ten squares in some line(row)arefilled,at which point the line is cleared. Multiple lines may be cleared with a single piece drop and will result in more points being earned.A player loses if the height of the pieces on his board exceeds the height of the board itself.The fully observable nature of the Tetris board and the sim-ple probabilistic transitions from state to state(i.e.adding a randomly selected piece to the end of the piece queue each turn)naturally suggest the use of reinforcement learning for Tetris.Specifically,Tetris can be modeled as a Markov Deci-sion Process.However,the state space of Tetris is extremely large-the number of ways tofill in a20×10board is2200, and the Tetris requirement that no row be completelyfilled only reduces this to(210−1)20.As a result,the complete MDP for Tetris is entirely intractable.In our paper,we will outline the different approaches we took to dealing with this intractability.We begin by describing our initial,unsuccess-ful attempts,and commenting on the reasons why they may have failed.Next,we describe the approach that eventu-ally suceeded-fitted value iteration-and give a detailed analysis of its results.Finally,we conclude by considering the strengths and weaknesses of our implementation offitted value iteration,and highlight other potential approaches that we did not try.Initial AttemptsIn this section we will outline our early approaches to the problem.In each case we describe the motivations for the ap-proach,the extent to which it was successful,and the reasons why it was ultimately unsuccessful.Block DropAs an initial pass at the problem of Tetris,we chose to imple-ment a simple game which we will call Block Drop.Block Drop is comparable to Tetris in that players drop pieces onto a rectangular grid,and gain points forfilling up an entire row (clearing a line).The main difference is that in Block Drop, the only type of piece available to players is a single unit square.Although the relationships between states in Block Drop are much more simple than in Tetris(and thus the op-timal policy is much easier to learn),the size of the state space for Block Drop and Tetris are similar.In this sense, the tractability of Block Drop can be used as a rough lower bound on the tractability of Tetris.We wrote a program to formulate an m×n game of Block Drop as an MDP.Note that a valid Block Drop gameboard never has any holes(where a certain grid square is notfilled but a higher up square in the same column isfilled).This al-lows us to represent each state by a set of n values correspond-ing to the height of the pieces in each of the n columns.These column heights can take on values from0to m−1indepen-dently of each other(columns heights are bounded by m−1 since the game ends if a column reaches height m).We also added a single terminal state to represent losing the game, so the total state space of Block Drop is m n+1.One pos-sible action exists for each column(since the player chooses which column to drop a piece in)so there are a total of n ac-tions.Transition probabilities are completely deterministic-dropping a piece in a particular column either increments that column’s value by1,or decrements the value of every column by1(in the case when a line is cleared).Our reward function assigns a reward of1for clearing a line and a cost of m for losing the game;all basic moves had0cost.In order to solve the Block Drop MDP,we used the ZMDP package written by Trey Smith(Smith&Simmons,2012). This package is designed primarily forfinding(approximate) solutions to Partially Observable MDPs,and as a result it re-quires that the input problem be described as a POMDP rather than an MDP.Unfortunately,we were unable tofind any other software that works for pure MDPs,so we decided to con-vert our problem into a POMDP by adding a single observa-tion that is always made.We assumed that this modification would not affect the tractability of the problem significantly. We found that the solution to Block Drop matched the in-tuitively obvious optimal policy,and provided a good sanity check for our use of MDPs to model the problem.The result-ing policy drops a single block into each column(from left to right)until a line is cleared,at which point it repeats.How-ever,the Block Drop problem also illustrated the tractability issues that we faced:4×4Block Drop took less than a sec-ond to solve,but5×4Block Drop took16seconds and6×4 Block Drop took over a minute;all sizes above7×5were unable to complete a single solver round.Figure1shows the time in seconds required for a solution to Block Drop as afunction of the board size (in terms of total number of grid squares).These results made it clear that the actual game of Tetris would be intractable without significant approxima-tions;our next task was to decide how to make these approx-imations.Figure 1:Solution times for Block DropNo-Holes TetrisIn Tetris,unlike in Block Drop,we cannot represent the states of the board by a single height for each column.It is possi-ble that holes -where a column has an empty block below the highest filled block -are formed in the board when pieces are dropped.In fact,the management of these holes is one of the more difficult aspects of the game.However,we decided to approximate board states in this manner by emulating the state representations for Block Drop and tracking only the height of the highest block in each column.We refer to the version of Tetris that results from this approximation as No-Holes Tetris.The number of possible boards for No-Holes Tetris is significantly smaller than for Tetris (m n compared to around 2mn )and was the main motivation for making this approximation.We attempted to counterbalance the inaccu-racy of this simplification by also modifying the reward func-tion of No-Holes Tetris,so that the player loses points upon taking actions which produce holes.This causes the opti-mal policy for No-Holes Tetris to avoid actions that lead it to states that differ from the corresponding true Tetris state.Furthermore,we observed that competitive Tetris players al-most never leave holes in their stacks.As a result we hoped that the optimal policy for our approximate No-Holes Tetris would remain close to the optimal policy of true Tetris.We then wrote a program to formulate an m ×n game of No-Holes Tetris as an MDP.As noted above,each board of No-Holes Tetris is represented by a vector of n values be-tween 0and m −1,representing the heights of the blocks in each column.In addition,each state of No-Holes Tetris also includes the next piece available to the player.For p pos-sible pieces,this gives a total state space of p ·m n +1(in-cluding the terminal state).The set of actions available to the player at each state is the set of ways (including rotations)that the given piece can be dropped onto the board.Transitionprobabilities are deterministic to the extent that they affect the board and uniformly random to the extent that they affect the piece,so each state/action pair leads uniformly to one of p possible next states.The reward function assigns a cost of m to losing the game,and the reward for a given state/action pair is the number of lines cleared minus the number of holes created.We found (unsurprisingly)that No-Holes Tetris is signif-icantly more difficult to solve than Block Drop.4×4No-Holes Tetris took around 20seconds to solve,while on larger boards our solver was unable to find good policies.The best policies found had regret values greater than 0.5on a 5×4board and greater than 4on a 6×4board,even when the solver is allowed to run for long periods.We tried varying the number of pieces to reduce complexity and found that the difficulty of the problem scaled extremely quickly with the number of pieces.No-Holes Tetris with only one piece could be solved about as quickly as Block Drop,but attempts to solve 6×4No-Holes Tetris with two pieces were unable to reduce regret below 1.Figure 2shows the solution times of No-Holes Tetris as a function of board size when different total numbers of pieces areused.Figure 2:Solution times for No-Holes TetrisThese results were disappointing,to say the least.We knew from our analysis of Block Drop that solving No-Holes Tetris would be intractable for board sizes above 7×5.How-ever,our hope was that we could find a ”good enough”pol-icy for the full 7-piece No-Holes Tetris on a ”large enough”board.This policy would serve as a starting point for local piece placement that we could use to build policies for larger boards.We planned to do this by dividing larger boards into smaller sub-boards and choosing actions based on some func-tion of the sub-board policies.Unfortunately,No-Holes Tetris proved to be too computationally difficult to find suitable sub-board policies.On the one hand,the 4×4board was small enough that the outputted policy for No-Holes Tetris was de-termined primarily by height constraints,and did not gener-alize well to larger boards.On the other hand,the best 5×4policy performed poorly even on the 5×4board.This left us with no good starting point from which to construct a policyfor larger boards,and no obvious way of applying our results effectively to the general problem of playing Tetris. Factored MDPsWe briefly considered representing Tetris(or No-Holes Tetris)as a factored MDP.Some of the existing literature on MDP and POMDP solution algorithms explores ways of op-timizing the solution process by representing the state space in factored form(Guestrin,Koller,Parr,&Venkataraman, 2003),(Poupart,2005).That is,we would represent each state in terms of somefinite number of state variables,and describe the transition probabilities and rewards as functions of these variables instead of defining a particular value for every state.Tetris,like most games,is highly factorable and therefore well suited to this kind of approach.Possible factor-izations of an m×n board include using a binary variable for each of the mn grid squares(which encodes the full version of Tetris),and representing each of the n columns as a vari-able that takes m possible values(which encodes No-Holes Tetris).Ultimately,however,we decided not to pursue this ap-proach.The benefits of representing MDPs in factored form come primarily from the ability to represent particular tran-sition probabilities and rewards as functions of strict subsets of the full set of state variables.In other words,in order for a factored MDP to be more efficiently solvable than its un-factored counterpart,the state variables of the factored form must be independent to some extent.Atfirst,No-Holes Tetris seems to exhibit a great deal of independence:each time a piece is dropped,at most four state variables(i.e.columns) will change in value,and these changes will depend only on the values of the other three state variables.However,the fact that state transitions must also take into account line clear-ances eliminates this independence.Line clearances only oc-cur when a particular row isfilled in at every column,and as a result,the transition probabilities for each state variable de-pend on the value of all n state variables.For this reason,we judged that the use of a factored representation would not pro-vide significant reductions in tractability and decided against this approach.Fitted Value IterationThe next approach that we tried(and the one that ultimately proved most successful)wasfitted value iteration.The ba-sic principle behindfitted value iteration is to choose a small set of features and represent each state in terms of this fea-ture set.In this respect,it is reminiscent of the factoring ap-proach.Importantly,however,the state space of a factored MDP is necessarily the same size as the original state space, while the set of possible combinations of feature values may be much smaller.Furthermore,fitted value iteration does not consider this full set of possible feature values,but instead re-stricts itself to the feature representations of a sampled set of states.This means that the tractability offitted value iteration depends on the number of samples rather than the size of the state space.In general,two basic properties must hold for a particular MDP in order forfitted value iteration to be effective.First, it must be possible to approximate the true value of a par-ticular state using only a small set of information about that state.This ensures that the”featurization”of the MDP is rea-sonably representative of the original.Second,the number of samples from the original state space needed to represent the relationship between state values and state features must be relatively small.This ensures that an accurate function from features to values can be learned using the sampled set of states.Intuitively,Tetris seems to satisfy both properties, so the application offitted value iteration is a natural choice. Our algorithm is roughly identical to thefitted value it-eration algorithm presented at the end of the Reinforcement Learning handout,with a few small caveats.It begins by sam-pling a set of states S s from the state space S and initializing a parameter vectorθto zero.Then,it alternates between two steps.In thefirst step it calculates,for each state,the maxi-mum over all actions of the expected utility of being in that state:y s=max a∈A R(s)+γE s ∼Psa[V(s )]for each s∈S s.In the second step,itfits the values of the parameters to this set of maximum expected utilities:θ:=argminθ12∑s∈Ss(θTφ(s)−y s)2.Here,A is the set of all actions,R(s)is the reward received from being in state s,P sa is the distribution over possible tran-sition states resulting from taking action a in state s,andφ(s) is the vector of features values for state s.The main difference between our algorithm and the one presented in the handout is that instead of sampling from the state space as a whole,we take samples by randomly placing pieces around the board.This seems reasonable,since a truly random Tetris board is unlikely to look anything like the kinds of boards found in Tetris gameplay,and the set of states that can be generated using this sampling process is identical to the set of states it is possible for our agent to encounter during test time.The other difference is that because we already know the transition probabilites in Tetris,we can compute R(s)+γE s ∼Psa[V(s )]directly instead of having to sample it. Once the algorithm converges,the parameter settings can be used by an agent to play Tetris.The agent chooses moves in a manner similar to thefirst step of the algorithm.When in state s,it chooses action a according to:a =argmax a∈A E s ∼Psa[V(s )].Feature SelectionThefitted value iteration algorithm given above assumes the existence of a functionφfrom states to feature values.A full implementation therefore requires the choice of a feature set and the definition ofφ.Note thatφcan be any conceivable functionφ:S→R k for some k∈N,where k is the number of features.This means that the”feature space”of possible functionsφis both enormous and difficult to define.Although techniques exist for automating the process of feature selec-tion,it is far more common to define features manually usingdomain knowledge(Hall,1999).We considered a variety of possible features based on our personal experience playing Tetris.Thefirst kind of features we considered were”summary”features that are functions of the whole board.The most straightforward summary features we examined were Max-Height,the maximum height of all the columns,and Num-Holes,the total number of holes on the board(where a”hole”is an unfilled grid square for which there exists afilled grid square higher up in the same col-umn).Other more complicated features included:Avg-Diff, the average of the absolute values of differences between ad-jacent columns;Max-Diff,the maximum of these absolute values;and Num-Covers,the total number of covers on the board(where a”cover”is afilled grid square for which there exists an unfiled grid square lower down in the same column). For each proposed feature,we conducted preliminary tests of the feature by running our algorithm using only this feature and a constant feature with value1for every state.We then measured the average lifespan(i.e.number of moves made before losing)for an agent that plays using the learned pa-rameters and compared it to the average lifespan of an agent that places pieces randomly.Only the features Max-Height, Num-Holes,and Num-Covers led to better than random per-formance.In addition to the summary features,we also considered sets of non-summary features that were functions of partic-ular columns.These feature sets corresponded to different summary features:examples include i-Col-Height,the height of the i th column,and i-Col-Diff,the absolute value of the difference between the i th and(i−1)th columns.However, when we tested each of these potential feature sets individu-ally,we found that an agent playing with the learned parame-ters performed no better than random play.This is most likely because the relationships between these feature sets and the true value function are non-linear.Since our algorithmfinds parameter settings using linear regression,it does not have the potential to learn non-linear relationships well and typ-ically ends up converging to arbitrary parameter values.In theory,the parameter update step could use a wide variety of machine learning algorithms,and using a more complex re-gression algorithm could allow it to learn more intricate rela-tionships between the features and the state values.Without such modifications,however,our algorithm cannot perform well using the non-summary feature sets considered and so we ultimately chose not to consider them further.ResultsAfter this preliminary testing,we were left with three features -Max-Height,Num-Holes,and Num-Covers-that seemed promising.We then conducted a series of more rigorous tests to determine which combinations of features led to the best performance.For each possible combination of these three features,we ran our algorithm to convergence5times and tested an agent for each of the resulting parameter vectors. The algorithm used1000sampled states and a discount factor of0.9on a full20×10,7-piece board,and iterated until the maximum difference between corresponding parameter val-ues from one iteration to the next dropped below0.01.In test-ing,each agent played100games and the average number of moves per game was computed.Table1shows the average of the parameter vectors and average lifespans over all5agents for each possible feature set.Thefirst value in each parame-ter vector is the value of the constant feature,followed by the values of the parameters corresponding to the other features in the order listed.Table1:Average agent lifespans and parameter vectors Feature Set Avg Lifespan Parameter VectorConstant25.8(-13.8)Max-Height56.8(21.8,-2.8)Num-Holes47.7(17.6,-2.5)Num-Covers40.5(12.6,-1.0)MH,NH126.0(8.2,-0.4,-0.7)MH,NC72.1(63.1,-3.7,-0.2)NH,NC42.0(-15.1,-5.5,-0.1)MH,NH,NC129.4(-17.0,-1.0,-0.9,-0.1) Max-Height seems to be the most informative feature since it gives the best single-feature performance and the two fea-ture pairs that use it perform better than the third.Num-Holes likely comes in second given the stark difference between performance with Max-Height,Num-Holes and with Max-Height,Num-Covers.Num-Covers has the worst individual performance and it makes a difference only when combined with Max-Height,so it seems to be the least informative of the three.These rankings seem to make intuitive sense since looking at the maximum column height and total number of holes on a Tetris player’s board is one of the most obvious ways to judge whether they are doing well.On the other hand,however,the differences in performance given by these three features may not be due to inherent dif-ferences in their informativeness about the true value func-tion,but rather a product of the random sampling process we used.In general,random play clears lines rarely but loses of-ten,and thus the set of sampled states contains many more examples of terminal states than of line clearances.Since Max-Height is especially indicative of whether the player is near a terminal state,while the other two features relate more to how easily lines can be cleared,it is possible that the cor-rect value of Max-Height is simply the easiest to learn from random play.To test this hypothesis,we conducted another set of trials for the three-feature set using a modified version of the algorithm with multiple training rounds.In thefirst round,we learn parameter settings for the three features us-ing a set of states sampled from random play.In subsequent rounds,the set of states is sampled from play using the pa-rameter settings from the previous round,with a probability of0.3to choose a random action.However,this algorithm showed no significant change in parameters from round to round(even though subsequent rounds reinitialized the pa-rameters to0after collecting the samples).This suggests that the parameter settings for the three-feature set given in Table 1are likely the optimal values achievable by our algorithm. The last method we tried for improving the effectiveness of our agent was to allow it to preview the upcoming two pieces and add a2-step look-ahead feature.Most versions of Tetris allow players to preview several upcoming pieces,so it seemed fair to allow our agent the same privilege.During training,our agent learns a parameter vector for the three-feature set as it did before.In gameplay,however,it maxi-mizes the expected value of the next state over all possible combinations of actions with the current piece and the twopreviewed pieces.In other words,if P sa1,s2a2,s3a3is the tran-sition distribution over next states after taking actions a1,a2, and a3in s1,s2,and s3respectively,the agent chooses an ac-tion a according to:a=argmax a1∈A R(s2)+R(s3)+E s ∼Ps1a1,s2a2,s3a3[V(s )].We found that this2-step lookahead agent was a highly effective Tetris player-over the course of several thousand moves,it never lost once and rarely entered a state with a maximum column height above half of the total board height. To ensure that this performance was influenced by our param-eter learning(as opposed to merely the lookahead),we tested lookahead agents that had no learned parameters and simply chose actions based on the maximum reward attained dur-ing lookahead.In this case,2-and3-step lookahead agents performed no better than random play and agents with larger lookaheads took dozens of seconds to choose moves.These results demonstrate how agents that utilize both low-order lookahead and parameter learning can significantly outper-form agents which use only one or the other.ConclusionsWe began this project with the hopes of creating an agent ca-pable of playing competitive Tetris against a human player. Very quickly on,however,we realized that we had underesti-mated the difficulty of basic,single-player Tetris.The beast of exponential size reared its ugly head,and it became clear that we could not simply encode the game as an MDP and expect it to be tractably solvable.Even reducing the board size by a factor of eight and ignoring holes was not enough to allow for the computation of an optimal policy,and the tricky nature of line clearances prevented us from taking advantage of independence between factored state variables.This forced us to change tactics and use a less straightfor-ward approach tofind an approximate solution to the MDP. Thefitted value iteration algorithm was a natural choice,since Tetris boards are well characterized by a small set of features that can be learned with relatively few sampled states.Af-ter some experimentation,we found three features-Max-Height,Num-Holes,and Num-Covers-that seemed to be roughly linearly correlated with the”true”value of Tetris states.An agent that learned parameter values correspond-ing to these features was able to survive for an average of around130moves.This was not a trivial feat:a random agent only survives for an average of25moves,and without the use of previewed pieces-a staple of almost all versions of Tetris-humans without plenty of experience would likely not perform much better.Furthermore,the addition of two previewed pieces and a2-step lookahead allowed our agent to stay alive indefinitely.Thisfinal instantiation of our agent is particularly interest-ing because of the way it combinesfitted value iteration and simple search and ends up performing better than with either technique individually.In this case,the paradigm of uniting machine learning and conventional algorithms proved to be extremely effective.Future versions of our agent could likely be made even more effective if we incorporated other stand-alone techniques as well.The linear regression step could easily be replaced by a more advanced regression technique that would allow our agent to learn non-linear relationships between the features and state values,and automated tech-niques for feature selection might be able tofind relationships that are not readily apparent to human players.Neverthe-less,ourfinal agent is a reasonably effective Tetris player as it stands,and could provide a solid foundation if we wished to extend our agent to the competitive Tetris arena.ReferencesGuestrin,C.,Koller,D.,Parr,R.,&Venkataraman,S.(2003). Efficient solution algorithms for factored mdps.J.Artif. Intell.Res.(JAIR),19,399–468.Hall,M.(1999).Correlation-based feature selection for machine learning.Unpublished doctoral dissertation,The University of Waikato.Poupart,P.(2005).Exploiting structure to efficiently solve large scale partially observable markov decision pro-cesses.Unpublished doctoral dissertation,Citeseer. Smith,T.,&Simmons,R.(2012).Point-based pomdp al-gorithms:Improved analysis and implementation.arXiv preprint arXiv:1207.1412.。