Engineering the Compression of Massive Tables An Experimental Approach SIAM 2000.
- 格式:pdf
- 大小:209.84 KB
- 文档页数:10
Aabrasive grinding 强力磨削abrasive[☜'breisiv]a.磨料的, 研磨的absence ['✌bs☜ns] n.不在,缺席accesssory[✌k'ses☜ri] n.附件accommodate[☜'k m☜deit]v. 适应accordingly[☜'k :di☠li]adv.因此,从而,相应地accuracy['✌kjur☜si]n精度,准确性actuate['✌ktjueit]vt.开动(机器), 驱动adequate['✌dikwit] a. 足够的adhesive[☜d'hi:siv] n. 粘合剂adjacent[☜'d✞eisnt] a. 邻近的adopt[☜'d pt] vt. 采用advance [☜d'v✌:ns] n.进步advisable [☜d'vaizbl]adj. 可取的agitate['✌d✞iteit] v. 摇动a large extent 很大程度algorithm ['✌l♈☜ri❆☜m] n. 算法align [☜'lain] v 定位,调准alignment[☜'lainm☜nt] n. 校直all-too-frequent 频繁allowance[☜'l☜uens]n. 容差, 余量alternate[' :lt☜nit] v.交替,轮流alternative[ :l't☜:n☜tiv]n. 替换物alternatively[ :l't☜:n☜tivli]ad. 做为选择, 也许aluminiun[ ✌lju'minj☜m] n.铝ample['✌mpl] adj. 充足的analysis [☜'n✌l☜sis] n. 分析ancillary['✌nsil☜ri]a.补助的, 副的angular ['✌♈jul☜] adj. 有角的annealing[☜'li:li☠] n.退火aperture ['✌p☜t☞☜] n.孔applied loads 作用力appropriate [☜'pr☜uprieit]a. 适当的arc[a:k] n.弧, 弓形arise[☜'raiz] vi. 出现, 发生arrange[☜'reid✞] v. 安排article['a:tikl] n.制品, 产品ascertain[ ✌s☜'tein]vt. 确定, 查明assemble[☜'sembl] vt.组装attitude ['✌titju:d] n 态度auxiliary [ :♈'zilj☜ri]adj. 辅助的avoid[☜'v id] v.避免axis['✌ksis] n.轴axle['✌ksl] n.轮轴, 车轴Bbackup['b✌k ✈p] n. 备份batch [b✌t☞] n 一批bearing['b☪☜ri☠] n.轴承,支座bed[bed] n. 床身behavior[bi'heivj☜] n. 性能bench-work 钳工工作bend[bend] v. 弯曲beneath[bi'ni: ] prep在···bin [bin] n. 仓,料架blank [bl✌☠k] n. 坯料blank [bl✌☠k] v. 冲裁,落料blanking tool 落料模blast [bl✈st] n.一阵(风)blemish['blemi☞]n. 缺点, 污点bolster['b☜ulst☜] n. 模座,垫板boost[bu:st] n. 推进boring['b :ri☠] n.镗削, 镗孔bracket ['br✌kit] n. 支架brass [br✌s] n.黄铜break down 破坏breakage ['breikid✞] n.破坏bridge piecebrine[brain] n. 盐水brittle['britl] adv.易碎的buffer [b✈f☜] n.缓冲器built-in 内装的bulging [b✈ld✞i☠] n. 凸肚burr [b☜:] n. 毛刺bush [bu☞] n. 衬套bush[bu☞] n. 衬套by far (修饰比较级, 最高级)···得多, 最by means of 借助于Ccabinet ['k✌binit] n.橱柜call upon 要求carbide['ka:baid] n.碳化物carburzing['ka:bjureti☠] n. 渗碳carriage['k✌rid✞]n.拖板, 大拖板carry along 一起带走carry down over 从···上取carry out 完成case hardening 表面硬化case[keis] n. 壳, 套cast steel 铸钢casting['ka:sti☠] n. 铸造,铸件category['k✌t☜♈☜uri] n. 种类caution ['k :☞☜n]n. 警告,警示cavity and core plates凹模和凸模板cavity['k✌viti] n.型腔, 腔, 洞centre-drilling 中心孔ceramic[si'r✌mik] n.陶瓷制品chain doted line 点划线channel['t☞✌nl] n.通道, 信道characteristic[k✌r☜kt☜'ristik]n.特性check[t☞ek] v.核算chip[t☞ip] n.切屑, 铁屑chuck [t☞✈k] n.卡盘chute [☞u:t] n. 斜道circa ['s☜k☜:] adv. 大约circlip['s☜:klip] n.(开口)簧环circuit['s☜:kit] n. 回路, 环路circular supoport block circulate['s☜:kjuleid]v.(使)循环clamp [kl✌mp] vt 夹紧clamp[kl✌mp] n.压板clay[klei] n. 泥土clearance ['kli☜r☜ns] n. 间隙clip [klip] vt. 切断,夹住cold hobbing 冷挤压cold slug well 冷料井collapse[k☜'l✌ps]vi. 崩塌, 瓦解collapsible[k☜'l✌ps☜bl]adj. 可分解的combination [k mbi'nei☞☜n] n. 组合commence[k☜'mens]v. 开始, 着手commence[k☜'mens] v. 开始commercial [k☜'m☜:☞☜l]adj. 商业的competitive[k☜m'petitiv]a. 竞争的complementary[ k mpli'ment☜ri]a. 互补的complexity [kem'pleksiti]n. 复杂性complicated['k mpl☜keitid]adj. 复杂的complication [k mpli'kei☞☜n] n. 复杂化compression [k☜m'pre☞☜n] n.压缩comprise[k☜m'prais] vt.包含compromise['k mpr☜maiz]n. 妥协, 折衷concern with 关于concise[k☜n'sais]a. 简明的, 简练的confront[k☜n'fr✈nt] vt. 使面临connector[k☜'nekt☜]n. 连接口, 接头consequent['k nsikw☜nt]a. 随之发生的, 必然的console ['k nsoul] n.控制台consume [k☜n'sjum]vt. 消耗, 占用consummate [k☜n's✈meit]vt. 使完善container[k☜n'tein☜] n. 容器contingent[ken'tind✞☜nt]a.可能发生的contour['k☜ntu☜] n.轮廓conventional[k☜n'ven☞☜nl]a. 常规的converge[k☜n'v☜:d✞]v. 集中于一点conversant[k n'v☜:s☜nt]a. 熟悉的conversion[k☜n'v☜:☞☜n]n 换算, 转换conveyer[ken'vei☜]n. 运送装置coolant['ku:l☜nt] n. 冷却液coordinate [k☜u' :dnit]vt. (使)协调copy machine 仿形(加工)机床core[k :] n. 型芯, 核心corresponding [ka:ri'sp di☠]n.相应的counteract [kaunt☜'r✌kt]vt. 反作用,抵抗couple with 伴随CPU (central processing unit)中央处理器crack[kr✌k ] v.(使)破裂,裂纹critical['kritikl] adj.临界的cross-hatching 剖面线cross-section drawn 剖面图cross-slide 横向滑板CRT (cathoder-ray tube)阴极射线管crush[kr✈☞] vt.压碎cryogenic[ krai☜'d✞enik ]a. 低温学的crystal['kristl] adj.结晶状的cubic['kju:bik]a. 立方的, 立方体的cup [k✈p]vt (使)成杯状, 引伸curable ['kjur☜bl]adj. 可矫正的curvature['k☜:v☜t☞☜] n.弧线curve [k☜:v] vt. 使弯曲cutter bit 刀头, 刀片cyanide['sai☜naid] n.氰化物Ddash [d✌☞] n. 破折号daylight ['deilait] n. 板距decline[di'klain]v. 下落,下降,减少, ,deform[di'f :m] v. (使)变形demonstrate['dem☜streit ]v 证明depict[di'pikt ] vt 描述deposite [di'p zit] vt. 放置depression[di'pre☞☜n] n. 凹穴descend [di'sent] v. 下降desirable[di'zair☜bl] a. 合适的detail ['diteil] n.细节,详情deterioration[diti☜ri:☜'rei☞☜n]n. 退化, 恶化determine[di't☜:min] v.决定diagrammmatic[ dai☜gr☜'m✌tik] a. 图解的,图表的dictate['dikteit] v. 支配die[dai] n.模具, 冲模, 凹模dielectric[daii'lektrik]n. 电介质die-set 模架digital ['did✞itl ]n. 数字式数字, a.数字的dimensional[dddi'men☞☜nl]a. 尺寸的, 空间的discharge[dis't☞a:d✞]n.v. 放电, 卸下, 排出discharge[dis't☞a:d✞] v.卸下discrete [dis'cri:t]adj. 离散的,分立的dislodge[dis'l d✞]v. 拉出, 取出dissolution[dis☜'lu:☞☜n] n.结束distinct [dis'ti☠kt]a.不同的,显著的distort [dis'd :t] vt. 扭曲distort[dis't :t]vt. (使)变形, 扭曲distributed system分布式系统dowel ['dau☜l] n. 销子dramaticlly [dr☜'m✌tikli]adv. 显著地drastic ['dr✌stik] a.激烈的draughting[dra:fti☠] n. 绘图draughtsman['dr✌ftsm☜n]n. 起草人drawing['dr :i☠] n. 制图drill press 钻床drum [dr✈m] n.鼓轮dual ['dju:☜l]adv. 双的,双重的ductility [d✈k'tiliti ] n.延展性dynamic [dai'n✌mik ]adj 动力的Eedge [ed✞] n .边缘e.g.(exempli gratia) [拉] 例如ejector [i'd✞ekt☜] n.排出器,ejector plate 顶出板ejector rob 顶杆elasticity[il✌'stisiti] n.弹性electric dicharge machining电火花加工electrochemical machining电化学加工electrode[i'lektr☜ud] n. 电极electro-deposition 电铸elementary [el☜'ment☜ri]adj.基本的eliminate[i'limineit]vt. 消除, 除去elongate[i'l ☠♈et]vt. (使)伸长, 延长emerge [i'm☜:d✞]vi. 形成, 显现emphasise['emf☜saiz]vt. 强调endeavour[en'dev☜] n. 尽力engagement[in'♈eid✞ment]n. 约束, 接合enhance[in'h✌ns]vt. 提高, 增强ensure [in'☞u☜] vt. 确保,保证envisage[in'vizid✞] vt.设想erase[i'reis] vt. 抹去, 擦掉evaluation[i'v✌lju ei☞☜n]n. 评价, 估价eventually[i'v☜nt☞u☜li ]adv. 终于evolution[ev☜'lu:☞☜n] n.进展excecution[eksi'kju:☞☜n]n. 执行, 完成execute ['ekskju:t] v. 执行exerte [i♈'z☜:t] vt. 施加experience[iks'piri☜ns] n. 经验explosive[iks'pl☜usiv]adj.爆炸(性)的extend[eks'tend] v. 伸展external[eks't☜:nl] a. 外部的extract[eks'tr✌kt] v. 拔出extreme[iks'tri:m] n. 极端extremely[iks'tri:mli]adv. 非常地extremity[iks'tmiti] n. 极端extrusion[eks'tru:✞☜n]n. 挤压, 挤出FF (Fahrenheit)['f✌r☜nhait]n. 华氏温度fabricate ['f✌brikeit]vt. 制作,制造facilitate [f☜'siliteit] vt. 帮助facility[f☜'siliti] n. 设备facing[feisi☠] n. 端面车削fall within 属于, 适合于fan[f✌n] n.风扇far from 毫不, 一点不, 远非fatigue[f☜'ti♈] n.疲劳feasible ['fi:z☜bl] a 可行的feature ['fi:t☞☜] n.特色, 特征feed[fi:d] n.. 进给feedback ['fi:db✌k] n. 反馈female['fi:meil]a. 阴的, 凹形的ferrule['fer☜l] n. 套管file system 文件系统fitter['fit☜] n.装配工, 钳工fix[fiks]vt. 使固定, 安装, vi. 固定fixed half and moving half定模和动模flat-panel technology平面(显示)技术flexibility[fleksi'biliti]n. 适应性, 柔性flexible['fleks☜bl] a. 柔韧的flow mark 流动斑点follow-on tool 连续模foregoing ['f :'♈☜ui☠]adj.在前的,前面的foretell[f :'tell]vt. 预测, 预示, 预言forge[f :d✞] n. v. 锻造forming[f :mi☠] n. 成型four screen quadrants四屏幕象限fracture['fr✌kt☞☜] n.破裂free from 免于Ggap[♈✌p] n. 裂口, 间隙gearbox['♈i☜b ks] n.齿轮箱general arrangementgovern['♈✈v☜n]v. 统治, 支配, 管理grain [♈rein] n. 纹理graphic ['♈r✌fik] adj. 图解的grasp [♈r✌sp] vt. 抓住grid[♈rid] n. 格子, 网格grind[♈raind]v. 磨, 磨削, 研磨grinding ['♈raindi☠]n. 磨光,磨削grinding machine 磨床gripper[♈rip☜] n. 抓爪, 夹具groove[♈ru:v] n. 凹槽guide bush 导套guide pillar 导柱guide pillars and bushes导柱和导套Hhandset['h✌ndset] n. 电话听筒hardness['ha:dnis] n.硬度hardware ['ha:dw☪☜] n. 硬件headstock['hedst k] n.床头箱, 主轴箱hexagonal[hek's✌♈☜nl]a. 六角形的, 六角的hindrance['hindr☜ns]n.障碍, 障碍物hob[h b] n. 滚刀, 冲头hollow-ware 空心件horizontal[h ri'z ntl]a. 水平的hose[h☜uz] n. 软管, 水管hyperbolic [haip☜'b lik]adj. 双曲线的Ii.e. (id est) [拉] 也就是identical[ai'dentikl] a同样的identify [ai'dentifai] v. 确定,识别idle ['aidl] adj.空闲的immediately[i'mi:dj☜tli] adv.正好, 恰好impact['imp✌kt] n.冲击impart [im'pa:t] v.给予implement ['implim☜nt]vt 实现impossibility[imp s☜'biliti]n.不可能impression[im'pre☞☜n] n. 型腔in contact with 接触in terms of 依据inasmuch (as)[in☜z'm✈t☞]conj. 因为, 由于inch-to-metric conversions英公制转换inclinable [in'klain☜bl]adj. 可倾斜的inclusion [in'klu☞☜n]n. 内含物inconspicuous[ink☜n'spikju☜s]a. 不显眼的incorporate [in'k :p☜reit]v 合并,混合indentation[ inden'tei☞☜n ]n. 压痕indenter[in'dent☜] n. 压头independently[indi'pein☜ntli]a. 独自地, 独立地inevitably[in'evit☜bli]ad. 不可避免地inexpensive[inik'spensiv]adj. 便宜的inherently [in'hi☜r☜ntli]adv. 固有的injection mould 注塑模injection[in'd✞ek☞☜n] n. 注射in-line-of-draw 直接脱模insert[in's☜:t] n. 嵌件inserted die 嵌入式凹模inspection[in'spek☞☜n]n. 检查,监督installation[inst☜'lei☞☜n]n. 安装integration [inti'♈rei☞☜n]n.集成intelligent[in'telid✞☜nt]a. 智能的intentinonally [in'ten☞☜n☜li]adv 加强地,集中地interface ['int☜feis] n.. 界面internal[in't☜:nl] a. 内部的interpolation [int☜p☜'lei☞☜n]n.插值法investment casting 熔模铸造irregular [i'regjul☜]adj. 不规则的,无规律irrespective of 不论, 不管irrespective[iri'spektiv]a. 不顾的, 不考虑的issue ['isju] vt. 发布,发出Jjoint line 结合线Kkerosene['ker☜si:n] n.煤油keyboard ['ki:b :d ] n. 健盘knock [n k] v 敲,敲打Llance [la:ns] v. 切缝lathe[lei❆] n. 车床latitude ['l✌titju:d] n. 自由lay out 布置limitation[limi'tei☞☜n]n. 限度,限制,局限(性)local intelligence 局部智能locate [l☜u'keit] vt. 定位logic ['l d✞ik] n. 逻辑longitudinal['l nd✞☜'tju:dinl]a. 纵向的longitudinally['l nd✞☜'tju:dinl]a. 纵向的look upon 视作, 看待lubrication[lju:bri'kei☞☜n ]n. 润滑Mmachine shop 车间machine table 工作台machining[m☜'☞i:ni☠] n. 加工made-to-measure 定做maintenance['meintin☜ns]n.维护,维修majority[m☜'d✞a:riti] n.多数make use of 利用male[meil] a. 阳的, 凸形的malfunction['m✌l'f✈☠☞☜n]n. 故障mandrel['m✌dtil] n.心轴manifestation[m✌nif☜s'tei☞☜n] n. 表现, 显示massiveness ['m✌sivnis ]厚实,大块measure['me✞☜] n. 大小, 度量microcomputer 微型计算机microns['maikr n] n.微米microprocessor 微处理器mild steel 低碳钢milling machine 铣床mineral['min☜r☜l] n.矿物, 矿产minimise['minimaiz]v.把···减到最少, 最小化minute['minit] a.微小的mirror image 镜像mirror['mir☜] n. 镜子MIT (Massachusetts Institute ofTechnology) 麻省理工学院moderate['m d☜rit]adj. 适度的modification [m difi'kei☞☜n ]n. 修改, 修正modulus['m djul☜s] n.系数mold[m☜uld]n. 模, 铸模, v. 制模, 造型monitor ['m nit☜ ] v. 监控monograph['m n☜♈ra:f]n. 专著more often than not 常常motivation[m☜uti'vei☞☜n]n. 动机mould split line 模具分型线moulding['m☜udi☠] n. 注塑件move away from 抛弃multi-imprssion mould多型腔模Nnarrow['n✌r☜u] a. 狭窄的NC (numerical control ) 数控nevertheless[ nev☜❆☜'les]conj.,adv.然而,不过nonferrous['n n'fer☜s]adj.不含铁的, 非铁的normally['n :mli] adv.通常地novice['n vis]n. 新手, 初学者nozzle['n zl] n. 喷嘴, 注口numerical [nju'merikl]n. 数字的Oobjectionable [☜b'd✞ek☞☜bl]adj. 有异议的,讨厌的observe[☜b'z☜:v] vt. 观察obviously [' bvi☜sli]adv 明显地off-line 脱机的on-line 联机operational [ p☜'rei☞☜nl]adj. 操作的, 运作的opportunity[ p☜'tju:niti]n. 时机, 机会opposing[☜'p☜uzi☠]a. 对立的, 对面的opposite[' p☜zit]n. 反面a.对立的,对面的optimization [ ptimai'zei☞☜n]n. 最优化orient [' :ri☜nt] vt. 确定方向orthodox [' : ☜d ks]adj. 正统的,正规的overall['☜uv☜r :l]a. 全面的,全部的overbend v.过度弯曲overcome[☜uv☜'k✈m]vt. 克服, 战胜overlaping['☜uv☜'l✌pi☠]n. 重叠overriding[☜uv☜'raidi☠]a. 主要的, 占优势的Ppack[p✌k] v. 包装package ['p✌kid✞] vt.包装pallet ['p✌lit] n.货盘panel ['p✌nl] n.面板paraffin['p✌r☜fin] n. 石蜡parallel[p✌r☜lel] a.平行的penetration[peni'trei☞☜n ]n.穿透peripheral [p☜'rif☜r☜l] adj外围的periphery [p☜'rif☜ri] n. 外围permit[p☜'mit] v. 许可, 允许pessure casting 压力铸造pillar['pil☜] n. 柱子, 导柱pin[pin] n. 销, 栓, 钉pin-point gate 针点式浇口piston ['pist☜n] n.活塞plan view 主视图plasma['pl✌zm☜] n. 等离子plastic['pl✌stik] n. 塑料platen['pl✌t☜n] n. 压板plotter[pl t☜] n. 绘图机plunge [pl✈nd✞] v翻孔plunge[pl✈nd✞] v.投入plunger ['pl✈nd✞☜ ] n. 柱塞pocket-size 袖珍portray[p :'trei] v.描绘pot[p t] n.壶pour[p :] vt. 灌, 注practicable['pr✌ktik☜b]a. 行得通的preferable['pref☜r☜bl]a.更好的, 更可取的preliminary [pri'limin☜ri]adj 初步的,预备的press setter 装模工press[pres]n. 压,压床,冲床,压力机, prevent [pri'vent] v. 妨碍primarily['praim☜rili]adv. 主要地procedure[pr☜'si:d✞☜]n. 步骤, 方法, 程序productivity.[pr☜ud✈k'tiviti] n. 生产力profile ['pr☜ufail] n. 轮廓progressively[pr☜'♈resiv]ad. 渐进地project[pr☜'d✞ekt] n.项目project[pr☜'d✞ekt] v. 凸出projection[pr☜'d✞ek☞☜n]n. 突出部分proper['pr p☜] a. 本身的property['pr p☜ti] n.特性prototype ['pr☜ut☜taip] n. 原形proximity[pr k'simiti] n.接近prudent['pru:d☜nt] a. 谨慎的punch [p✈nt☞] v. 冲孔punch shapper tool 刨模机punch-cum-blanking die凹凸模punched tape 穿孔带purchase ['p☜:t☞☜s] vt. 买,购买push back pin 回程杆pyrometer[pai'n mit☜] n. 高温计Qquality['kwaliti] n. 质量quandrant['kw dr☜nt] n. 象限quantity ['kw ntiti] n. 量,数量quench[kwent☞] vt. 淬火Rradial['reidi☜l] adv.放射状的ram [r✌m] n 撞锤.rapid['r✌pid]adj. 迅速的rapidly['r✌pidli] adv. 迅速地raster['r✌st☜] n. 光栅raw [r :] adj. 未加工的raw material 原材料ream [ri:m] v 铰大reaming[ri:mi☠]n. 扩孔, 铰孔recall[ri'k :l] vt. 记起, 想起recede [ri'si:d] v. 收回, 后退recess [ri'ses]n. 凹槽,凹座,凹进处redundancy[ri'd✈nd☜nsi]n. 过多re-entrant 凹入的refer[ri'f☜:]v. 指, 涉及, 谈及,reference['ref☜r☜ns]n.参照,参考refresh display 刷新显示register ring 定位环register['red✞st☜]v. 记录, 显示, 记数regrind[ri:'♈aind](reground[ri:'gru:nd]) vt. 再磨研relative['rel☜tiv]a. 相当的, 比较的relay ['ri:lei] n. 继电器release[ri'li:s] vt. 释放relegate['rel☜geit] vt.把··降低到reliability [rilai☜'biliti] n. 可靠性relief valves 安全阀relief[ri'li:f] n. 解除relieve[ri'li:v ] vt.减轻, 解除remainder[ri'meind☜]n. 剩余物, 其余部分removal[ri'mu:vl] n. 取出remove[ri'mu:v] v. 切除, 切削reposition [rip☜'zi☞☜n]n. 重新安排represent[ repri'zent☜]v 代表,象征reputable['repjut☜bl]a. 有名的, 受尊敬的reservoir['rez☜vwa: ]n.容器, 储存器resident['rezid☜nt] a. 驻存的resist[ri'zist] vt. 抵抗resistance[ri'zist☜ns]n.阻力, 抵抗resolution[ rez☜'lu:☞☜n] n.分辨率respective[ri'spektiv]a. 分别的,各自的respond[ris'p nd]v. 响应, 作出反应responsibility[risp ns☜'biliti]n. 责任restrain[ris'trein] v. 抑制restrict [ris'trikt] vt 限制,限定restriction[ris'trik☞☜n] n. 限制retain[ri'tein] vt.保持, 保留retaining plate 顶出固定板reveal [ri'vil] vt.显示,展现reversal [ri'v☜sl] n. 反向right-angled 成直角的rigidity[ri'd✞iditi] n. 刚度rod[r d] n. 杆, 棒rotate['r☜uteit] vt.(使)旋转rough machining 粗加工rough[r✈f] a. 粗略的routine [ru:'ti:n] n. 程序rubber['r✈b☜] n.橡胶runner and gate systems 流道和浇口系统Ssand casting 砂型铸造satisfactorily[ s✌tis'f✌ktrili] adv. 满意地saw[a :] n. 锯子scale[skeil] n. 硬壳score[sk :] v. 刻划scrap[skr✌p]n. 废料, 边角料, 切屑, screwcutting 切螺纹seal[si:l] vt.密封secondary storagesection cutting plane 剖切面secure[si'kju☜] v. 固定secure[si'kju☜]vt. 紧固,夹紧,固定segment['se♈m☜nt] v. 分割sensitive['sensitiv] a. 敏感的sequence ['si:kw☜ns] n. 次序sequential[si'kwen☞☜l]a. 相继的seriously['si☜ri☜sli] adv.严重地servomechanism['s☜:v☜'mek☜nizm] n.伺服机构Servomechanism aboratoies 伺服机构实验室servomotor ['s☜:v☜m☜ut☜] n. 伺服马达setter ['set☜] n 安装者set-up 机构sever ['sev☜] v 切断severity [si'veriti] n. 严重shaded[☞✌did] adj. 阴影的shank [☞✌☠k] n. 柄.shear[☞i☜] n.剪,切shot[☞t] n. 注射shrink[☞ri☠k] vi. 收缩side sectional view 侧视图signal ['si♈nl] n. 信号similarity[simi'l✌riti] n.类似simplicity[sim'plisiti] n. 简单single-point cutting tool 单刃刀具situate['sitjueit] vt. 使位于, 使处于slide [slaid] vi. 滑动, 滑落slideway['slaidwei] n. 导轨slot[sl t] n. 槽slug[sl✈♈] n. 嵌条soak[s☜uk] v. 浸, 泡, 均热software ['s ftw☪☜] n. 软件solid['s lid] n.立体, 固体solidify[s☜'lidifai] vt.vi. (使)凝固, (使)固solution[s☜'lu:☞☜n] n.溶液sophisiticated [s☜'fistikeitid]adj.尖端的,完善的sound[saund] a. 结实的, 坚固的)spark erosion 火花蚀刻spindle['spindl] n. 主轴spline[splain] n.花键split[split] n. 侧向分型, 分型spool[spu:l] n. 线轴springback n.反弹spring-loaded 装弹簧的sprue bush 主流道衬套sprue puller 浇道拉杆square[skw☪☜] v. 使成方形stage [steid✞] n. 阶段standardisation[ st✌nd☜dai'zei☞☜n] n. 标准化startling['sta:tli☠]a. 令人吃惊的steadily['sted☜li ] adv. 稳定地step-by-step 逐步stickiness['stikinis] n.粘性stiffness['stifnis] n. 刚度stock[st k] n.毛坯, 坯料storage tube display储存管显示storage['st :rid✞] n. 储存器straightforward[streit'f :w☜d]a.直接的strain[strein] n. 应变strength[stre☠] n.强度stress[stres] n. 压力,应力stress-strain 应力--应变stretch[stret☞] v.伸展strike [straik] vt. 冲击stringent['strind✞☜nt ] a.严厉的stripper[strip☜] n. 推板stroke[strouk] n. 冲程, 行程structrural build-up 结构上形成的sub-base 垫板subject['s✈bd✞ikt] vt.使受到submerge[s☜b'm☜:d✞] v.淹没subsequent ['s✈bsikwent] adj.后来的subsequently ['s✈bsikwentli]adv. 后来, 随后substantial[s☜b'st✌n☞☜l] a.实质的substitute ['s✈bstitju:t] vt. 代替,.替换subtract[s☜b'tr✌kt] v.减, 减去suitable['su:t☜bl]a. 合适的, 适当的suitably['su:t☜bli] ad.合适地sunk[s✈☠k](sink的过去分词)v. 下沉, 下陷superior[s☜'pi☜ri☜] adj.上好的susceptible[s☜'sept☜bl]adj.易受影响的sweep away 扫过symmetrical[si'metrikl]a. 对称的synchronize ['si☠kr☜naiz]v. 同步,同时发生Ttactile['t✌ktail] a. 触觉的, 有触觉的tailstock['teilst k] n.尾架tapered['teip☜d] a. 锥形的tapping['t✌pi☠] n. 攻丝technique[tek'ni:k] n. 技术tempering['temp☜r☠] n.回火tendency['tend☜nsi]n. 趋向, 倾向tensile['tensail]a. 拉力的, 可拉伸的拉紧的, 张紧的tension ['ten☞☜n] n.拉紧,张紧terminal ['t☜:m☜nl ] n. 终端机terminology[t☜:mi'n l☜d✞i ] n. 术语, 用辞theoretically [ i:☜'retikli ] adv.理论地thereby['❆☪☜bai] ad. 因此, 从而thermoplastic[' ☜:m☜u'pl✌stik] a. 热塑性的,n. 热塑性塑料thermoset[' ☜:m☜set] n.热固性thoroughly[' ✈r☜uli] adv.十分地, 彻底地thread pitch 螺距thread[ red] n. 螺纹thrown up 推上tilt [tilt] n. 倾斜, 翘起tolerance ['t l☜r☜ns] n..公差tong[t ☠] n. 火钳tonnage['t✈nid✞]n. 吨位, 总吨数tool point 刀锋tool room 工具车间toolholder['tu:lh☜uld☜]n. 刀夹,工具柄toolmaker ['tu:l'meik☜]n 模具制造者toolpost grinder 工具磨床toolpost['tu:l'p☜ust] n. 刀架torsional ['t :☞☜nl] a扭转的.toughness['t fnis] n. 韧性trace [treis] vt.追踪tracer-controlled milling machine仿形铣床transverse[tr✌ns'v☜:s]a. 横向的tray [trei] n. 盘,盘子,蝶treatment['tri:tm☜nt] n.处理tremendous[tri'mend☜s]a. 惊人的, 巨大的trend [trend] n.趋势trigger stop 始用挡料销tungsten['t✈☠st☜n] n.钨turning['t☜:ni☠] n.车削twist[twist ] v.扭曲,扭转two-plate mould双板式注射模Uultimately['✈ltimitli] adv终于.undercut moulding侧向分型模undercut['✈nd☜k✈t]n. 侧向分型undercut['✈nd☜k✈t] n. 底切underfeed['✈nd☜'fi:d]a 底部进料的undergo[✈nd☜'♈☜u] vt. 经受underside['✈nd☜said]n 下面,下侧undue[✈n'dju:]a.不适当的, 过度的uniform['ju:nif :m]a. 统一的, 一致的utilize ['ju:tilaiz] v 利用Utopian[ju't☜upi☜n]adj. 乌托邦的, 理想化的Vvalve[v✌lv] n. 阀vaporize['veip☜raiz]vt.vi. 汽化, (使)蒸发variation [v☪☜ri'ei☞☜n] n.变化various ['v☪☜ri☜s]a. 不同的,各种的,vector feedrate computation向量进刀速率计算vee [vi:] n. v字形velocity[vi'l siti] n.速度versatile['v☜s☜tail]a. 多才多艺的,万用的vertical['v☜:tikl] a. 垂直的via [vai☜] prep.经,通过vicinity[v☜'siniti] n. 附近viewpoint['vju:p int] n. 观点Wwander['w nd☜] v. 偏离方向warp[w :p] v. 翘曲washer ['w ☞☜] n. 垫圈wear [w☪☜] v. 磨损well line 结合线whereupon [hw☪☜r☜'p n]adv. 于是winding ['waindi☠] n. 绕, 卷with respect to 相对于withstand[wi❆'st✌nd]vt. 经受,经得起work[w☜:k] n. 工件workstage 工序wrinkle['ri☠kl] n.皱纹vt.使皱Yyield[ji:ld] v. 生产Zzoom[zu:] n. 图象电子放大。
桥梁建设英语In the realm of civil engineering, the art of constructing bridges is nothing short of a marvel that intertwines human ingenuity with the natural landscape. Imagine standing at the edge of a vast chasm, pondering the challenge of connecting two distant points, and then behold, a bridge emerges, a testament to the power of engineering and the human spirit. The process of bridge construction is a symphony of meticulous planning, cutting-edge technology, and sheer determination.Starting with the conceptualization, engineers and architects collaborate to design a structure that not only meets the functional requirements but also stands as an architectural landmark. The design phase is a meticulous dance of calculations, where forces of gravity, tension, and compression are carefully considered to ensure the bridge's stability and longevity.Once the design is finalized, the real work begins. The site preparation involves clearing the area, setting up access routes, and ensuring the ground is ready to support the massive structure. Foundations are laid, often requiring deep excavations and the installation of piles that reach down into the earth's core, providing a solid base for the bridge's weight.The construction itself is a spectacle of coordinationand precision. Steel beams, concrete slabs, and cables are assembled with the precision of a Swiss watch, each piece playing a crucial role in the overall integrity of the bridge. The use of heavy machinery, such as cranes and piling rigs,is a common sight, as they hoist and position the massive components into place.Safety is paramount during bridge construction. Workersare equipped with the necessary personal protective equipment, and safety protocols are strictly enforced to prevent accidents. As the bridge takes shape, it becomes a beacon of progress, a symbol of overcoming obstacles, and a gateway to new opportunities.Upon completion, the bridge stands as a testament to the hard work and dedication of all those involved. It is notjust a passage over an obstacle; it is a connection between communities, a link to economic growth, and a bridge to the future. The engineering marvel that is a bridge is a story of overcoming challenges, of uniting people, and of creating pathways where none existed before.。
模具机械专用词汇中英文对照模具机械专用词汇中英文对照来源:未知时间:2006—10-26 15:20:49 浏览数:【收藏这篇文章】abrasive grinding 强力磨削abrasive 磨料的,研磨的absence 不在,缺席accesssory 附件accommodate 适应accordingly 因此,从而,相应地accuracy 精度,准确性actuate 开动(机器),驱动adequate 足够的adhesive 粘合剂adjacent 邻近的adopt 采用advance 进步advisable 可取的agitate 摇动a large extent 很大程度algorithm 算法align 定位,调准alignment 校直all—too-frequent 频繁allowance 容差,余量alternate 交替,轮流alternatively 做为选择,也许aluminiun 铝ample 充足的analysis 分析ancillary 补助的,副的angular 有角的annealing 退火aperture 孔applied loads 作用力appropriate 适当的arc 弧,弓形arise 出现,发生arrange 安排article 制品,产品ascertain 确定,查明assemble 组装attitude 态度auxiliary 辅助的avoid 避免axis 轴axle 轮轴,车轴alternative 替换物backup 备份batch 一批bearing 轴承,支座bed 床身behavior 性能bench—work 钳工工作bend 弯曲beneath 在???下bin 仓,料架blank 坯料blank 冲裁,落料blanking 落料模blast 一阵(风) blemish 缺点,污点bolster 模座,垫板boring 镗削,镗孔bracket 支架brass 黄铜break down 破坏breakage 破坏brine 盐水brittle 易碎的buffer 缓冲器built-in 内装的bulging 凸肚burr 毛刺bush 衬套by far ???得多,最by means of 借助于boost 推进cabinet 橱柜call upon 要求carbide 碳化物carburzing 渗碳carriage 拖板,大拖板carry along 一起带走carry down over 从???上取下carry out 完成case hardening 表面硬化case 壳,套cast steel 铸钢casting 铸造,铸件category 种类caution 警告,警示cavity and core plates 凹模和凸模板cavity 型腔,腔,洞centre—drilling 中心孔ceramic 陶瓷制品chain doted line 点划线channel 通道,信道characteristic 特性check 核算chip 切屑,铁屑chuck 卡盘chute 斜道circa 大约circlip (开口)簧环circuit 回路,环路circulate (使)循环clamp 夹紧clamp 压板clay 泥土clearance 间隙clip 切断,夹住cold hobbing 冷挤压cold slug well 冷料井collapse 崩塌,瓦解collapsible 可分解的combination 组合commence 开始,着手commence 开始commercial 商业的competitive 竞争的complementary 互补的complexity 复杂性complication 复杂化compression 压缩comprise 包含compromise 妥协,折衷concern with 关于concise 简明的,简练的confront 使面临connector 连接口,接头consequent 随之发生的,必然的console 控制台consume 消耗,占用consummate 使完善container 容器contingent 可能发生的CPU (central processing unit) 中央处理器conventional 常规的converge 集中于一点conversant 熟悉的conversion 换算,转换conveyer 运送装置coolant 冷却液coordinate (使)协调copy machine 仿形(加工)机床core 型芯,核心corresponding 相应的counteract 反作用,抵抗couple with 伴随contour 轮廓crack (使)破裂,裂纹critical 临界的cross—hatching 剖面线cross-section drawn 剖面图cross—slide 横向滑板CRT (cathoder—ray tube)阴极射线管crush 压碎cryogenic 低温学的crystal 结晶状的cubic 立方的,立方体的cup (使)成杯状,引伸curable 可矫正的curvature 弧线curve 使弯曲cutter bit 刀头,刀片cyanide 氰化物complicated 复杂的dash 破折号daylight 板距decline 下落,下降,减少deform (使)变形demonstrate 证明depict 描述deposite 放置depression 凹穴descend 下降desirable 合适的detail 细节,详情deterioration 退化,恶化determine 决定diagrammmatic 图解的,图表的dictate 支配die 模具,冲模,凹模dielectric 电介质die-set 模架digital 数字式数字dimensional 尺寸的,空间的discharge 放电,卸下,排出discharge 卸下discrete 离散的,分立的dislodge 拉出,取出dissolution 结束distinct 不同的,显著的distort 扭曲distort (使)变形,扭曲distributed system 分布式系统dowel 销子dramaticlly 显著地drastic 激烈的draughting 绘图draughtsman 起草人drawing 制图drill press 钻床drum 鼓轮dual 双的,双重的ductility 延展性dynamic 动力的edge 边缘e。
岩土力学英文版IntroductionGeotechnical Engineering, also known as Soil Mechanics or Rock Mechanics, is a branch of civil engineering that deals with the behavior of soil and rock materials under various conditions. It is an important field of study as it helps engineers understand the properties and characteristics of these materials, which in turn enables them to design and construct safe and stable structures.Soil MechanicsSoil Mechanics is the study of the behavior of soil materials, including its formation, classification, and properties. Various aspects of soil mechanics are essential in geotechnical engineering, such as soil compaction, permeability, and soil stability.Soil formation is a complex process that involves the weathering and erosion of existing rocks, resulting in the formation of different soil types. The composition and particle size distribution of soil influence its properties, including its bearing capacity, shear strength, and compressibility.Soil classification is an important step in understanding the behavior of various soil types. The Unified Soil Classification System, which categorizes soils based on their particle size and organic content, is widely used in geotechnical engineering. Common soil types include gravel, sand, silt, clay, and organic soils.Understanding soil properties is crucial in determining its suitability for construction projects. Soil compaction refers to the process of densifying soil by applying mechanical force, ensuring stability and reducing settlement. Permeability is the ability of soil to transmit fluids such as water or gas, which is essential in designing drainage systems.Shear strength is another critical property of soil, as it determines its ability to resist sliding or deformation. Soil stability can be assessed through various laboratory tests, such as direct shear tests or triaxial tests, which simulate the conditions that soil experiences in real-world applications.Rock MechanicsRock Mechanics, on the other hand, is the study of thebehavior of rock materials, including its strength, deformation, and stability. It plays a crucial role in the design and construction of underground structures, such as tunnels and mines, as well as in slope stability analysis.Rock strength is an essential characteristic to consider when designing structures in rock formations. Different rock types have varying strength properties, with factors such as mineral composition, rock structure, and geological history influencing their behavior. Lab testing, such as uniaxial compression tests or point load tests, is typically conducted to determine the rock's strength.Rock deformation refers to the response of rock materials to applied stresses, including compression, tension, and shear. Understanding the deformation behavior of rock is crucial in predicting stability and designing support systems for underground excavations. Slope stability analysis is a critical aspect of geotechnical engineering, especially in hilly or mountainous regions. An unstable slope can lead to landslides or slope failures with disastrousconsequences. Various methods, including limit equilibrium analysis and numerical modeling, are used to assess slope stability and design appropriate reinforcement measures.ConclusionGeotechnical engineering plays a vital role in the construction industry as it helps design safe and stable structures by understanding the behavior of soil and rock materials. Soil mechanics focuses on the properties and characteristics of soil, including its formation, classification, and behavior under various conditions. Rock mechanics, on the other hand, studies the properties of rock materials such as strength, deformation, and stability. These fields of study are essential for engineers to ensure the safety and integrity of construction projects.。
22 a couple of 几个,二,三个13 a large proportion 大部分22 a range of 一系列的22 a variety of 种种6 abutment 支座,墩座2 academic 学术的13 accommodate 接纳12 accomplishment 成就22 acpuire 获得15 adequately 充分的,适当的5 adhere 粘附2 adjacent 相邻的5 adjacent 相邻的4 adjust to 适应7 aerial 空中的,航空的16 aggradation 填积2 air mass 气团24 alignment 定线23 allowance 余地,容许量4 alluvial fan 冲积扇1 alternative 比较方案7 altitude 高度15 amplitude 振幅15 anabranches 重汇支流6 anchorage 锚固25 anchorage 锚地3 ancient 古老的5 angulardeformation角变形23 annex 附录24 anticipate 预期3 appearance 外观5 application offorce施力7 approach 接近,方法16 arroyo 旱谷4 as a result of 作为···的结果25 asphalt 沥青23 astronomical 天文的13 Atlantic Ocean 大西洋12 auxiliary 辅助的23 availability 有效性13 Baltic Sea 波罗的海15 bankfull stage 漫滩水位4 bar 沙洲1 bar examination 律师合格考试24 be exposed to 受到23 be involved in 包括1 be responsible for 对···负责6 beam 梁,船宽24 beam 船宽,梁24 bear in mind 记住25 bedding 垫层15 bend 弯道23 berth 泊位7 blunder 大错6 bond 粘合力25 bonding 砌合15 braided 分叉的25 breakwater 防波堤12 breeding ground 繁殖场,育种场2 broad detail 概括性的24 bulk carriers 散装货船23 bulk vessel 散装货船23 bulletin 公告6 cable 缆25 caliber 质量3 canyon 峡谷23 capital 资本1 career 业务4 catastrophic 灾难性的4 catch fire 着火12 cement 水泥2 channel 槽,水道,航道,4 characteristic 特征1 chemistry 化学4 chlorine 氯2 circulation 循环1 civil engineering 土木工程4 civilization 文明,文化7 classification 分类24 clearance 净空16 coarse 粗糙的22 coast zone 海岸带25 cobber 伙伴5 coefficient 系数12 cofferdam 围堰5 cohesion 内聚力4 commodity 商品 12 compact 压实 23 comparative 相当的 16 composition 组成 6 compression压缩1 computer programing 编制计算机程序 15 concave凹的 2 concern themselves with关心22 concomitant with 与...相伴的 1 concrete 混凝土2 condense 冷凝25 confine 限制16 confluence 汇合点,汇流点 23 congress 会议24 conservative 保守的2 constant 不变的,常数 1 construction 施工 23 container集装箱23 container berth 集装箱泊位23 container vessel 集装箱船24 container vessel 集装箱船7 contaminate 混淆,沾染13 continent 大陆25 coping顶层15 correspondingly 相应的25 course 层6 crack 开裂4 cradle 摇篮,发源地 6 creep 蠕动,徐变25 crib木笼24 criteria 标准4 crop庄稼,植物 24 cross wind 横风24 crosscurrent 横流6 cross —section 横截面,断面,剖面 12 crush碎16 cultivation 耕作16 cure治愈1 curriculum 全部课程 7 curvature 曲率12 cut stone 琢石25 cyclopean 毛石1 data数据 24dead weight (=dwt) 载重吨位,载重量 23 deficit亏损 4 definite integral 定积分 6 deflection 挠度 21 degenerate 退化7 delineate描绘,勾画 16 delivery ratio 递送比 7 demonstrate 表明 21 denote 表示 5density密度 24Department of Commerce 商业部4 deposits 沉积物 16desert 沙漠 1 design设计6 design load 设计载荷 22 destructive 破坏性的 7detect察觉 2 determinant 决定因素 24 dictate支配 24 differentation 区别 5 dimension 量纲,维5diminish [di'mini ʃ] 减小3 dirt泥土4 discharge 排放,泄放 25 disintegrate 崩解2 disperse 分散,散发 1disrepair 失修15 dissipate 消散,消耗 21 distinction 区别 21 distinguish 区别 2distribution 分布12 divert 改道,转向 2 domain 范围,领域 25 dowel 销钉 15 downcut 下切23 dredge 挖泥,疏浚 13 dredger 挖泥船,疏浚船 2drought 干旱12 duct 导管,通道 4 dump倾倒 22 duration历时 2earth science地球科学1 earth-fillembankment dam填土坝12 earthquake 地震5 edge condition 边界条件2 elevation 海拔,高程5 elevation 高程13 eliminate 排除21 ellipse 椭圆21 ellipse 椭圆12 embankment 填土堤坝,路基12 embankment dam 填筑坝,土石坝6 embed 埋置2 embrace 包括12 enclosure 围栏21 energy flux 能通量,能流22 energy-transfer 能量传递1 environment 环境5 equivalent[i’kwivələnt]相当于16 erosion 侵蚀2 erraatic 不稳定的7 error 误差15 evaluation 估计值2 evaporation 蒸发2 eventually 最终的16 evidence 证据21 exception 例外22 expanse 浩瀚,广阔24 explicit 明显的,明确的,显式24 explicitly 明显的4 exploration 探险3 explore 探险23 exposure 方位24 extrapollation 外延,外推2 extremes 极值,极端情况12 facility 设施,设备6 failure 破坏6 faulty 毛病1 feasibility 可行性7 features 特征,地物4 fertile 肥沃的4 fertilizer 肥料22 fetch 风浪区22 fetch area 受风水域7 field work 野外作业,外业1 finance 资助2find its way 。
基于网格分割的三维模型轻量化算法及构建金伟祖;潘伟龙【摘要】三维建模指通过三维制作软件构建出具有三维数据的模型.三维建模技术可以大大增加物体的真实感.传统的三维建模技术首先对真实物体进行抽象,用多边形构造物体的三维模型,往往伴随着缓慢的生成速度.为保证三维建模以及模型展现的效率,三维模型的轻量化技术必不可少.基于重复的网格分割算法通过发现重复单元来提高模型展现效率,但是该算法对于单连通模型的效果不佳,因此在引入几何描述符的基础上,提出一种基于网格分割的模型轻量化算法,该算法包括网格分割,单元匹配和网格重建三个环节,算法能够发现经过旋转,缩放,平移后的重复单元,很好的提高轻量化效率.实验表明,该分割算法在网格轻量化中可以取得较高的存储压缩率.【期刊名称】《实验室科学》【年(卷),期】2015(018)005【总页数】4页(P20-23)【关键词】三维模型;模型轻量化;重复单元;网格分割【作者】金伟祖;潘伟龙【作者单位】同济大学软件学院,上海201804;同济大学软件学院,上海201804【正文语种】中文【中图分类】TP37三维模型特有的真实感,使得三维模型在各类应用决策中正得到广泛应用。
如何保证三维建模以及模型展现的效率至关重要,核心问题就是三维模型的轻量化,减少三维模型数据的存储量是研究的焦点之一。
本文具体分析了现有的、基于轻量化的模型分割算法,并分析它们各自的优劣和适用场景,最终在这些算法的基础上提出一种针对模型轻量化的网格分割算法,并以该算法为基础,提出了三维模型的轻量化构建步骤。
1 基于重复单元的轻量化研究重复单元是指模型中相同或相似的部分,这些相似的部分可以通过一系列几何变换来完成。
重复单元的轻量化是指通过提取模型中的重复单元,只保存一份模型数据和几何变换信息代替整个模型,达到大幅度减少数据存储量的目的。
在三维模型中,常常含有很多相似或者可以通过几何变换而变得相似的部分,例如一个建筑中含有很多窗口,而这些窗口模型之间是可以通过平移来转换的,即可以通过保存一份窗口模型的数据和平移信息来实现对窗口的保存,这样就可以使用更少的数据来保存窗口模型。
Engineering MechanicsEngineering mechanics is a fundamental discipline within the field of engineering that deals with the behavior of solid bodies subjected to external forces. It is a branch of mechanics that focuses on the study of the motion and deformation of objects under the influence of forces. This field plays a crucial role in various engineering applications, such as designing structures, machines, and systems that can withstand different types of loads and forces. One of the key concepts in engineering mechanics is statics, which deals with the equilibrium of bodies under the action of forces. Statics is essential for analyzingstructures and determining the forces acting on them to ensure stability and safety. By studying statics, engineers can design structures that can supporttheir own weight as well as any additional loads they may be subjected to during their lifespan. Another important aspect of engineering mechanics is dynamics, which deals with the motion of bodies under the influence of forces. Dynamics is crucial for understanding how objects move and deform in response to external forces, such as gravity, friction, and inertia. By studying dynamics, engineers can predict the behavior of systems and machines, allowing them to optimize their performance and efficiency. In addition to statics and dynamics, engineering mechanics also encompasses the study of materials and their mechanical properties. Understanding how materials behave under different conditions is essential for designing structures and machines that can withstand the stresses and strains they will be subjected to during operation. By studying materials science, engineers can select the most suitable materials for a given application, ensuring the durability and reliability of their designs. Furthermore, engineering mechanics plays a vital role in the field of civil engineering, where structures such as buildings, bridges, and dams are designed and constructed to withstand various loads and environmental conditions. Civil engineers rely on the principles of engineering mechanics to ensure the safety and stability of their projects, taking into account factors such as wind, earthquakes, and temperature variations. Moreover, engineering mechanics is also crucial in the field of mechanical engineering, where machines and mechanical systems are designed to performspecific tasks efficiently and reliably. Mechanical engineers use the principlesof engineering mechanics to analyze the forces and stresses acting on machine components, ensuring that they can withstand the forces generated during operation. By applying the principles of engineering mechanics, mechanical engineers can design machines that are safe, durable, and efficient. In conclusion, engineering mechanics is a fundamental discipline within the field of engineering that plays a crucial role in designing structures, machines, and systems that can withstand different types of loads and forces. By studying statics, dynamics, and materials science, engineers can analyze the behavior of objects under the influence of forces, ensuring the safety, stability, and reliability of their designs. Whetherin civil engineering or mechanical engineering, the principles of engineering mechanics are essential for ensuring the success of engineering projects and the safety of the built environment.。
标准与质量中图分类号:TU528.07文献标识码:A文章编号:1001-6945(2023)06-44-02本栏编辑:冯凯混凝土是现代建设工程中必不可少的原材料之一,其抗压强度是混凝土最基本的性能之一,也是混凝土最重要的物理性能;混凝土物理性能与建设工程中混凝土工程的质量和耐久性能息息相关,直接关系整个建设工程的质量安全。
混凝土抗压强度检测是混凝土性能检测中最为基本的项目,对于混凝土抗压强度检测,许多学者均对其影响因素开展了研究,主要是混凝土成型、浇筑、养护等方面,但对于混凝土抗压强度的不确定度评定的研究相对较少[1-2]。
不确定度评定是通过一种分析各种实验因素对实验结果造成偏差的方法[3]。
本文通过开展混凝土抗压强度检测中不确定度评定,分析影响检测过程中各种因素对检测结果造成的影响程度分析,可以明确影响较大因素,进而为检测准确性的提升提供技术支撑。
1检测过程测量对象:混凝土试件。
测量依据:GB/T 50081-2019《混凝土物理力学性能试验方法标准》[4];GB/T 27418-2017《测量不确定度评定和表示》[5]。
测量仪器:微机控制电液伺服万能试验机(量程1000kN ,精度1级),游标卡尺(量程200mm ,分度值0.02mm ),塞尺(精度0.01mm )。
其中,扩展不确定度及扩展因子由设备计量单位提供。
测量环境:温度20℃,湿度60%。
试验方法:按照GB/T 50081-2019《混凝土物理力学性能试验方法标准》中轴向抗压强度要求,使用游标卡尺测量150mm×150mm×150mm 的立方体混凝土试块尺寸,精确至0.1mm 。
采用钢板尺和塞尺测量试件承压面的平面度,精确至0.01mm 。
将试件放置在承压板中心处,设置加荷速度为0.65MPa/s ,直至试件破坏,记录破坏力值,按照下式计算混凝土试件抗压强度。
考虑试件制作偏差可能导致评定无效,选取10块试件进行评定。
Value Engineering1建筑大体积混凝土施工现状1.1需要加强施工管理人员和操作人员的技术水平施工现场质量管理的决定性因素之一是施工管理人员的专业能力和责任感。
我国目前的建设过程缺乏高素质的管理人才,管理人才的技术水平低下。
他们无法完全理解设计图纸上加密节点的详细信息(例如梁和箍筋)的重要性。
一旦被管理人员忽略,施工人员可能会错过并减少钢筋的放置,这会影响施工质量并导致潜在的安全隐患。
但是,在施工过程中,大多数建筑工人是农民工,缺乏安全生产所需的专业技能和知识。
在施工过程中,无法按照相应的施工规定进行操作,从而难以保证工程质量。
1.2影响温度的因素在房屋建筑工程中浇筑大体积混凝土结构时,温度的影响很容易引起结构裂缝,直接影响工程的整体质量。
更确切地说,第一个是水泥的水热现象。
水泥在水合过程中会散发一定量的热量。
由于大体积混凝土结构的壁厚,大量的热量积聚在结构内部。
另外,混凝土的导热性差并且不能在短时间内散发热量。
最终,混凝土在固化过程中会产生巨大的拉伸应力,超过其自身的拉伸强度,会导致裂纹。
其次,外界温度的影响。
在建造大体积混凝土结构时,由于外部环境温度的影响,很可能会发生裂缝。
混凝土的温度将随着外部温度的变化而变化,特别是当外部温度急剧下降时,混凝土结构的内部和表面之间将存在较大的温差。
温差越大,热应力越大,这将导致混凝土结构在硬化过程中收缩。
一系列问题,例如不均匀和不均匀膨胀,会导致裂纹。
1.3不平等地基大体积混凝土结构主要在房屋建设项目中起一定作用,直接影响到工程的稳定性和安全性。
在实际施工中,由于对地基的处理不当,工程的地基部分在外力的作用下不均匀沉降,在混凝土结构中产生内应力。
安装基础时的质量一旦应力超过拉伸范围,就会出现裂纹。
诸如断裂等一系列问题严重影响了项目的稳定性,并降低了项目的整体性能和寿命。
2建筑中大体积混凝土施工技术2.1施工前准备在高层建筑进行大体积混凝土施工之前,应根据建筑物设计的要求选择材料。
To appear in Proc.11th ACM-SIAM Symp.on Discrete Algorithms,San Francisco,CA c SIAM2000.Engineering the Compression of Massive Tables:An Experimental Approach Adam L.Buchsbaum Donald F.Caldwell Kenneth W.Church Glenn S.FowlerS.MuthukrishnanAbstractWe study the problem of compressing massive tables.We devise anovel compression paradigm—training for lossless compression—which assumes that the data exhibit dependencies that can belearned by examining a small amount of training material.Wedevelop an experimental methodology to test the approach.Ourresult is a system,pzip,which outperforms gzip by factors oftwo in compression size and both compression and uncompressiontime for various tabular data.Pzip is now in production use in anAT&T network traffic data warehouse.1IntroductionWe study the problem of compressing massive tables,whicharises naturally in corporate data warehouses.Our goalis to provide a working system that can be put into pro-duction use and achieve100:1compression,in particular,one that can compress10s of TB of data into100s of GB.We devise a novel compression strategy—training for loss-less compression—which can leverage standard compressionmethods,and we demonstrate its effectiveness experimen-tally.In the process,we identify the requirements for ourcompression application and design algorithmic solutions tovarious technical problems.The system we built,pzip,cancompress1TB of data from AT&T’s network traffic datawarehouse into about28GB,small enough tofit on one PCdisk and a two-fold improvement over existing solutions(intime as well as space).Pzip is now in production use in thewarehouse.Our motivating applications are tables of traffic datafrom telecommunication networks.For example,the AT&Tvoice communications network generates a record of eachphone call it carries.A typical record consists of severalhundred bytes and depicts network-level information(e.g.,endpoint exchanges),time-stamp information(e.g.,startand end times),and billing-level information(e.g.,appliedtariffs).This application generates about1TB of data permonth and is just one example of the many different tablesthat AT&T and other corporations generate.Others includeswitch-and router-level traffic data,equipment sensor data(e.g.,alarm status),and credit card transaction records.These data have certain unifying characteristics:they haveand maintained.Finally,it is preferable that any solution be general,applying to the many different tables,or portions of tables,that are processed.Thus we cannot exploit domain-specific syntactic or semantic information.Studying massive tables is a new focus in compression research.To distinguish our context from extant ones,con-sider the relatedfield of database compression,where rela-tional data may be viewed as tables.This differs from our table compression problem in many ways.First,the goals are different.Database compression stresses the preserva-tion of indexing—the ability to retrieve an arbitrary record—under compression[7].Table compression does not re-quire indexing to be preserved.Next,the data are different. Database records are often dynamic,unlike table data,which have a write-once discipline.Databases consist of hetero-geneous data,possibly with several stringfields of variable length;table data are more homogeneous,withfixedfield lengths.Also,non-tabular databases are not routinely TBs in size.(An exception is NASA’s EOSDIS database[13], which anticipates processing1TB of satellite images every two weeks.)Finally,the approaches to database compression include lightweight techniques such as compressing each tu-ple by simple encodings[7,8]and tiling the entire table[8]. These approaches are not appropriate for table compression: the former is too wasteful,and the latter too expensive and cumbersome.Our contribution is a novel approach for the table com-pression problem:lossless compression via training.The idea is to construct a compression plan for the table by study-ing a very small training set off line.To do so,we assume that the data can be modeled by an underlying source that can be learned from a small sample.We further assume and exploit dependencies in the columns of the data in one of two ways:(1)implicitly,by grouping the columns that compress well together;and(2)explicitly,by determining a depen-dency tree among the columns.We then employ the com-pression plan on the entire dataset.To test our assumptions, we implement algorithms to construct compression plans on some training sets,and we compress test sets with respect to the plans.We compare the resulting compression to the straightforward approach of treating the tables as text and applying Lempel-Ziv compression[20,21].It will be clear that comparable performance would falsify our assumptions about the data dependencies.In all cases,however,our algorithms provide substantial compression improvements. While training has been applied to lossy compression,e.g., in speech coding[12,15],ours is thefirst known instance of applying training to lossless compression.For our primary application,compression exploiting implicit dependencies outperformed that using explicit de-pendencies.We have implemented the corresponding algorithm—optimum partitioning—in pzip,a fully work-ing software system for compressing table data,which has been deployed in the AT&T network traffic data warehouse. Pzip achieves factors of about2improvement in compres-sion size and both compression and uncompression time over gzip,the method previously used in this application.In Section2,we discuss the table compression problem further and define our assumptions regarding data dependen-cies.In Section3,we present technical problems that ex-ploit our assumptions,and we give algorithmic solutions to these problems.In Section4,we present our experimental results,and in Section5,we discuss the pzip system and some additional applications.In Section6,we summarize our contributions and present directions for future work.2Problem DiscussionOur input consists of a table,,of a large number of rows,each of length bytes.We define column to be the projection of the th byte of each row,for .(A byte is the smallest unit of data that can be easily and rapidly accessed;moreover,this level of granularity captures patterns among larger lexical units.)The table compression problem is to compress,such that the requirements discussed in Section1are satisfied.From an information-theoretic point of view,can be treated as a string,e.g.,of bytes in row-major order.It would thus suffice to perform Lempel-Ziv[20,21]or Huffman[9] compression,yielding provably optimal asymptotic perfor-mance in terms of certain ergodic properties of the source that generates the table.This does not,however,adequately solve the table compression problem.For specific classes of inputs,e.g.,tables of network traffic data,the optimality re-sults may not necessarily hold.In particular,the optimality results hold only with respect to compression methods that likewise treat as a(byte)string;i.e.,methods that do not account for complex dependencies in.Some compression does result,however,and we use this method as our bench-mark in Section4.We need a few technical definitions.Denote by the th column of.Denote by the interval of columns through of.Finally,denote by the size of the result of compressing some interval of columns,in row-major order,using an arbitrary butfixed compressor.2.1Assumptions.Our approach to the table compression problem assumes that there are dependencies among the columns of.In particular,we will consider dependencies of two types:combinational and differential.D EFINITION2.1.Two contiguous intervals of columnsand are combinationally dependent ifCombinational dependency is an implicit dependency between intervals.It merely expresses that intervals ofcolumns are dependent,without determining which columns are dependent on the others.D EFINITION2.2.Column is differentially dependent on column ifwhere is the column formed by taking the row-wise difference between columns and.Differential dependency is an explicit dependency be-tween columns,in that it determines which column is de-pendent on the other.In general,we might compress and by different methods,and we might consider other transformations.This does not affect the ensuing discussion.Our approach also makes the important assumption that the data is generated by some source that is well behaved, in particular,that dependencies(such as those above)among columns,if they exist,can be captured by examining a small amount of data,independent of the size of.3Algorithmic IssuesWe design compression schemes based on the assumptions embodied in Definitions2.1and2.2.3.1Combinational Approach.We can exploit combina-tional dependencies as follows.Consider a partition,,of into intervals, such that and.We refer to an intervalin as a class.We define the cost of to beThe goal isfind an optimum partition,i.e.,such that.We canfind an optimum partition as follows.Define to be the cost of an optimum partition of for ,and define.Then for,Assuming that the cost has been computed for all,we can compute(and the corresponding partition)in time by simple dynamic programming.We call this optimum partitioning.This gives the following compression plan for compressing:compress each class in the optimum partition independently in row-major order.We can further speed up the dynamic programming, under the assumption that there are many optimum or near optimum partitions for compressing,and that among these are some in which the classes are not too wide.To do so,we develop a“chunking”approach,in which,for some ,we divide the columns into pairwise-disjoint intervals of size at most each,and solve our general problem on each such interval.The running time becomes .We call this chunk partitioning,and it likewise returns a compression plan.3.2Differential Approach.We can exploit differential dependencies as follows.Consider a partition of the columns of into two sets,and.We treat the columns in as source columns and those in as derived columns.Given a mapping,we define the cost,to beThe goal is tofind a pair of minimum cost.This is precisely the facility location problem[17].We will assume that the differential cost is a metric.In general, this depends on the base compressor.We apply the simple, greedy algorithm for this problem[14].At any time,we have a candidate pair.We determine the smallest cost solution,,obtained by1.removing a column from,2.adding a column to,or3.substituting one of the columns in for one not in. (Ties are broken arbitrarily.)If,then we set and and iterate.Otherwise we are done.We call this greedy differential compression.Thefinal solution is roughly-optimal under the metric assumption [14].Better approximations[3,4,5,11]are known,but the greedy algorithm suffices for our purpose of testing the presence of differential dependencies.Greedy differential compression produces the follow-ing compression plan:compress each column in indepen-dently,and for each column,compress.3.3Lossless Compression via Training.Our overall ap-proach is thus the following.1.Select a small subset as training material.ing,compute a compression plan,,for byeither the combinational or differential approach.press with the compression plan.Our assumption that the amount of training data needed is independent of the size of implies that,once we havegenerated a compression plan,we can use it to compress future tables generated by the same source as.Training is thus an off-line procedure.So far,we have abstracted the base problem of comput-ing and.Rather than develop our own base compression method,we decided to use one of the stan-dard programs,which have already been well optimized: e.g.,compress[18,21],gzip[20],and vdelta[10]. Each is fast,on-line,and well-suited to our application.Of other available compressors,we note that PPM[6,19],which exploits context sensitivity and thus seems applicable to ta-ble data,and bzip[1]are too slow for our environment, although attempts have been made to tune PPM for speed at the expense of compression size[16].We therefore do not use bzip and PPM in our compression scheme,but we do compare our scheme against bzip and PPM by themselves. We note but do not consider in this paper hybrid approaches, in which we pick the best compressor for a given interval. We can even nest the differential approach within the combi-national approach.4Experiments4.1Methodology.We summarize our assumptions as fol-lows.1.Our data sets present combinational dependencies.2.The combinational approach is likely to induce some(near)optimum partition in which no class is wide.3.Our data sets present differential dependencies.4.The above dependencies can be detected with a smallamount of training data,independent of the size of.Wefix gzip as our underlying compression method. While this does not explore the range of possible base com-pressor options,it suffices to test our approach.As bench-mark R,we apply gzip to in row-major order,corre-sponding to the usage of gzip without off-line training;as benchmark C,we apply gzip to in column-major order, corresponding to the other extremal partition in which no combinational or differential dependencies exist.We thus designed experiments to compare the performance of1.optimum partitioning to the benchmarks,2.chunk partitioning to optimum partitioning,and3.the greedy differential compression to both optimumpartitioning and benchmark C.Each experiment has the potential to falsify one of our assumptions.If either benchmark outperforms optimum par-titioning(with respect to output size),then our data sets do not present combinational dependencies.If optimum parti-tioning significantly outperforms chunk partitioning,then all (near)optimum partitions must have at least one wide class. If benchmark C outperforms the greedy differential compres-sion,then our data sets do not present differential dependen-cies.We discuss testing assumption(4)below.For each experiment,we produced a compression plan by running the corresponding algorithm on a training data ing the resulting plan,we compressed a disjoint test data set,and we compared the compression performance (time and size)to that of the benchmark(s)for that goal. Although size of compressed output is the metric by which our assumptions can be falsified,we also measured running times,to assess the practicality of our methods.This method-ology extends to assess other,similar compression systems.We also varied the amount of training data available, to gauge the effect of training size on compression perfor-mance.This is only thefirst step in testing assumption(4). If we do not see compression performance stabilize at some point as we increase the amount of training data,then as-sumption(4)is likely falsified.After observing this stabi-lization,however,a second test will be required:namely, tofix the training set size above the point at which we ob-served stability and increase the test set size arbitrarily.If the relative performance of the compression systems being compared does not remain stable,again assumption(4)is likely falsified.Otherwise,we will have evidence supporting assumption(4).Although this second experiment remains to be conducted,based on our results we do not expect assump-tion(4)to be falsified.Finally,we compared the best results from the above experiments with isolated usages of compressors based on the Burrows-Wheeler transform[1]and prediction by par-tial match(PPM)[6,19].The goal was to assess empir-ically the benefit of our training scheme,which leverages standard compression technology,versus other methods that claim improved compression via sophisticated analyses of the source data.Data.We used100,000records from a network traffic data warehouse.Each record is781bytes and pertains to an individual network event.The warehouse receives approximately one billion records per month,so effective compression is critical to this application.From the781 columns of bytes,we extracted the90with the highest frequency:i.e.,the number of times the value of the byte changed as the column was scanned top-down.We explain this in Section5;basically,in our real application,the other691columns were compressed more effectively using incomparable methods.We divided the100,000(now90-byte)records into training and test sets.The training set was1/11th of the data;the test set was the rest.We chose the training set in two ways:(1)thefirst9091records,which we will call the ordered training set,and(2)a randomly chosen set of9091 records,which we will call the random training set.Eachway left the corresponding rest of the data as the test set. In all our experiments,we used the training sets to generate the corresponding compression plans,and we conducted the compression experiments using those plans and the test sets.Software.To run the experiments,we implemented the following tools.pin.Given a training set,pin computes a compression plan based on optimum partitioning.pzip.Given a compression plan computed by pin,pzip compresses a test set with respect to the plan.It encodes enough of the plan into the output(which is included in the output size results reported)so that,given a com-pressedfile,pzip will uncompress it without needing the original plan.(Pin and pzip actually form our working system,and we discuss them in greater detail in Section5.)colsel.Given a training set,colsel computes a com-pression plan based on the greedy differential algo-rithm.cszip.Given a compression plan computed by colsel, cszip compresses a test set with respect to the plan.It encodes enough of the plan into the output(which is included in the output size results reported)so that, given a compressedfile,cszip will uncompress it without needing the original plan.Pzip and cszip use the zlib library,version1.1.3,1 to compress the intervals and columns,respectively.System.All the training and experiments were run on one250MHz MIPS R10000processor on a16-processor SGI Origin2000running IRIX6.5,with10GB of main memory.Each time reported is the median offive runs, summing user and system time for each run.4.2Experiments and Results.Optimum Partitioning.To test assumption(1)and as-sess how optimum partitioning affects compression perfor-mance,we used pin to compute optimum partitions on pieces of the training data of increasing size.We ran pzip with each resulting compression plan on the test set and com-pared the compression time and resulting size to those of the benchmarks;we also compared uncompression times.We performed this experiment using both the ordered and ran-dom training sets.Figures1and2displays the results,as a function of the amount of training material used.Training on the2%-size(w.r.t.the test set size)data set,at which we see the results stabilize,took about2.27CPU minutes.Because we anticipate using the same compression plan with multi-ple tables from afixed source,though,training should beTraining set size / test set size (%)0.00.51.01.5R e l a t i v e p e r f o r m a n c e(a)Training set size / test set size (%)0.00.51.01.5R e l a t i v e p e r f o r m a n c e(b)Figure 1:Results of optimum partition compression,as a function of amount of training material.Shown is the relative performance of optimum partitioning over benchmark R in terms of compressed size,compression time,and uncompression time.(a)Ordered training set;(b)random training set.in training time.We offer a caveat:pzip has undergone significantly more code optimization than cszip ,which partially ex-plains the relative difference in running times.We believe that we can improve the running time of cszip by combin-ing the column differencing and compression of the derived columns into a single pass.4.3General Discussion.In all the experiments above,the difference between using ordered and random training sets was negligible,although the ordered sets did provide slightly better results,suggesting the need for future experiments to assess the effects of contiguity in the training data.We did observe stabilization of compression perfor-mance in all the experiments.Perhaps most remarkable is that this stabilization occurred at training set sizes of 1–2%of the test set size.Again,further experiments with increas-ingly larger test sets and a fixed training set size are required before assumption (4)can be assumed with confidence.4.4Comparison to Other Methods.We compared the result of optimum partitioning (using the 2%training size)to Burrows-Wheeler [1]and PPM [6,19]compression in isolation.For Burrows-Wheeler,we used Seward’s bzip2,version 0.9.5d.2For PPM,we used Bloom’s ppmz ,versionCompress.Uncompress.Method gzip 6.340e-1 1.202e-0 1.129e-0col-major 7.768e-14.344e-12.165e-1PPM3/bloom/src/ppmz.htmlTraining set size / test set size (%)0.00.51.01.5R e l a t i v e p e r f o r m a n c e(a)Training set size / test set size (%)0.00.51.01.5R e l a t i v e p e r f o r m a n c e(b)Figure 2:Results of optimum partition compression,as a function of amount of training material.Shown is the relative performance of optimum partitioning over benchmark C in terms of compressed size,compression time,and uncompression time.(a)Ordered training set;(b)random training set.PPM,the relative speed difference was orders of magnitude.The results suggest that,for our table application,opti-mum partitioning using gzip as the underlying compression method outperforms isolated usage of bzip and PPM ,which by themselves purport to outperform gzip .Since bzip did out-compress gzip with only a slight time penalty,it is worth future experimentation to assess the performance of optimum partitioning using bzip as the underlying com-pression method.5Partition Compression System and ApplicationsPin and pzip actually form our production compression system.Recall that the experiments in Section 4used only the 90highest frequency columns from the original data set.Prior to determining an optimum partition,pin calculatescolumn frequencies.It actually computes the optimum par-tition only on the projection of the high frequency columns.(How it determines low from high is a heuristic outside the scope of this paper.)Furthermore,before computing the par-tition,it employs another heuristic to reorder the high fre-quency columns to improve compression size further.Again,this heuristic is outside the scope of this paper and was turned off for the experiments in Section 4.Pzip then compresses the low frequency columns by differential encoding,additionally gzip ping the output of of that phase,and the high frequency columns with respect to the (reordered)partition.On the low frequency columns of the full network traffic data set,this method outperformed the gzip benchmark (R)by two orders of magnitude in com-pression size and almost an order of magnitude in compres-sion time.To measure how the full system works on the original data set,we repeated the structure of the optimum partitioning experiment,allowing pin to detect the low fre-quency columns and pzip to use differential encoding on them.The system setup was as in Section 4,and again we fixed gzip as the underlying compression method.We com-pared to benchmark R,corresponding to off-the-shelf use of gzip ,which was the method of choice in the AT&T net-work traffic data warehouse prior to our work.The results are shown in Figure 5.For this experiment,we used only the ordered training set,as the random training set destroys the frequency information.The results indicate overall improvements relative to straight gzip of 55%in compression size and 40–50%in both compression and uncompression time,supporting the argument that the space improvement by optimum partition-ing is worth the extra time.The time savings for the low frequency columns more than paid for the extra time needed to compress the high frequency columns via optimum parti-tioning.Again,training sets of 1–2%of the test set size suf-ficed to achieve these results.Applied to the network traf-fic warehouse,pin/pzip compresses the raw data for an entire month from the original 1TB to about 28GB,small enough to fit on a large PC disk.By comparison,gzip com-pressed the data only to about 65GB.Chunk size 012R e l a t i v e p e r f o r m a n c e(a)Chunk size12R e l a t i v e p e r f o r m a n c e(b)Figure 3:Results of chunk partition compression,as a function of chunk size.Shown is the relative performance of chunk partitioning over optimum partitioning in terms of compressed size,compression time,uncompression time,and training time.(a)Ordered training set;(b)random training set.5.1Additional Test Sets.The AT&T network switch statistics project.In this project,statistics are collected at 15-minute intervals from ATM switches in a network.A record corresponds to a re-placeable circuit component in one of the switches and con-sists of a 16-byte hardware identifier,a 4-byte statistic iden-tifier,a 4-byte time stamp,and a 4-byte count value.Each 15-minute interval produces a file with about 80,000records for a 6-switch network.The items are sorted by the 16-byte hardware identifier;sorted identifiers usually differ by one byte from one item to the next.The file format is determined by the switch manufacturer and has an irregular structure:variable length headers and interspersed sequencing records make the records variable length in general.The average file size is about 2.2MB and gzip s to about 192kB,making the daily space requirement 18.4MB.To use pzip ,each record was padded to a fixed 32bytes.This expanded the average file size to about 2.6MB.Training data produced 10high frequency columns,for which an optimum (reordered)partition was generated.The resulting average pzip ped file size was 10.3kB,for a daily space requirement of 1MB,a 95%improvement over straight gzip .Compression and uncompression time improvement was only about 20%.U.S.census data.We took a portion of the United States 1990Census of Population and Housing Summary Tape File 3A (a.k.a.STF3A )[2].The data format is fixed length ASCII records.We used field group 301,level 090,for all states.Table 2:Summary of results.Size and times are ratios of pzip values to the corresponding gzip values.Size time time .45.62.54Ntwk.switch.56.50.33This generated a 342MB file with 932-byte records.Gzip compressed the file to 31.5MB.Pin determined that 186columns were high frequency.In the optimum partition generated,the largest class was 56bytes wide,indicating high combinational dependence.Pzip compressed the file to 17.5MB,a 44.4%improvement over gzip .The compression time improvement was 50%,and uncompression time improvement was 67%.5.2Discussion.Table 2summarizes the results in this section.Based on its performance on the network traffic data,pzip has been put into production use in the AT&T network traffic warehouse,using a compression plan generated by pin on about 100,000records.Although not in a controlled setting,this will provide an “in-production”experiment thatTraining set size / test set size (%)0.00.51.01.5R e l a t i v e p e r f o r m a n c e(a)Training set size / test set size (%)0.00.51.01.5R e l a t i v e p e r f o r m a n c e(b)Figure 4:Results of greedy differential compression,as a function of amount of ordered training material.(a)Shows the relative performance of greedy differential compression over optimum partitioning in terms of compressed size,compression time,uncompression time,and training time.(b)Shows the relative performance of greedy differential compression over benchmark C in terms of compressed size,compression time,and uncompression time.can help assess assumption (4),because going forward,pzip will be compressing an arbitrary amount of data based on the fixed-size training set.6Concluding RemarksWe have presented massive tables as a new focus in data compression research.We have given a systematic approach for solving the problem,based on the experimental vali-dation of data dependency assumptions.The result is a new compression paradigm:training for lossless compres-sion.By exploiting data dependencies,our scheme outper-forms standard methods based on information theoretic re-sults,e.g.,Lempel-Ziv [20,21].We tested two such depen-dencies.For our application,optimum partitioning is better,and it is in production use within AT&T,in the pzip sys-tem.We anticipate instances for which the differential ap-proach will outperform the combinational approach and also instances that favor a hybrid approach.We leave as an open problem to find other data dependencies.Our results demonstrate the utility of training for loss-less compression.Given multiple tables from a commonsource,training becomes an off-line operation,suitable for computationally expensive optimizations.The bottleneck in our dynamic programming algorithm for optimum partition-ing is the computation of for all ,which requires running the base compressor (gzip ,in our case)on intervals of columns.A quick way to esti-mate the compressed size of an interval of columns,such as providing a suitable lower bound on their joint entropy—a fundamental problem of independent interest—would there-fore be valuable in speeding the overall algorithm.Two aspects of our work that are now only heuristic are as follows.Permuting the columns before partitioning them effects greater compression.The problem of optimally per-muting the columns can be abstracted in combinatorial op-timization terms as versions of the Hamiltonian path prob-lem or clustering.We suspect that these formulations will prove to be hard,but proving their hardness is non-trivial.Any reduction must capture required costs by constructing columns whose compressed size using a particular program (such as gzip )will match required costs in the reduction.From a practical point of view,an efficient heuristic with good performance is desirable.Our second heuristic involves the choice of low frequency columns that are removed prior to training.In our data sets,simple rules of thumb sufficedto identify such columns,but a formal approach would be desirable.In the differential approach,we focused on the case in which derived columns can be assigned only to source columns.In general,however,we can build a tree of derivations,which implies an interesting variation of the facility location problem that can be solved exactly by a reduction to minimum spanning trees.We also leave as open problems to explore the effect on compression plans。