Ergonomic evaluation of interior design of Shoka v人机工程外文文献
- 格式:doc
- 大小:62.00 KB
- 文档页数:8
615 CAD System for Human-Centered DesignKunwoo Lee11Seoul National University, kunwoo@snu.ac.krABSTRACTSevere competition in today's marketplace has imposed additional demands on the products beingsold. In addition to basic requirements, such as function, quality, price, and appearance, modernproducts must be increasingly comfortable and easy to use. The current discussion will introducetwo fundamental approaches intended to improve the comfort and user-friendliness of newproducts. The first approach involves designing a product to reflect the geometry of the humanbody. This approach can be exploited when designing custom-tailored products, such as shoes,clothes, and helmets. Specialized CAD systems for custom tailoring these products will beintroduced together with human body scanners and human body modelers. The other approach isto design a product based on the simulation of both product behavior and human motion. Thisapproach can be used to design products that are used in close association with the human bodyto maximize its comfort and ease of use. To realize the simulation environment, a CAD system thatallows modeling the human and the product in the same environment and provides the simulationresults is necessary. In this paper, a prototype CAD system for this purpose is also introduced.Keywords: Human-Centered Design, CAD, custom-tailoring, human motion simulation1. INTRODUCTIONDue to globalization and increasingly competitive market environments, modern products must meet more requirements than ever before. They must not only be functional and aesthetically pleasing, but comfortable and ergonomic, as well. Consumers increasingly demand products that conform to the human body in a customized way in order to maximize comfort and mitigate user fatigue. Similar design elements are also becoming apparent in the design and layout of factory spaces. It is now common for factories to adopt new ergonomic designs in order to reduce the possibility of dangerous situation and to increase the efficiency of interactions between workers and machines. These types of design that consider the static geometry or dynamic responses of the human body are referred to as ‘human-centered design’.Current CAD systems are unable to fully support a human-centered product design. Most of the research regarding ergonomic design has primarily focused on simple geometric analysis and simulation and has considered only the human model rather than the interaction between the product and the human model. As a result, it is difficult to apply the method to the traditional design process. Consequently, such designs must be completed manually using trial-and-error iterative experiments on physical prototypes.Some integrated CAD systems, such as CATIA from Dassault Systems, provide simple simulation and analysis tools for the human body, such as HBR (Human Builder), HME (Human Measurement Editor), HPA (Human Posture Analysis), and HAA (Human Activity Analysis). However, the functions and capabilities of these internal modules or add-ons are quite limited when compared to that of dedicated human simulation tools [3]. Particularly, the human analysis process in CATIA can only be used for static analysis, and not for dynamic analysis. Recently, many applications to simulate the human motion and the corresponding muscle reactions have been developed. These tools are very useful for simulating the man-machine behavior simultaneously. Furthermore, some commercial software, such as SIMM from Motion Analysis Company and LifeMOD from Biomechanics Research Group, have been designed for this purpose [12][20]. However, these software packages were developed considering only the human motion simulation and did not provide the design environment for the products which interact with the human motion. They also do not provide an easy way to define the interaction between the human and the product and to run the simulation for both the human and the product.Computer-Aided Design & Applications, Vol. 3, No. 5, 2006, pp 615-628616Due to the aforementioned limitations, the design of human-friendly products primarily involves trial-and-error. This trial-and-error process is undesirable because it inherently requires human operators and their subjective evaluations. Accordingly, an integrated approach combining the CAD/CAE system and human simulation tools is strongly recommended. Also, an in-depth study of the interaction conditions between the human body and product model is required. In order to achieve the goal of human-centered design, the following research activities must be conducted: •Human body modeling.•Human motion and muscle force simulation.•Design environment for modeling both human body and product.The related works on the first two topics in the above list will be described in the following chapters, and then the third topic will be demonstrated with several case studies performed at the Seoul National University.2. HUMAN BODY MODELINGThe human models used for human-centered design can be classified into four categories: simple geometric models, multi-body models, finite element models, and a hybrid models that combine multi-body and finite element models. 2.1 Simple Geometric ModelSimple geometric models of the human body are often used to design custom-tailored products, to perform simple ergonomic analysis for applications like reach evaluation, and to create animations. Geometric models of the whole human body or the portions thereof are obtained by scanners or section data from computer tomography or laser bean scanners. The scanned data are usually triangulated and the anatomical features are represented as a facet model.2.2 Multi-body ModelsThe musculoskeletal systems of a human body are usually modeled as a set of rigid bones connected by joints. These joints include muscle-tendon actuators that span the joints and develop force, thus generating moments about the joints. These models allow passive simulation as well as active simulation (inverse and forward dynamics). Multi-body models can be used in impact simulation and analysis, ergonomics, comfort study, biomechanical analysis, movement simulation and analysis, and surgical planning.MADYMO [14] provides a series of multi-body models whose geometries are represented by ellipsoids or facet surfaces. LifeModeler [12], developed by the Biomechanics Research Group, is another commercially available human motion simulation program that uses ADAMS (Automatic Dynamic Analysis of Mechanical Systems) as a dynamics solver. It provides a default multi-body model of the human body that can be modified by changing anthropometric sizes. Simulation examples using LifeModeler will be illustrated in Section 3. Another commercially available software known as AnyBody [1] is similar to LifeModeler, in that it provides the simulation of the human body motion with the human body stored in the program.The Neuromuscular Biomechanics Lab (/group/nmbl/) at Stanford University created a software package called SIMM (Software for Interactive Musculoskeletal Modeling). This software enables users to develop, alter, and evaluate models of musculoskeletal structures. In SIMM, a musculoskeletal model consists of a set of body segments that are connected by joints. The joints are modeled using detailed, accurate kinematics functions, or simple pin or ball-and-socket joints. The model includes muscle-tendon actuators that span the joints and develop force, thus generating moments about the joints. This software can be used for biomechanical analysis, movement analysis, surgical planning, and ergonomics.2.3 Finite Element ModelsFinite element models of a whole human body, or portions thereof, can be used to evaluate the effects of a product on the human body during use.2.3.1 WSU modelThe Wayne State University (WSU) Bioengineering Center () is among the pioneers in human modeling. The WSU Human Injury Investigation Model was developed for use with simulation programs such as MADYMO, LS-DYNA [13] and PAM-CRASH [19].Computer-Aided Design & Applications, Vol. 3, No. 5, 2006, pp 615-628617Fig. 1. WSU model.The WSU Bioengineering Center also determined a unique data set to simulate head impact. The data set contained a geometric mesh which described all the components of the head, including the scalp, skull, inner brain membranes, and hemispheres of the brain, cerebellum, brain stem, ventricles, cerebrospinal fluid, and bridging veins. All of these components were made up of 38,000 elements and 29,000 nodes. They also developed finite element models of other portions of a human body, such as head and neck, thorax, abdomen, spine, pelvis and hip, and knee.2.3.2 HUMOS-1 & HUMOS-2HUMOS is full human body model that was funded by the EU and developed by a consortium of European car companies, research institutes, universities and software vendors.HUMOS-1 shown in Fig. 2 is a human model of a male in a driving position close to the 50th percentile [2].Fig. 2. HUMOS-1 model.HUMOS-2 models [24] are the successor of HUMOS-1 models. They also provide the positioning tool to put the human models in a proper pose for each specific application.2.3.3 THUMSTHUMS (Total Human Model for Safety) is a family of human models developed by Toyota Central R&D Labs (http://www.tytlabs.co.jp/eindex.html) and are illustrated in Fig. 3.Computer-Aided Design & Applications, Vol. 3, No. 5, 2006, pp 615-628Computer-Aided Design & Applications, Vol. 3, No. 5, 2006, pp 615-628618Fig. 3. THUMS Human Model Family: AM 50, AF50, AF05, 6 year old child, pedestrian and a girl.Each model contains more than 80,000 finite elements and 60,000 nodes [18]. THUMS also provides a brain model to simulate a brain injury.2.3.4 H-ModelsThe H-Model family was developed at the Hong-ik University in Korea with the IPSI/ESI partner companies for impact biomechanics, pedestrian impact and riding comfort.Fig. 4. Elements of the H-Model.2.4 Hybrid ModelSometimes, finite element models require too many calculations and the associated simulation may not serve as a good design tool. In this case, hybrid models that combine rigid bodies and finite elements are used.619Fig. 5. Hybrid models that are mixtures of rigid bodies and finite elements.Fig. 5. shows an example of a hybrid model where rigid bodies are used for most segments, but a deformable structure is used for the thorax.3. HUMAN MOTIOIN AND MUSCLE FORCE SIMULATIONAs mentioned in the previous section, there are many commercially available simulation software products, such as LifeModeler, MADYMO, AnyBody, and SIMM. These all basically provide generic models for the human body and allow the user to scale the human model, position it to the same position as that in operation, and solve the dynamic equation by rigid dynamics solver or finite element analysis code. The results of the simulation are the human motion, forces exerted by the muscles, and the stresses or strains at the desired location of the human body. From these results, the comfort or ease of operation can be estimated for a given task. Figures 6-9 illustrate the example output of some simulations.(a) (b) (c)(d)Fig. 6. Injury simulations: (a) Falling, (b) Bungee jumping, (c) Car crash and (d) Motorcycle crash.(a) (b) (c) (d) (e)Fig. 7. Performance: (a) Bicycle riding, (b) Golfing, (c) Exercise, (d) Fighting and (e) Dancing.Computer-Aided Design & Applications, Vol. 3, No. 5, 2006, pp 615-628Computer-Aided Design & Applications, Vol. 3, No. 5, 2006, pp 615-628620Fig. 8. Optimizer: Car driver muscle works.Fig. 9. Impact simulation with a belt system.4. CASE STUDY 1 - CUSTOM TAILORING PRODUCTS USING HUMAN MODELAs the standard of living continues to improve, consumers tend to purchase products that are custom tailored to fit their bodies perfectly. Many CAD systems for custom tailoring products are currently being developed to meet this growing need. In this presentation, two examples of custom production systems developed at Seoul National University are introduced: custom tailoring systems for shoes and for wigs.Computer-Aided Design & Applications, Vol. 3, No. 5, 2006, pp 615-6286214.1 Custom Tailoring ShoesWe implemented the shoelast design system as part of a system to create custom tailored shoes. The shoelast design system uses the scanned data from the foot scanner to design a new shoelast fitted to the scanned data.The proposed shoe last design system consists of three parts. In the first part, the pre-processor, the point cloud data from the scanner is converted to a data format that is easy to handle. An existing shoelast that is closest to the scanned foot is selected from the database and is modified to fit to the customer’s foot. At the same time, a customized insole is also designed so that its surface exactly fits the bottom of the foot. The post-processor then generates the tool path data for manufacturing the shoelast and the insole. Figures 10-16 illustrate the result of each step.Fig. 10. Foot scanner. Fig. 11. Scanned result.Fig. 12. The parameters of foot.Fig. 13. The parameters of standard shoelast.Computer-Aided Design & Applications, Vol. 3, No. 5, 2006, pp 615-628622Fig. 14. The modified shoelast.Fig. 15. The machined shoelast.Fig. 16. The machined insole.4.2 Custom Tailoring WigsA customized wig manufacturing system was based on the same concept as the customized shoe manufacturing system. The head is scanned in order to generate a net, on which hairs are implanted. Since the net has the same shape as the head while the shape of the shoelast differs from that of the foot, the development of customized wigs will be inherently easier than for shoes. Figures 17-18 show the general procedure.623Fig. 17. Procedure of custom tailoring wigs.Fig. 18. CAM system for customized wig manufacturing.5. CASE STUDY 2 - PRODUCT DESIGN CONSIDERING EFFECT ON HUMAN OPERATORAs stated in the previous sections, many commercial programs already exist to simulate the human body and muscle forces for a given motion. However, traditional biomechanical research has largely focused on human modeling and analysis, has largely ignored the design of the product. For example, many attempts have been made to estimate the optimal pedaling rate or direction while cycling [4],[15-17][22]; however, relatively little effort has been expended in order to determine the optimal position of the crank with respect to the saddle when the size and posture of the human model are changed [11]. This is similar to past situations in which CAD and FEA systems were run independently on different platforms. These days, we run FEA inside CAD systems, and the design alteration is performed based on the FEA results right in the design stage.Jung and Kee suggested an integrated design framework for ergonomic product design [6]. A new man-machine interface model is proposed to integrate the ergonomic evaluation tools into the CAD system; the visibility and reach functions are used for the ergonomic evaluation function. The systematic scenario is quite similar to this study. They modeled the human body as a multi-link system and used the Denavit—Hartenberg notation of robotics for the kinematic representation. However, their system can handle only the evaluation of visibility and reach functions. Also, the geometry of the human model is represented by the boundary-representation (B-rep), and its main function is only to graphically visualize the human body. Therefore, it cannot be directly used as design references. Finally, there is nospecific interface relation or constraint between the human body and solid objects.Computer-Aided Design & Applications, Vol. 3, No. 5, 2006, pp 615-628Computer-Aided Design & Applications, Vol. 3, No. 5, 2006, pp 615-628624 Sooyeup Lee tried to determine the objective function value of the motion of various bicycle models operated by a musculo-skeleton human model in a computer’s virtual space [11]. By minimizing the value of this objective function the muscle can exercise with minimal force, leading to minimum muscle exhaustion. His research represents an ideal approach to determine the important design parameters in the product model. However, the approach is limited to the pedaling motion of bicycles and the total computing time is very long. Consequently, a more general approach to enable human-centered design in the conventional design process is needed.In this case study, we propose a new design platform that is constructed upon commercial CAD systems. The human body and product are closely interrelated by certain combinations of interactions defined in the platform. A product design can be evaluated for many representative human bodies without re-defining the relationships between them. The integrated platform is developed to have the following features.General Feature-based CAD capabilities.Input of human model into the platform. Interactive specification of human-product interaction [21]. Automatic inverse kinematics on the human body model for the initial posture determination from thehuman-product interaction [5][7-10][23],[25].Import motion data from motion capture camera. Generation of input file for human motion simulation program.Graphic postprocessor to display the results of human motion simulation.In this case study, the integrated platform was constructed on a commercially available CAD system and the featuresHuman Model Agent Human ModelManangerModeling Engine ReferenceEntities Human AnalysisApplicationInteraction MgrRefEnt MgrPostureBuilder Exporting Simul Mgr Config Human Model Collision Mgrsimple different human models.For the detailed dynamics simulation and analysis, “Human Model Manager” modules can create the batch command file for the external simulation software. All the time-consuming and iterative steps can be minimized.625The HMA (Human Model Agent) and HMM (Human Model Manager) were implemented as add-on programs for SolidWorks 2005. We also used the BRG.LifeMOD Biomechanics Modeler from the Biomechanics Research Group, as the human motion analysis software. This software is based on the popular ADAMS software product. Interaction Type Attachment(with A and B) Possibly Contact(with A and B) Fix(with A and B) Hand-holding Stepping-on Sitting-on Lean-against Lift an object in armsTab. 1. Classification of interactions between human model and product model.‘bushing’ force conditions ‘contact’ force condition ‘fix’ conditionsIn this case study, we classified the interactions between the human and product models into 8 categories (shown in Tab. 1), depending on their postural meaning and interaction characteristics. With these interaction types, the posture of a human body is more easily specified compared to when we use LifeModeler or other simulation programs.(a)(b)(c)(d)(e)Fig. 20. (a) Hand-holding (pronation, supination), (b) Stepping-on (with right foot), (c) Sitting-on, (d) Leaning-against and (e) Lift an object in arms.Each interaction type can be modeled differently upon user input. For example, the ‘stepping-on Object A with left foot’ interaction can be modeled as the ‘bushing’ model or ‘contact’ model with different parameters, Fig. 20. ‘Attachment’, ‘Possibly Contact’, and ‘Fix’ are the same as the ‘bushing’ force, ‘contact’ force, and ‘fix’ condition, respectively. The other 5 interactions have more postural meaning and interaction characteristics, but they are internally expressed as one of the ‘Attachment’, ‘Possibly Contact’, or ‘Fix’.Computer-Aided Design & Applications, Vol. 3, No. 5, 2006, pp 615-628626When these interactions are assigned, the posture of the human model is also adjusted for the postural meaning of the interaction. It is very easy to model the interaction between the human model and product model because assigning the interaction also changes the human posture automatically. 5.1 Home Fitness Cycling Machine In this example, we will examine the effect of the height of the saddle on the motion of the human body. For the two different positions of the saddle, the bio-mechanic analysis is performed, and the results are compared to each other. We first use the proposed system to model the human and bicycle in the normal saddle position (Fig. 21). We then run the simulation in LifeMOD with the exported files from the system. The simulation result is shown in Fig. 22. Subsequently, we alter the position of the saddle in the cycle machine. If we move the saddle part downward, the human model is automatically repositioned. Finally, we get the modified exported files.Fig. 21. Modeling example using proposed system.Fig. 22. Result of simulation (10 selected frames).(a)(b) Fig. 23. Comparison between 2-different saddle positions.The result can be quantitatively compared by selecting a few representative muscle forces during the simulation (5.0 sec, 100 time-steps) and plotting them in Fig. 23. For the normal saddle position, the maximum force for all muscles is smaller than 600N (Fig. 23(a).). However, for the low saddle position, the maximum force is greater than 1400N (Fig. 23(b).). The result of the simulation is consistent with our own experience. When we ride on the small bicycle (such as a child’s bicycle), pushing the pedals is difficult because there is a very small free region of movement.Computer-Aided Design & Applications, Vol. 3, No. 5, 2006, pp 615-6286275.2 Elliptical Trainer The second example involves another fitness machine frequently referred to as an Elliptical Trainer. The overall combined motion of this product and the human model is quite complex. As a result, it is difficult to estimate the human motion in this example, even for an average person. The interactions and stabilizing constraints are assigned as shown in Fig. 24.Fig. 24. Assigning of interactions and stabilizing constraints.Fig. 25. Simulation result of elliptical trainer example.The basic procedure for modeling the human and the elliptical trainer together is almost the same as the first example. However, as mentioned previously, the stabilizing constraints are needed in this example. The stabilizing constraints are used to generate the initial motion of the human body during the inverse dynamics simulation. Using the implemented system, the model can be simulated successfully (3 sec, 150 time-steps). The result of the simulation is shown in Fig. 25. We conclude that the proposed system can efficiently handle this complex example. 6. DISCUSSION The competitive nature of the global economy is placing strong pressure for companies to design products that are increasingly easy for humans to use. Thus, current CAD systems should be modified to provide the function by which this ‘user-friendliness’ can be evaluated at any time in the design process and the evaluation result can be implemented directly into the design process. Although Finite Element Analysis tools were introduced a long time ago, these tools have only recently been integrated into CAD systems, enabling analysis and design to be carried out simultaneously. We are experiencing a similar phenomenon in the case of human motion and muscle force simulation tools. These tools must eventually be integrated into the CAD system and the product design should be performed with consideration to the effect of the product on the human user. There are several problems to be solved to realize this design paradigm. First, an integrated framework to accommodate product CAD models and digital human models is required. This system needs to be devised in such a way to prevent losing all the features of CAD systems and simulation programs for human motion and muscle forces. Second, digital human models of various resolutions have to be developed. It will be necessary to use the right resolution model depending upon the application. Third, the simulation programs for human motion and muscle forces should become fast enough for the interactive design task. FEA tools were integrated into VAD systems after very fast FEA tools became available.Computer-Aided Design & Applications, Vol. 3, No. 5, 2006, pp 615-628628Fourth, good algorithms should be developed so that the human model can be positioned properly to maximize comfort and ease of use for a given product. We hope the proposed approach in this paper will make a contribution toward this goal. 7. ACKNOWLEDGEMENTS This research is supported by the Laboratory of Excellency Research Laboratory Program of the Ministry of Commerce, Industry and Energy in Korea. This research was also supported in part by the Institute of Advanced Machinery and Design at Seoul National University. 8. REFERENCE [1] AnyBody, , Anybody Technology. [2] Behr, M., Arnoux, P. J. and Serre, T., A Human Model for Road Safety: From Geometrical Acquisition to Model Validation with Radioss, Computer Methods in Biomechanics and Biomedical Engineering, Vol. 6, No. 4, 2003, pp 263-273. [3] CATIA v5, , Dassult Systems. [4] Chen, G., Kautz, S. A. and Zajaca, F.E., Simulation analysis of muscle activity changes with altered body orientations during pedaling, Journal of Biomechanics, Vol. 34, 2001, pp 749-756. [5] Jung, E. S. and Choe, J., Human reach posture prediction based on psychophysical discomfort, International Journal of Industrial Ergonomics, Vol. 18, 1996, pp 173-179. [6] Jung, E. S. and Kee, D., Man–machine interface model with improved visibility and reach functions, Computers and Industrial Engineering, Vol. 30, No. 3, 1996, pp 475-486. [7] Jung, E. S., Kee, D. and Chung, M.K., Reach posture prediction of upper limb for ergonomic workspace evaluation, Proceedings of the 36th Annual Meeting of the Human Factors Society, Atlanta, GA, Part 1, Vol. 1, 1992, pp 702-706. [8] Jung, E. S., Kee, D. and Chung, M.K., Upper body reach posture prediction for ergonomic evaluation models, International Journal of Industrial Ergonomics, Vol.16, 1995, pp 95-107. [9] Kee, D. and Kim, S.H., Analytic generation of workspace using the robot kinematics, Computers & Industrial Engineering, Vol. 33, 1997, pp 525-528. [10] Kee, D., Jung, E.S., and Chang, S., A man-machine interface model for ergonomic design, Computers & Industrial Engineering, Vol. 27, 1994, pp 365-368. [11] Lee, S., Motion of musculoskeletal human model minimizing the metabolic energy, MS Thesis, Seoul National Univ., 2003. [12] LifeMOD, , Biomechanics Research Group Inc. [13] LS-DYNA, , Livermore Software Technology Corp. [14] MADYMO, http://www.automotive.tno.nl, TNO Automotive. [15] Neptune, R. R. and Kautz, S. A., Knee Joint Loading in Forward versus Backward Pedaling: Implications for Rehabilitation Strategies, Clinical Biomechanics, Vol. 15, 2000, pp 528-535. [16] Neptune, R. R. and Hull, M. L., A Theoretical Analysis of Preferred Pedaling Rate Selection in Endurance Cycling, Journal of Biomechanics, Vol. 32, No. 4, April 1999, pp 409-415. [17] Neptune, R. R. and Hull, M. L., Evaluation of performance criteria for simulation of submaximal steady-state cycling using a forward dynamic model, Journal of Biomechanics Engineering, Vol. 120, 1998, pp 334-341, [18] Oshita, F., Omori, K., Nakahira, Y. and Miki, K., Development of a finite element model of the human body, 7th International LS-DYNA Users Conference, 2002. [19] PAM-CRASH, , ESI Group. [20] SIMM, , Musculographics Inc. (Division of Motion Analysis Cooperation) [21] Song, Y. J., Integrated Product Design Environment to Support the Human-Centered Design, IJCC (International Journal of CAD/CAM) Workshop on Digital Engineering, 2006. [22] Thelen, D. G., Anderson, F. C. and Delp, S. L., Generating dynamics simulations of movemnet using computed muscle control, Journal of Biomechanics, Vol. 36, 2003, pp 321-328. [23] Tolani, D., Goswami, A. and Badler, N. I., Real-time inverse kinematics techniques for anthoropomorphic limbs, Graphical Models, Vol. 62, No. 5, Sept. 2000, pp 353-388. [24] Vezin, P., Verriest, J. P., Development of a set of numerical human models for safety, /pdf/nrd-01/esv/esv19/05-0163-O.pdf [25] Wang, X. G., A behavior-based inverse kinematics algorithm to predict arm prehension postures for computeraided ergonomic evaluation, Journal of Biomechanics, Vol. 32, 1999, pp 453-460. Computer-Aided Design & Applications, Vol. 3, No. 5, 2006, pp 615-628。
Interior design stylesMary v. knackstedt shanghai2004Abstract:Architectural style derived from a variety of interior design style, according to designers and owners of the different aesthetic and loving. and turned into a variety of body.Although the hybrid design style eclectic, using a variety of style, but the design is still Originality, in-depth scrutiny shape, color, material so the overall composition and visual effects.Key words:Style Classical Simple ExquisiteArticle:HOW DO YOU CHOOSE AN interior design style that suits you and your home? you can draw inspiration from numerous styles but try to keep to a similar theme so your interior is coordinated.Modern retro interior design incoprates bright colours and old patternsThe best place to start is by looking at styles that appeal to you, which suit your budget and lifestyle. You will soon work out what you like and what you don `t. After looking through a few of these you will begin to understand the design process, along with the variety of materials and styles available.Interior design style and genre, an indoor environment in the spirit of the art modeling and functional areas. Interior design style and genre and construction is often the furniture as well as in close connection with the style and genre; sometimes to the corresponding period of the painting, plastic arts, and even literature, music, etc, in close connection with the style and genre; sometimes to the corresponding period of the painting, plastic arts, and even literary, musical style and genre, such as their origin and mutual influence. For example in architecture and interior design of the “post-modernism ”and the meaning of the word, and ”camp style “is the distinctive characteristics of a plastic arts schools. Can be seen that the construction addition to interior design and material art materials, engineering and technology beyond the characteristics of close contact, and also literature, music and painting, sculpture and other categories of communication between the arts.The formation of interior design styles, different ideas and characteristics of the times, by creative ideas and performance, and gradually developed into a representative form of interior design. The form of a typical style, usually in the humanities and local factors and is closely related to natural conditions and the need for creativity in design and modeling characteristics. The formation of external and internal style factors.Although the performance in the form of style, but the style of art, culture, social development, such as profound connotation; from this deeper meaning; the style does not stop or equivalent form. A style or genre, once formed, it can turn a positive or a negative impact on culture, arts and many social factors can not be limited to as a form of performance and visual experience.The style of interior design features in the embodiment of art and creative personality at the same time, relatively speaking, to that style across a longer time, will include wide geographical. Interior design style can be divided into; Traditional style, modern style, post-modern style, natural style as well as hybrid style.Architectural style derived from a variety of interior design style, according to designers and owners of the different aesthetic and loving, and turned into a variety of body. Here, to brief Members on the present 10 kind of the more common style of interior design. Classical Style (luxurious rich)The rise in the age of just decoration, decoration of most of the pursuit of prosperity is more luxurious style. Especially in the early 20th century, interior decoration to show off their status in often a special form. Owners will be asked to design a variety of embedded symbolic luxury decoration, such as painting the glass ceiling, fireplaces, decorative panels, decorative wood and soon, but basically similar to the Baroque style combined with the material there The main decorative way.Simple style (arbitrary)The 20th century, in some areas a Home heat. Due to technical constraints, the time has not yet the true sense of the Home Designer for guidance, so casual in the best portrayal of that time. Owners have begun the pursuit of a clean and bright indoor effect. Today, this style is still the most firs-time home buyers are the first choice for the decoration.Exquisite style (noble solemn)After nearly 10 years of exploration, with the improvement of living standards of residents, opening to the outside world increased, people began to yearn for and the pursuit of high-quality life. From about the mid-20th century, there is the decoration of the decorative use of sophisticated materials and furniture, especially at this time, the designer of the domestic into the ranks of Home Design, wish ha brought a new concept of decoration. Natural style (art)The beginning of the 20th century decorative boom has brought many people to the concept of decoration. large number of emerging market. Taiwan Hong Kong journal of decorative eye-opener for people, before we stone such as tactics have appeared in the reality of the design. In particular, e used to see red elm caused y large-scale use of “national renovation a yellow after the phenomenon of the decoration, hose to nature has become a people one of the objectives pursued. Natural style of advocacy,”return to nature,”highly aesthetic nature, combined with the natural order in today` s high-tech, high- social life, so that people can achieve the balance of physical and psychological, so indoor use of wood, fabric, stone and other natural materials shows that the material texture, fresh and elegant. In addition, because of their similar aims and practices, which may be classified as natural style of garden-style category .ptoral style sought in indoor use of wood, fabric stone and other natural materials shows that the material texture, fresh and elegant. In addition, because of their similar aims and practices, which maybe classified as a natural style of garden-style category. Pastoral style sought in indoor environment performance of leisurely , leisurely, comfortable, natural life of the countryside, and is often the use of natural wood, stone, rattan, bamboo and other materals simple texture. Coincidentally in the green room settings, creating a natural, simple, elegant atmosphere.Light style (hold Dafang)The mid-20th century, home design ideas have been a lot of liberation, people began to pursue a wide range of design, including modernism, postmodernism, such as the design of a series of more comprehensive system formed in the interior design. Turning to the decoration of people, these” isms” frequently appear in the mouth. This style basically cherry wood finishes as a major.Soft style(smooth independent)In the last century, the beginning of this century, a kind of smooth pursuit in the club like a little luxury design all kinds of real estate properties in the model room and the office appeared. and then appear in a large number of home decoration in general. This style of art is a relatively simple compared to emphasize the content and yet decorative forms. and gradually formed a black walnut wood decorative panels for the main style. Among them, the simplicity and minimalism the beginning of the surface.Style(quiet gentle)This is not in the last century. The beginning of this century as a design style, it is basically a wallpaper based on the surface as the main decoration material, combined with the woodworking practices Mixing this style emphasized the harmony of proportion and color. People will be the last part of a wall and the ceiling the same color, and the wall with the use of a light texture of the wallpaper. The whole of it is elegant and quiet, with no trace of the impetuous.City style(independent personality)The 21st century, the housing reform, and many young first-time home buyers who appear to have this style impetus. Young people have just bought a house, many of them cos I’m always skint, but this time the real estate and are essentially rough Housing(anon-basic style of decoration), and these young people are forced to carry out the renovation of the revolution. Limited by its financial resources, people are starting to come through a variety of forms has been stressed that “decoration”of perception , in which large-scale use of bright colors is a typical example. It will be at home in the large-scale use of a variety of colors, sometimes even in the same space, the use of three or more than three colors.Fresh style(freehand pastel)This is a simple and derived under the influence with a “petty bourgeoisie” flavor of the interior design style. Particularly as many of the emergence of single people, the petty bourgeoisie in the style of a great deal of variety in the decoration of the apartment.. As many times, they do not consider the decoration of many of the functions of the problem. They are often stressed and dull of an arbitrary. White 纱窗buoyant with a soft fabric of the sofa, and then stacked with a variety of colors of the pillow pile to form a full atmosphere of the interior space lazily. Chinese style(retro)With the emergence of large number of modernist doctrine, the rise of the domestic and a retro style, it is the revival of Chinese-style decoration. Chinese painting, calligraphy and the Ming and Qing furniture constitute the most important elements of Chinese design. However, these expensive retro furniture has become a major obstacle to the lovers.Hybrid style: In recent years, the architectural design and interior design in general has diversified, inclusive of the situation. Indoor layout also tends to have a modern and practical, but also learn the characteristics of traditional, in the decoration and furnishings in the ancient and modernChina and the West melting into one, such as traditional screens, and table setting, with modern style decoration of the walls, windows and doors. A new type of sofa; European classical glass lamps and wall decoration. with the traditional oriental furniture and furnishings in Egypt, and so short. Although the hybrid design style eclectic, using a variety of style, but the design is still Origin, in-depth scrutiny shape, color, material and so the overall composition and visual effects.。
Title:Unleashing Creativity:Exploring the World of Interior DesignIntroduction:In this essay,I will delve into the captivating world of interior design and share my experience in creating functional,aesthetically pleasing,and personalized spaces.From conceptualization to execution,interior design allows me to unleash my creativity and transform ordinary spaces into extraordinary havens.Body:Understanding Client Needs:Interior design begins with understanding the needs and preferences of the client.I communicate closely with clients to grasp their vision, lifestyle,and functional requirements.This understanding forms the foundation for creating spaces that cater to their unique needs and reflect their personality and style.Space Planning and Layout:Effective space planning is key to optimizing the functionality and flow of a space.I analyze the layout,dimensions,and purpose of each room, considering factors such as traffic flow,natural light,and the desired ambiance.By strategically arranging furniture,fixtures,and architectural elements,I create harmonious and well-utilized spaces.Color and Material Selection:Color and material selection play a crucial role in setting the mood and ambiance of a space.I carefully choose colors that complement eachother and align with the desired style and atmosphere.Additionally,I select materials,such as flooring,wall finishes,and textiles,that enhance the overall design aesthetic and provide durability and visual interest.Furniture and Accessories:Selecting the right furniture and accessories is essential in creating a cohesive and inviting interior.I consider the scale,style,and functionality of each piece,ensuring they align with the overall design concept.Accessories,such as artwork,lighting fixtures,rugs,and decorative accents,add personality and create focal points within the space.Lighting Design:Lighting design is a critical aspect of interior design.I carefully plan the placement of natural and artificial lighting sources to create a balance of illumination and ambiance.Proper lighting enhances the functionality of the space,highlights architectural features,and sets the desired mood for each room.Attention to Detail:Attention to detail is paramount in interior design.I pay close attention to the finishing touches,such as trim work,hardware selection,and styling,to create a polished and cohesive look.Every element,from window treatments to the arrangement of accessories,is thoughtfully considered to ensure a harmonious and visually appealing space.Conclusion:Interior design is a dynamic and creative process that allows me to transform ordinary spaces into extraordinary havens.From understanding client needs to space planning,color and material selection to furniture and accessory choices,interior design involves careful consideration of each element to create functional,aesthetically pleasing,and personalized spaces.Through my designs,I aim to create environments that inspire,nurture,and reflect the unique personality and style of the inhabitants.The world of interior design is a realm of endless possibilities,and I am excited to continue unleashing my creativity to create spaces that leave a lasting impression.。
企业家精神测度的文献综述近年来,企业家精神成为了学者和实践者关注的焦点之一。
对于企业家精神的定义也越来越广泛,其中包括创新、冒险精神、自信、信念和坚韧不拔等特质。
本文将从测度企业家精神的角度,综述相关的文献。
首先,测量企业家精神的方法主要有两种:问卷调查和情境测试。
问卷调查是较为普遍的方法,主要包括以下两类:独立发明者量表(IEM),企业家特质量表(ENTR)。
独立发明者量表(IEM)首次由荷兰心理学家埃德斯特姆(Edustrom)于1977年开发,它是一种针对企业家的测量工具,以调查个体的“独立发明者”能力和行为。
该工具包括知识领域、创造力、问题解决能力、挑战和创新的态度等方面的评估,评价对象为独立发明者的运用它以在自己的领域内发展出新的商业活动。
它适用于评估创新性和创造性的商业能力,评估总成绩越高,商业创新能力越强的结论。
企业家特质量表(ENTR)是创业者个性测量工具,以评估候选人与成功企业家之间的差异为目的,由Shaver和Scott(1991)发展而来。
ENTR是一个多维量表,包括个性特质、态度和动机、行为和行动方面,它由5个维度和22个项目组成,其中包括自信、自控、冒险和创新,坚定和坚韧不拔等。
ENTR是一个有效的根据个性和行为特质对企业家进行筛选和评估的测量工具。
其次,情境测试是在模拟的真实环境下对候选人进行测试,其目的是测量候选人在不同情境下的商业决策和行动。
情境测试包括以下几种:商业情境展示、团队合作等。
商业情境展示是在一个真实的商业环境中对被试进行考察,以评估其商业决策和行为能力。
此方法可以通过评估被试的决策速度、正确性和情境响应性来评估其天生的商业能力。
团队合作测试是评估企业家精神中领导和合作的特质。
在这种测试中,候选人参加一系列的团队活动,这些活动旨在展示候选人的领导力和合作能力,以及他们解决问题和采取行动的能力。
尽管这些测量方法有助于评估企业家精神,但它们都存在一定局限性。
Designing Efficient Cooperative Caching Schemes for Multi-Tier Data-Centersover RDMA-enabled NetworksS UNDEEP N ARRAVULA,H YUN-W OOK J IN,D HABALESWAR K.P ANDATechnical ReportOSU-CISRC-6/05-TR39Designing Efficient Cooperative Caching Schemes for Multi-Tier Data-Centersover RDMA-enabled NetworksSundeep Narravula Hyun-Wook Jin Dhabaleswar K.PandaComputer Science and EngineeringThe Ohio State Universitynarravul,jinhy,panda@AbstractCaching has been a very important technique in improv-ing the performance and scalability of web-serving data-centers.Research community has proposed cooperationof caching servers to achieve higher performance benefits.These existing cooperative cache designs often partially du-plicate cached data redundantly on multiple servers forhigher performance while optimizing the data-fetch costsfor multiple similar requests.With the advent of RDMA en-abled interconnects these cost estimates have changed thebasic factors involved.Further,utilization of large scaleof resources available across the tiers in todays multi-tierdata-centers is of obvious importance.Hence,a systematicstudy of these various trade-offs involved is of paramountimportance.In this paper,we present cooperative cacheschemes that are designed to benefit in the light of the abovementioned trends.In particular,we design schemes tak-ing advantage of RDMA capabilities of networks and multi-ple tier resources of modern multi-tier data-centers.Ourdesigns are implemented on InfiniBand based clusters towork in conjunction with Apache based servers.Our ex-perimental results show that our schemes show throughputimprovement of up to35%better than the basic cooperativecaching schemes and180%better than the simple singlenode caching schemes.1IntroductionBanking on their high performance-to-cost ratios,clus-ters have easily become the most viable method of host webservers.The very structure of clusters,with several nodesconnected with high performance local interconnects,hasseen the emergence of a popular web-serving systems:data-centers.With the explosive growth of the adoption of Inter-Since lack of duplication of content incurs higher data-transfer overheads for cache retrieval,traditional net-work hardware/software architectures that impose signifi-cant load on the server CPUs and memory cannot benefit fully from these.On the other hand,Remote Direct Mem-ory Access(RDMA)enabled network interface cards(e.g. InfiniBand)are capable of providing reliable communica-tion without server CPU’s intervention.Hence,we design our cache cooperation protocols using one-sided operations to alleviate the possible effects of the high volume of data transfers between individual cache and sustain good overall performance.Further,current generation data-centers have evolved into complex multi-tiered structures presenting more inter-esting design options for cooperative caching.The nodes in the multi-tier data-center are partitioned into multiple tiers with each tier providing a part request processing function-ality.The end client responses are generated with a collec-tive effort of these tiers.The front-end proxy nodes typi-cally perform caching functions.Based on the resource us-age,we propose the use of the available back-end nodes to assist the proxies in the caching services.Also,this back-end server participation and access to cache is needed for several reasons:(i)Invalidating caches when needed,(ii) Cache usage by the back-end and(iii)Updating the caches when needed[11].The constraint of providing these ac-cess mechanisms to the back-end nodes increases the over-all complexity of the design and could potentially incur ad-ditional overheads.In our design,we handle this challenge by introducing additional passive cooperative cache system processing modules on the back-end servers.The benefits of these cache modules on the back-end servers are two-fold in our design:(i)They provide the back-end servers access to the caching system and(ii)They provide better overall performance by contributing a limited amount resources of the back-end server when possible.We implement our system over InfiniBand using Apache Web and Proxy Servers[10].We further evaluate the var-ious design alternatives using multiple workloads to study the trade offs involved.The following are the main contri-butions of our work:A Basic RDMA based Cooperative Cache design:wepropose an architecture that enables proxy servers like apache to cooperate with other servers and deliver high performance by leveraging the benefits of RDMASchemes to improve performance:We propose three schemes to supplement and improve the performance of cooperative caches-(i)Cooperative Cache With-out Redundancy,(ii)Multi-Tier Aggregate Coopera-tive Cache and(iii)Hybrid Cooperative CacheDetailed Experimental evaluation and analysis of the trade offs involved.Especially the issues associatedwith working-set size andfile sizes are analyzed in de-tailOur experimental results show throughput improvements of up to35%for certain cases over the basic cooperative caching scheme and improvements of Upton180%over simple caching methods.We further show that our schemes scale well for systems with large working-sets and large files.The remaining part of the paper is organized as follows: Section2provides a brief background about InfiniBand,and multi-tier data-centers.In Section3we present the design detail of our implementation.Section4deals with the de-tailed performance evaluation and analysis of our designs. In Section5,we discuss current work in relatedfields and conclude the paper in Section6.2Background2.1InfiniBand ArchitectureInfiniBand Architecture[4]is an industry standard that defines a System Area Network(SAN)that offers high bandwidth and low latency.In an InfiniBand network, processing nodes and I/O nodes are connected to the fab-ric by Host Channel Adapters(HCA)and Target Channel Adapters,respectively.An abstraction interface for HCA’s is specified in the form of InfiniBand Verbs.InfiniBand sup-ports both channel and memory semantics.In channel se-mantics,send/receive operations are used for communica-tion.To receive a message,the receiverfirst posts a receive descriptor into a receive queue.Then the sender posts a send descriptor into a send queue to initiate data transfer. In channel semantics there is a one-to-one match between the send and receive descriptors.Multiple send and receive descriptors can be posted and consumed in FIFO order.The memory semantic operation allows a process to write to a virtually contiguous buffer on a remote node.Such one-sided operation does not incur software overhead at the re-mote side.Remote Memory Direct Access(RDMA)Read, RDMA Write and Remote Atomic Operations(fetch-and-add and compare-and-swap)are the supported one-sided operations.2.2Multi-Tier Data-CentersA typical data-center architecture consists of multiple tightly interacting layers known as tiers.Each tier can contain multiple physical nodes.Figure1shows a typi-cal Multi-Tier Data-Center.Requests from clients are load-balanced by the edge services tier on to the nodes in the front-end proxy tier.This tier mainly does caching of con-tent generated by the other back-end tiers.The other func-tionalities of this tier include embedding inputs from vari-ous application servers into a single HTML document(forframed documents for example),balancing the requests sent to the back-end based on certain pre-defined algorithms.The middle tier consists of two kinds of servers.First, those which host static content such as documents,im-ages,musicfiles and others which do not change with time. These servers are typically referred to as web-servers.Sec-ond,those which compute results based on the query itself and return the computed data in the form of a static docu-ment to the users.These servers,referred to as application servers,usually handle compute intensive queries which in-volve transaction processing and implement the data-center business logic.The last tier consists of database servers.These servers hold a persistent state of the databases and other data repos-itories.These servers could either be compute intensive or I/O intensive based on the query format.For simple queries, such as search queries,etc.,these servers tend to be more I/O intensive requiring a number offields in the database to be fetched into memory for the search to be performed.For more complex queries,such as those which involve joins or sorting of tables,these servers tend to be more compute intensive.3Design and Implementation of Proposed Cooperative Cache SchemesIn this section,we propose four schemes for cooperative caching and describe the design details of our schemes.At each stage we also justify our design choices.This section is broadly categorized into four main parts:(i)Section3.1: RDMA based design and implementation of basic cooper-ative caching,(ii)Section3.2:Design of No Redundancy scheme,(iii)Section3.3:Multi-tier extensions for Cooper-ative caches and(iv)Section3.4:A combined hybrid ap-proach for Cooperative caches.Wefirst start with a detailed design description of the common components of all our schemes.External Module:As described earlier in Section2, proxy server nodes provide basic caching services in a multi-tier data-center.The traditional data-center applica-tions service requests in two ways:(i)by using different server threads for different concurrent requests or(ii)by using single asynchronous server to process to service re-quests.Catering to both these approaches,our design uses an asynchronous external helper module to provide cooper-ative caching support.Figure2shows the the typical setup on each node.This module handles inter-node communi-cation by using InfiniBand’s native Verbs API(V API)and it handles intra-node communication with the basic data-center applications using IPC.This module is designed to be asynchronous to handle multiple overlapping requests from the data-center applications.Soft Shared State:The cache meta-data information is maintained consistent across the all the servers by using a home node based approach.The cache entry key space (called key-table)is partitioned and distributed equally among the participating nodes and hence all the nodes han-dle a similar amount of meta-data key entries.This ap-proach is popularly know as the home node based approach. It is to be noted that in our approach we just handle the meta-data on the home node and since the actual data itself can reside on any node,our approach is much more scalable than the traditional home node based approaches where the data and the meta-data reside on the home node.All modifications to thefile such as invalidations,loca-tion transfers,etc.are performed on the home node for the respectivefile.This cache meta-data information is period-ically broadcasted to other interested nodes.Additionally, this information can also be requested by other interested nodes on demand.The information exchange uses RDMA Read operations for gathering information and send-receive operations for broadcasting information.This is done to avoid complex locking procedures in the system.Basic Caching Primitives:Basic caching operations can be performed using a small set of primitives.The in-ternal working caching primitives needs to be designed ef-ficiently for scalability and high performance.Our vari-ous schemes implement these primitives in different ways and are detailed in the following sub-sections.The basic caching primitives needed areCache Fetch:To fetch an entity already present in cacheCache Store:To store a new entity in cacheCache Validate:To verify the validity of a cached en-tityCache Invalidate:To invalidate a cache entity when neededBuffer Management:The cooperative cache module running on each node reserves a chunk of memory.This memory is then allocated to the cache entities as needed. Since this memory needs to be pooled into the global coop-erative cache space,this memory is registered(i.e.locked in physical memory)with the InfiniBand HCA to enable effi-cient memory transfers by RDMA.Several researchers have looked at the different aspects of optimizing this limited buffer usage and have suggested different cache replace-ment algorithms for web caches.Our methods are orthog-onal to these issues and can easily leverage the benefits of the proposed algorithms.3.1Basic RDMA based Cooperative Cache(BCC)In our design,the basic caching services are provided by a set of cooperating modules residing on all the participatingInternetApplications ApplicationsFront−end Mid−tierBack−end Figure 1.A Typical Multi-Tier Data-Center (Courtesy CSP Architecture design [12])Figure 2.External Module based Designserver nodes.Each cooperating module keeps track of the local cache state as a set of local page-tables and places this information in the soft shared state for global access.The basic RDMA based Cooperative Caching is achieved by designing the cache primitives using RDMA operations.The communication messages between the modules are divided into two main components:(i)control messages and (ii)data messages.The control messages are further classified into (i)meta-data read messages and (ii)meta-data update messages.Since data messages form the bulk volume of the total communications we use RDMA operations for these.In addition,the meta-data read mes-sages use the RDMA Read capabilities.Meta-data update messages are exchanged using send-receive operations to avoid concurrency control related issues.The basic cache primitives are handled by BCC in the following manner:Cache Fetch involves three simple steps:(i)finding the cache entry (ii)finding a corresponding amount of local free space and (iii)fetching the data using RDMA Read opera-tion.Cache Store involves the following steps:in case the local node has enough free space the entity is cached and key-table is updated.In cases where local node has no free memory,the entity is stored into a temporary buffer and the local copies of all page tables are searched for a suitable candidate remote node for a possible free space.A control message is sent to that node which then performs an RDMA Read operation of this data and notifies the original node of the transfer.Once a control message is sent with a storerequest to a remote node,then the current entity is consid-ered to be a responsibility of the remote node.For both these primitives,in cases where free space is not available system-wide,a suitable replacement is chosen and data is stored in place of the replacement.Cache Validate and Cache Invalidate involve a meta-data read or a meta-data update to the home node respectively.As mentioned earlier,RDMA Read is used for the read op-eration.Although this scheme provides a way to share cache across the proxy nodes,there may be redundancy in the en-tries across the system.3.2Cooperative Cache Without Redundancy(CCWR)In this scheme,the main emphasis is on the redundant duplicates in the system.At each step of request processing,the modules systematically search the system for possible duplicate copies of cache entities and these are chosen for replacement.In aggregate,the cache replacement decisions are taken in the following priority:(i)Local free space,(ii)Remote node free space,(iii)Local redundant copies of en-tries cached elsewhere in the system,(iv)remote redundant copies have duplicates in the system and (v)replacement of suitable entity by removing an existing entry to make space for the new entry.We again describe the details of designs of the cache primitives.The case of Cache Fetch presents interesting design op-tions.The data from remote node is fetched into local free space or in place of local redundant copy in the same prior-ity order.However,in case there are no free buffer spaces or local duplicates available for getting the data,remote cache entity is swapped with some local cached entity.In our de-sign,we select a suitable local replacement,send a store message to the remote cache for this local replacement and followed by a RDMA Read of the required remote cache entity.The remote node follows a similar mechanism to decide on storage and sends back an acknowledgment.Fig-ure3shows the swap case of this scheme.The dotted lines shown in thefigure are control messages.Cache Store design in this case is similar to the previ-ous approach,the main difference being the priority order described above.The memory space for storing new cache entries being searched in the order of free space,redundant copies and permanent replacements.The CCWR scheme benefits significantly by increas-ing the total amount of memory available for cooperative caching by removing redundant cache entries.For large working sets this yields higher overall performance.3.3Multi-Tier Aggregate Cooperative Cache(MTACC)In typical multi-tier data-centers proxy servers perform all caching operations.However,the system can benefit sig-nificantly by having access to additional memory resources. There are several back-end nodes in the data-center that might not be using their memory resources to the maximum extent.In MTACC,we utilize this free memory on servers from other tiers of the multi-tier data-center.This provides us with more aggregate system memory across the multi-ple tiers for cooperative caching.Further,the involvement back-end modules in caching can be possibly extended to the caching support for dynamically changing data[11].MTACC scheme is designed with passive cooperative caching modules running on the back-end servers.These passive modules do not generate cache store or retrieve re-quests themselves,but help the other modules to utilize their pooled memory.In addition,these passive modules do not act as home nodes for meta-data storage,minimizing the necessity for cache request processing overheads on these back-end servers.In addition,in certain scenarios like in case of cache in-validates and updates,the back-end servers need to initi-ate these invalidate operations[11].Utilizing the modules existing on the back-end nodes,the back-end nodes can perform operations like invalidations,etc.efficiently with the help of the closer and direct access to cache to achieve significant performance benefits.Figure4shows a typical setup for MTACC.3.4Hybrid Cooperative Cache(HYBCC)Though the schemes CCWR and MTACC can achieve good performance by catering to larger working sets,they have certain additional working overhead to remove redun-dant cache entries.While this overhead does not impact the performance in cases when the working set is large or when the requestedfile is large,it does impact the performance of the smaller cache entities or smaller working setfiles to a certain extent.CCWR adds certain overhead to the basic cache process-ing.The added lookups for duplicates and the higher cost of swapping make up these overheads.MTACC also adds similar overheads.This aggregated cache system size can cause higher overheads for request processing.To address these issues,we propose the use of the Hybrid Cooperative Caching Scheme(HYBCC).In this scheme, we employ different techniques for differentfile sizes.To extent possible,smaller cache entities are not checked for duplications.Further,the smaller cache entities are stored and their lookups are performed on only the proxy servers without using the web servers.So smaller cache entities are not stored on the passive nodes and are duplicated to the extent possible reducing the effect of the associated over-heads.Our experimental results show that this can achieve a good balance for different kinds of traces.4Experimental ResultsIn this section,we present a detailed experimental eval-uation of our designs.Here,we compare the following lev-els of caching schemes:(i)Apache default caches(AC)(ii) BCC,(iii)CCWR,(iv)MTACC and(v)HYBCC.Experimental Testbed:For our experiments we used 20nodes with dual Intel Xeon 2.66GHz processors. InfiniBand network connected with Mellanox InfiniHost MT23108Host Channel Adapters(HCAs).The clusters are connected using a Mellanox MTS14400144port switch. The Linux kernel version used was2.4.20-8smp.Mellanox IBGD1.6.1with SDK version3.2and the HCAfirmware version3.3was used.These nodes were setup with two web-servers and with the number of proxy servers varying from two to eight.The client requests were generated from multiple threads on10 nodes.The web-servers and application servers used in the reference implementation are Apache2.0.52.All proxy nodes we configured for caching of data.Web server nodes were also used for caching for the schemes MTACC and HYBCC as needed.Each node was allowed to cache64 MBytes of data for any of the experiments.Traces Used:Four synthetic traces representing the working sets in Zipf[15]traces were used.Thefiles sizes in the traces were varied from8k bytes to64k bytes.Since theRequest forFile B Present File BIn CacheFigure 3.Cooperative Caching Without RedundancyServer NodeProxy TierBack−End TierGlobal Cache SpaceLocal CacheFigure 4.Multi-Tier Aggregate Cooperative Cachingworking sets of Zipf traces all have similar request probabil-ities,a trace comprising of just the working set is seemingly random.The working set sizes for these traces are shown in Table1.These present us with a number of cases in which the working sets are larger than,equal to or smaller than the total cache space available to the caching system.Table 1shows the comparison working set size and the system cache size for various cases.Trace2nodes8nodes8k-trace80M/128M80M/512M 16k-trace160M/128M160M/512M 32k-trace320M/128M320M/512M 64k-trace640M/128M640M/512M Table1.Working Set and Cache Sizes for VariousConfigurations4.1Basic PerformanceAs an indication of the potential of the various caching schemes,we measure the overall data-center throughput. Figures5and6show the throughput measured for the four traces.We see that the basic throughput for all the coopera-tive caching schemes are significantly higher than the base case of basic Apache caching(AC).Impact of Working Set Size:We notice that the per-formance improvements from the AC scheme to the other schemes show steep improvements when the cooperative caching schemes can hold the entire working set of that trace.For example,the throughputs for the cooperative caching schemes for the8k-trace for two nodes in Fig-ure5are about10000TPS,where as the performance for AC is just above5000TPS.This shows a performance im-provement of about a factor of two.This is because the AC scheme cannot hold the working set of the8k-trace which is about80MBytes.Since each node can hold64 MBytes,AC incurs cache misses and two node coopera-tive caching shows good performance.We see similar per-formance jumps for all cases where the working setfits in cache.Figure7clearly shows a marked improvement for larger traces(32k-trace and64k-trace)for MTACC and HY-BCC.This benefit comes from the fact that MTACC and HYBCC can accomodate more of the working set by aggre-gating cache from nodes across several tiers.Impact of Total Cache Size:The total cache size of the system for each case is as shown in Table1.For each con-figuration,as expected,we notice that the overall system performance improves for the cases where the working-set sizes are larger then the total system cache size.In partic-ular,the performance of the64k-trace for the8node case achieves a throughput of about9500TPS while using the memory aggregated from the web server for caching.This clearly shows an improvement of close to20.5%improve-ment over basic caching scheme BCC.Impact of System Size:The performance of the8k-trace in Figure8shows a drop in performance for the CCWR and the MTACC cases.This is because as a result of aggre-gated cache across tiers for MTACC its total system size increases,hence the total overheads for each lookup also increase as compared to CCWR.On the other hand,since HYBCC uses CCWR for small cache entities and MTACC for large cache entities,its improvement ratios of HYBCC in Figure8clearly show that the HYBCC scheme does well in all cases.4.2Detailed AnalysisIn this section,we discuss the performance benefits seen for each of the schemes and analyze the same.AC:These numbers show the system throughput achiev-able by using the currently available and widely used simple single node caching.Since all the nodes here take local de-cisions the performance is limited by the amount of cache available on individual nodes.BCC:As shown by researchers earlier,the performance of the BCC scheme marks significant performance improve-ment over the AC scheme.These performance numbers hence represent throughput achievable by basic cooperative caching schemes.In addition,the trends for the BCC per-formance also show the effect of working-set size as men-tioned earlier.We see that as we increase the number of proxy servers,the performance benefit seen by the BCC scheme with respect to AC increases.The performance ben-efit ratio as shown in Figures7and8clearly shows this marked improvement.CCWR:From the Figures5and6,we observe that the performance for the CCWR method shows two interesting trends:(i)the performance for the traces16k-trace,32k-trace and64k-trace show improvement of up to32%as compared to the BCC scheme with the improvement grow-ing with higher size traces and(ii)the performance of the 8k-trace shows a drop of about5%as compared to the BCC scheme.The primary reason for this performance drop is the cost of additional book-keeping required for eliminat-ing copies.We measured this lookup cost for this scheme to be about5%to10%of the total request processing time for afile of8Kbytes size.Since this cost does not grow withfile size,its effect on largerfile sizes is negligible.MTACC:The main difference between the CCWR scheme and the MTACC scheme is the increase in the to-tal system cache size and the total system meta-data infor-mation size.The additional system size improves perfor-mance by accommodating more entities in cache.On the other hand,the higher meta-data size incurs higher lookup and synchronization costs.These reasons both show effecton the overall performance of the data-center.The8node case in Figure6shows that the performance of8k-trace de-creases with MTACC as compared to BCC and CCWR and the performance improves for16k-trace,32k-trace and64k-trace.We observe similar trends for the2node case in Fig-ure5.HYBCC:HYBCC overcomes the problems of lower performance for smallerfiles as seen above by using a hy-brid scheme described in Section3.4.In this case,we ob-serve in Figures5and6that the HYBCC scheme matches the best possible performance.Also,we notice that the im-provement of the HYBCC scheme over the BCC scheme is up to35%.5Related WorkSeveral researchers[11][8][2][5]have focussed on the various aspects of caching.Cooperation of multiple servers is proposed as an improtant technique in caching[5][9].A popular approach of cooperative caching proposed in[5] uses application level redirects of requests to enable coop-erative caching.This approach needs all the data-center servers to have different external IP addresses visible to the client and incurs higher overheads.On the other hand ap-proaches like[9][1]use either a home node based approach for the data or use a single node for management activities. These approaches could easily lead to performance bottle-necks.In our approach,we use the concept of home node for just the meta-data instead of the actual cached data.This alleviates the bottleneck problem to a large extent.In our approach,we use an approach similar to the N-Chance approach proposed in thefile-system research con-text in XFS[13].Significant work[7][6][3]has done in respect to the cache replacement algorithms.Our proposed schemes are orthogonal to these and can easily leverage the benefits of these.6ConclusionThe importance of Caching as an instrument for improv-ing the performance and scalability of web-serving data-centers is immense.Existing cooperative cache designs of-ten partially duplicate cached data redundantly on multiple servers for higher performance while optimizing the data-fetch costs for multiple similar requests.With the advent of RDMA enabled interconnects these cost estimates have changed the basic factors involved.Further,the utilization of the large scale of resources available across the tiers in today’s multi-tier data-centers is of obvious importance.In this paper,we have presented cooperative cache schemes that have been designed to benefit in the light of the above mentioned trends.In particular,we have designed schemes that take advantage of RDMA capabilities of net-works and the resources spread across the multiple tiers of modern multi-tier data-centers.Our designs have been im-plemented on InfiniBand based clusters to work in conjunc-tion with Apache based servers.We have evaluate these with appropriate request traces.Our experimental results have shown that our schemes perform up to35%better than the basic cooperative caching schemes for certain cases and 180%better than the simple single node caching schemes.We further analyze the performances of each of our schemes and propose a hybrid caching scheme that shows high performance in all cases.We have observed that sim-ple caching schemes are better suited for cache entities of small sizes and advanced schemes are better suited for the larger cache entities.As future work we propose to extend our work to support dynamic data cooperative caching.7AcknowledgmentsWe thank Karthikeyan Vaidyanathan and Pavan Balaji for their contribution towards the basic data-center setup and their valuable comments.References[1]WhizzBee Web Server./.[2]Marc Abrams,Charles R.Standridge,Ghaleb Abdulla,StephenWilliams,and Edward A.Fox.Caching proxies:limitations and po-tentials.In Proceedings of the4th International WWW Conference, Boston,MA,December1995.[3]Multimedia Proxy Across.Cost-based cache replacement and serverselection for.[4]Infiniband Trade Association.http://www.infi.[5]Scott M.Baker and Bongki Moon.Distributed cooperative Webputer Networks and ISDN Systems,31(11-16):1215–1229,May1999.[6]P.Cao and S.Irani.Greedydual-size:A cost-aware www proxycaching algorithm,1997.[7]Pei Cao and Sandy Irani.Cost-aware WWW proxy caching algo-rithms.In Proceedings of the1997Usenix Symposium on Internet Technologies and Systems(USITS-97),Monterey,CA,1997.[8]Francisco Matias Cuenca-Acuna and Thu D.Nguyen.Cooperativecaching middleware for cluster-based servers.In Tenth IEEE Inter-national Symposium on High Performance Distributed Computing (HPDC-10).IEEE Press,2001.[9]Li Fan,Pei Cao,Jussara Almeida,and Andrei Broder.Summarycache:A scalable wide-area Web cache sharing protocol.In Pro-ceedings of the ACM SIGCOMM’98conference,pages254–265, September1998.[10]The Apache Foundation./.[11]S.Narravula,P.Balaji,K.Vaidyanathan,S.Krishnamoorthy,J.Wu,and D.K.Panda.Supporting Strong Coherency for Active Caches in Multi-Tier Data-Centers over InfiniBand.In Proceedings of System Area Networks(SAN),2004.[12]Hemal V.Shah,Dave B.Minturn,Annie Foong,Gary L.McAlpine,Rajesh S.Madukkarumukumana,and Greg J.Regnier.CSP:A Novel System Architecture for Scalable Internet and Communication Ser-vices.In the Proceedings of the3rd USENIX Symposium on Internet Technologies and Systems,pages pages61–72,San Francisco,CA, March2001.。
Accurate Circuit Model of Interdigital Capacitor and Its Application to Design of New Quasi-Lumped Miniaturized Filters with Suppression ofHarmonic ResonanceLei Zhu,Member,IEEE,and Ke Wu,Senior Member,IEEEAbstract—A general-purpose circuit model of microstrip inter-digital capacitor(IDC)is presented in this paper for use in thedesign of new quasi-lumped miniaturized filters.This computer-aided-design-oriented model is developed as a versatile admittance(a)(b)(c)(d)Fig.1.Cross-sectional view and physical layout description of three types of IDC’s with predesignated geometrical dimensions and parameters of structure.(a) Cross section of multiple finger.(b)IDC(Type A).(c)IDC(Type B).(d)IDC(Type C).(a)(b)(c)Fig.2.Physical topology and equivalent-circuit representation prepared for MoM characterization of a generalized IDC attached with two different lines.(a) Physical layout.(b)Susceptance network.(c)J-inverter network.a unified equivalent-circuit network approach[12].Two direct-coupled bandpass filters are designed and attractive features ofa new class of filters using series-connected inductance line areobserved theoretically and experimentally as to size reductionand suppression of harmonic resonances.II.C IRCUIT M ODEL OF IDC S TRUCTUREFig.1(a)describes the cross-sectional view of a typical IDCwith multiple fingers while Fig.1(b)–(d)depict a geometricalsketch of three types of IDC’s[1]–[4]to be studied that are at-tached with two lines as the input/output.In the following,a gen-eral-purpose circuit model is described for efficient design andoptimization of IDC-related low-loss miniaturized filters basedon the versatile lumped-element network approach.To do so,webegin with a deterministic MoM algorithm to simulate IDC’s[9].The SOC technique is then applied to extract a CAD-ori-ented circuit model from field calculations and it is embedded inthe MoM algorithm[13]–[16].In the end,a andZHU AND WU:ACCURATE CIRCUIT MODEL OF IDC 349gates,and circuitdiscontinuitymatrixrepresenting the error term,,andand(1)inwhich and-inverter topology,consisting of a suscep-tance and two equivalent lines,as shown in Fig.2(c).This al-ternative notation is useful for design and optimization using the network-based synthesis technique [12].Theis directly proportional to the coupling (or strength)of the IDC,while the equivalence of two attached electricallines350IEEE TRANSACTIONS ON MICROWA VE THEORY AND TECHNIQUES,VOL.48,NO.3,MARCH2000(a)(b)Fig.4.Extracted parameters of the equivalent network of an IDC (Type A)as a function of frequency versus the number of finger with N =2,4,6,8,10(even).(a)J -inverter susceptance J .(b)Equivalent electrical line length =2.(2b)(2c)inwhich.=2andZHU AND WU:ACCURATE CIRCUIT MODEL OF IDC351(a)(b)Fig.6.Predicted and measured insertion and return losses for a simple IDC structure (Type B with N =5)connecting with two 50- lines.(a)Layout sketch.(b)Frequency responses.point,the susceptance is strongly enhancedfor and ittends to be gradually saturatedforversus frequency,except for minor nonlinear disturbance for alarge-inverter network param-eters for the IDC attached with two identical50-(5,7,9).With reference to Fig.5(a),similardistributed behaviors can also be observed,and the saturation effect takes place around 2.5GHz in,for example,the caseofGHz for thethree cases indicate that the conventional approximate lumped model is no longer valid.As plotted in Fig.5(b),the two elec-trical lengths have similar variation with frequency as compared with Fig.4(b),but there is a visible difference due to the choice of odd number of fingers.In particular,the two electrical phases are not identical with the structure asymmetry.To validate our proposed circuit model,a simple IDC circuit is fabricatedwith=2and352IEEE TRANSACTIONS ON MICROWA VE THEORY AND TECHNIQUES,VOL.48,NO.3,MARCH2000(a)(b)Fig.8.Equivalence of a global circuit network of an IDC line resonator designed for CAD-oriented model and its predicted and measured frequency responses.(a)Physical layout and network equivalence.(b)Measured and predicted frequency responses.IV .F ILTER D ESIGN WITH S IZE R EDUCTION AND S UPPRESSIONOF H ARMONIC R ESONANCE In this section,a class of IDC-related filters is designed with the proposed circuit model and our interests are to realize size reduction and suppression of filtering harmonic resonance with new topologies.With the developed circuit model for the three types of IDC presented above,we choose one of the asymmet-rical IDC’s for use as a tight coupling element for low-loss filter and resonator design.In our study,one IDC line resonator and two bandpass filters are studied theoretically and experi-mentally.Considering some practical aspect of our experiments,a dielectric substrate of Duroid is fixed with the thickness of1.27mmandthrough the following calculations.A.IDC Line ResonatorFig.8(a)displays the proposed layout and complete CAD-ori-ented circuit model of a line resonator interconnected by twoasymmetrical IDC structures.The two external lines(50)in stripwidth,and the number offingerand the two identical equiv-alent linelengths-inverter,as shown inFig.8(a).Since thedeveloped,the bandpass becomes broader for the higher order of resonant frequency.This is essentially due to the coupling enhancement at higher frequency,as shown in Fig.7(a),through the charac-teristicsof.B.Miniaturized IDC Bandpass Filters with Suppression ofHarmonic ResonanceThe above example points to the effectiveness and generality of our proposed model for the IDC circuit design using a net-work approach.In this section,a very efficient network-based synthesis and optimization procedure,as detailed in [13]–[15],is applied to the design of high-quality IDC circuits with mini-mized computational efforts as opposed to direct,but time-con-suming,field-theory-based techniques [10].In this way,the en-tire filter is represented in design by an equivalent cascaded cir-ZHU AND WU:ACCURATE CIRCUIT MODEL OF IDC353(a)(b)(c)Fig.10.Extracted design curves of the equivalent network as a function of frequency versus the length of finger section for an asymmetrical IDC.(a)Structure and network sketches.(b)Self-susceptance.(c)Mutual-susceptance.cuit topology that consistsof-inverter parameterscan be calculated in an explicit form with reference to a series of standard design tables or closed-form equations in [13]–[15].This is usually related to a specific frequency response,such as a maximally flat or Chebyshev type.In our case,each line resonator is designed to operate at the central frequency and its equivalent sloping susceptance parameter is derived from the differentiation of a complete circuit network,as indicated in [21].In this study,our interest is to design Chebyshev-type IDC bandpass filters.In particular,a new class of miniaturized IDC filters is proposed to suppress the unwanted harmonic resonance over the upper frequency range,thus significantly improving the filter performance and adjacent channel rejection.Fig.9(a)and (b)presents the physical layouts and CAD-ori-ented model of two types of three-pole direct-coupled bandpass filters that use the uniform-line resonator and inductive line-loaded resonator,respectively.These filters operate with quasi-lumped elements whose circuit parameters can accu-rately be extracted from the above-described SOC technique.Fig.9(a)shows two asymmetrical IDC coupling elements for interconnecting the line resonators so that a tight coupling is conceptually achieved.Simple gap-coupling structures are used to provide a weak coupling between two adjacent-line resonators.This type of filter structure is usually long in size because the line resonator takes up an appreciable fraction of wavelength.To achieve a dimensional compactness,the resonator lines are split into two pieces and a series-connected shorted stub element having electrically narrow width is inserted to bridge them,as shown in Fig.9(b).The elementgenerates inductive effect such that the quasi-lumped resonance can take place at lower frequency and the new filter size can significantly be reduced as compared to Fig.9(a).Obviously,accurate parameter extraction of the quasi-lumped elements is critical for successful design of the proposed filter topology,which should be done with the developed MoM algorithm and SOC technique.Fig.10(b)and (c)plots a number of design curves for the ex-tracted self-and mutual-susceptance parameters as a functionof frequency for the asymmetrical IDCwith,as shown in Fig.10(a).Within the frequency range of interest (global-po-sitioning-system (GPS)band around 1.575GHz),the suscep-tances exhibit a lumped capacitance that gradually increases with extended finger length.Of course,the distributed effects are pronounced over the higher frequency range over where the wavelength is comparable to the finger length.Fig.11(b)presents a series of curves for the equivalent capacitance versus gap spacing for the gap coupling topology at three different frequencies,whose layout parameters and circuit representa-tion are described in Fig.11(a).Theself-capacitancethat are veryuseful in design of the new size-reduced IDC filters.It is con-firmed that the mutual admittance really points to the effect of a quasi-lumped series inductance,while the shunt susceptance points to a shunt capacitance.The extracted circuit parameters of Figs.10–12are used to design two three-pole bandpass filters with a center frequency354IEEE TRANSACTIONS ON MICROWA VE THEORY AND TECHNIQUES,VOL.48,NO.3,MARCH2000(a)(b)Fig.11.Calibrated equivalent capacitance of a gap coupling structure for three different frequencies.(a)Physical layout and network equivalence.(b)Self-capacitance and mutualcapacitance.(a)(b)Fig.12.Calibrated equivalent inductance and capacitance of a series-connected shorted stub element as a function of frequency versus the length of the loaded line section.(a)Physical layout and network equivalence.(b)Shunt capacitance and series inductance.of 1.575GHz and a bandwidth of 2.7%(not necessarily a GPS requirement).Fig.13(a)shows predicted and measured inser-tion and return losses of our first filter without inductive com-pensation over a narrow frequency range of 1.48–1.66GHz.The(a)(b)Fig.13.Predicted and measured results of frequency response for the simple IDC filter that is made up of uniform line resonators.(a)Zoom-in for narrow-band characteristics.(b)Broad-band frequency response.predicted center frequency is slightly down (within 1.5%)com-pared with the measured result,and the measured bandwidth is 2.4%.Fig.13(b)presents the panoramic frequency response,and the second harmonic resonance frequency is found around 3.15GHz,approximately equal to twice the fundamental res-onance.The reduction of 2.4dB in the insertion loss around 1.56GHz can be attributed to the conductor loss of the narrow strip of the IDC multiple finger.As described in [19]and [20],an alternate way of overcoming this problem is to use a super-conducting conductor for the design of extremely narrow-band filters.The length of above-described filter is a rather large 106.90mm.Now let us look at the design aspects of the inductively compensated IDC filters with the proposed circuit model.Fig.14(a)shows predicted and measured results for a filter that uses the same IDC and gaps as in the above filter design for the purpose of comparison.The length of series-connected shorted stub element is chosenasmm,as discussed in Fig.12(b),so that the total length of the new filter is significantly shortened to 68.58mm for the same operating frequency bandwidth as the above example.A slightZHU AND WU:ACCURATE CIRCUIT MODEL OF IDC355(a)(b)Fig.14.Predicted and measured results of frequency response for thecompensated IDC filter with inductive series-connected shorted stub loadedresonators and feature presentation for suppression of the second harmonicresonance.(a)Zoom-in for narrow-band characteristics.(b)Broad-bandfrequency response.increase in return loss is observed and it can be improved withsuitable adjustment of coupling enhancement of the IDC struc-ture.Fig.14(b)depicts the prediction of complete frequencyresponse over a broad band.It is interesting to find out thatthe second harmonic resonance is completely suppressed withthe inductive effect,attributed by the series-connected shortedstub element.This is a very attractive feature for practicalapplications.This interesting aspect of our design is alsodiscussed with regard to stepped-impedance resonator(SIR)filtering circuits[21].The additional insertion loss observed inthe experiments can be explained by the conductor loss of theadded narrow series-connected shorted stub elements[20].V.C ONCLUSIONA general-purpose circuit model is proposed for design andcharacterization of IDC-related structures using an accurate356IEEE TRANSACTIONS ON MICROWA VE THEORY AND TECHNIQUES,VOL.48,NO.3,MARCH 2000[14]J.Rautio,“Comments on revisiting characteristic impedance and its def-inition of microstrip line with a self-calibration 3-D MoM scheme,”IEEE Trans.Microwave Theory Tech.,vol.47,pp.115–117,Feb.1999.[15]L.Zhu and K.Wu,“Author's reply,”IEEE Trans.Microwave TheoryTech.,vol.47,pp.117–119,Feb.1999.[16],“Network equivalence of port discontinuity related to source plane in a deterministic 3-D method of moments,”IEEE Microwave Guided Wave Lett.,vol.8,pp.130–132,Mar.1998.[17]R.A.Soares,P.Gouzien,P.Legaud,and G.Follot,“A unified mathe-matical approach to two-port calibration techniques and some applica-tions,”IEEE Trans.Microwave Theory Tech.,vol.37,pp.1669–1674,Nov.1989.[18]L.Zhu and K.Wu,“A general-purpose circuit model of interdigitalcapacitor for accurate design of low-loss microstrip circuit,”in IEEE MTT-S Int.Microwave Symp.Dig.,Baltimore,MD,June 1998,pp.1755–1758.[19]R.R.Mansour,S.Ye,V .Dokas,B.Jolley,G.Thomson,W.C.Tang,andC.M.Kudsia,“Design considerations of superconductive input multi-plexers for satellite application,”IEEE Trans.Microwave Theory Tech.,vol.44,pp.1213–1228,July 1996.[20]ncaster,F.Huang,A.Porch,B.Avenhaus,and J.Hong,“Minia-ture superconducting filters,”IEEE Trans.Microwave Theory Tech.,vol.44,pp.1339–1346,July 1996.[21]M.Sagawa,M.Makimoto,and S.Yamashita,“Geometrical structuresand fundamental characteristics of microwave stepped-impedance res-onators,”IEEE Trans.Microwave Theory Tech.,vol.45,pp.1078–1085,July1997.Lei Zhu (S'91–M'93)received the Bachelor degree and Master degree of engineering from the Nanjing Institute of Technology (now Southeast University),Nanjing,Jiangsu,China,in 1985and 1988,respec-tively,and the Doctor Degree of Engineering from the University of Electro-Communications,Tokyo,Japan,in 1993.From 1985to 1989,he studied a variety of mil-limeter-wave passive/active circuits and leaky-wave antennas using a grooved nonradiative waveguide (GNRD).From 1989to 1993,he conducted studieson the full-wave characterization of planar transmission lines and components for monolithic microwave integrated circuits (MMIC's).From 1993to 1996,he was with the Matsushita-Kotobuki Electronics Industries,Ltd.Japan,as a Research Engineer,where he conducted research/development works in the design of linearly polarized (LP)and circularly polarized (CP)planar antennas/arrays with compact-size and high radiation gain/efficiency for applications in wireless communications.Since 1996,he has been with Ecole Polytechnique,University of Montréal,Montréal,P.Q.,Canada,where he is currently a Research Associate.His current works/interests include the study of planar dual-mode bandpass filters,planar antenna elements/arrays,uniplanar passive/active circuits and antennas,as well as full-wave 3-D modeling techniques,numerical calibration (deembedding)procedures,and field-theory-based CAD techniques.Dr.Zhu was awarded the Japanese Government (Monbusho)Graduate Fel-lowship (1989–1993),the First-Order Achievement Award in Science and Tech-nology presented by the National Education Committee of China (1993),the Silver Award of Excellent Invention presented by the Matsushita-Kotobuki Elec-tronics Industries Ltd.,Japan (1996),and the 1997Asia–Pacific Microwave Prize Award presented at the Asia–Pacific Microwave Conference,Hong Kong(1997).Ke Wu (M'87–SM'92)was born in Liyang,Jiangsu Province,China.He received the B.Sc.degree with distinction in radio engineering from the Nanjing Institute of Technology (now Southeast University),Nanjing,China,in 1982,and the D.E.A.and Ph.D.degree with distinction in optics,optoelectronics,and microwave engineering from the Institut Na-tional Polytechnique de Grenoble (INPG),Grenoble,France,in 1984and 1987,respectively.He conducted research in the Laboratoire d’Elec-tromagnetisme,Microondes et Optoelectronics(LEMO),Grenoble,France,prior to joining the Department of Electrical and Computer Engineering,University of Victoria,Victoria,B.C.,Canada.He subsequently joined the Department of Electrical and Computer Engineering,Ecole Polytechnique de Montréal (Faculty of Engineering,University of Montréal),Montreal,Canada,as an Assistant Professor,and is currently a Full Professor.He held a 1995Visiting Professorship in France (Telecom-Paris and INP-Grenoble),a 1996-1997Visiting Professorship at the City University of Hong Kong,a 1999Guest Professorship at the Swiss Federal Institute of Technology (ETH-Zürich),Zürich,Switzerland,as well as several short-term Visiting Professorships at other universities.He also holds an Honorary Visiting Professorship at the Southeast University,China.He is currently a Visiting Professor at the National University of Singapore.He has been the Head of the FCAR Research Group of Quebec,where he is involved with RF and millimeter-wave electronics,and the Acting Director of the Poly-Grames Research Center.He has authored or co-authored over 250refereed journals and conference papers and several book chapters.His current research interests involve 3-D hybrid/monolithic planar and nonplanar integration techniques,ac-tive and passive circuits,antenna arrays,advanced field-theory-based CAD and modeling techniques,and development of low-cost RF and millimeter-wave transceivers.He is also interested in the modeling and design of microwave photonic circuits and systems.He was chairperson of the 1996ANTEM Pub-licity Committee and vice-chairperson of the Technical Program Committee (TPC)for the 1997Asia–Pacific Microwave Conference (APMC’97).He has served on the FCAR Grant Selection Committee (1993–1996,1998–1999),and the TPC committee for the TELSIKS and ISRAMT.He has also served on the ISRAMT International Advisory Committee.He was the general co-chair of the 1999SPIE International Symposium on Terahertz and Gigahertz Photonics,Denver,CO.He serves on the Editorial Review Board for Microwave and Optical Technology Letters .Dr.Wu is a member of the Electromagnetics Academy.He has served on the Editorial or Review Boards of various technical journals,including the IEEE T RANSACTIONS ON M ICROW A VE T HEORY AND T ECHNIQUES ,the IEEE T RANSACTIONS ON A NTENNAS AND P ROPAGATION ,and the IEEE M ICROW A VE AND G UIDED W A VE L ETTERS .He served on the 1996IEEE Admission and Advancement (A&A)Committee and the Steering Committee for the 1997joint IEEE APS/URSI International Symposium.He has also served as a TPC member for the IEEE Microwave Theory and Techniques Society (IEEE MTT-S)International Microwave Symposium.He received a URSI Young Scientist Award,the Institute of Electrical Engineers (IEE)Oliver Lodge Premium Award,the Asia–Pacific Microwave Prize Award,the University Research Award "Prix Poly 1873pour l’Excellence en Recherche,"presented by the Ecole Polytechnique on the occasion of its 125th anniversary,and the Urgel-Archambault Prize (the highest honor)in the Field of Physical Sciences,Mathematics and Engineering from the French-Canadian Association for the Advancement of Science (ACFAS).。
Journal of Navigation and Port Research International EditionVol.35, No.9 pp. 707~715, 2011 (ISSN-1598-5725)DOI :/10.5394/KINPR.2011.35.9.707An Ergonomic Approach for Optimized Layout of Training Ship’s BridgeJun Hyuk Jang*, Hongtae Kim**, Joung Hoon Sim***, †Hea Sik Kim*, ** Korea Ocean Research & Development Institute, Daejeon 305-343, Republic. of Korea*** , †Department of Marine Systems Engineering, Dong-A University, Busan 530-729, Republic. of KoreaAbstract :To propose an ergonomic layout on the bridge, this study conducted a usability evaluation on the on-bridge navigational equipment of a college training ship that is in use at present. Through the usability evaluation on the training ship navigators, the possibility of operational errors with navigational equipment, the possibility of readout errors with display devices and even the effect of navigational equipment on navigation were evaluated beyond the scope of existing layouts, which adopt only the importance and usage frequency of navigational equipment. By taking ergonomic variables into consideration, this study suggested an optimized layout for on-bridge navigational equipment using a mathematical programming model.Key words :ship's bridge, layout, ergonomics, usability evaluation, maritime accident* jang@moeri.re.kr 042)866-3658** kht@moeri.re.kr 042)866-3643*** jhsim71@ 051)200-6974†Corresponding author, g9772690@donga.ac.kr 051)200-6975Note) This paper was presented on the subject of "An Ergonomic Approach for Optimized Layout of Training Ship's Bridge" in 2010 AsiaNavigation Conference proceedings(Songdo, InCheon, Korea, Nov.4-6, 2010, pp.3-10).1. IntroductionShips are one of the major forms of transportation forinternational exchange, and due to increases in global tradeand developments in scientifictechnology, ships are becoming increasingly larger, faster and more fully- automated. However, marine accidents still occur regularly. The fact that 60%∼80% of actual marine accidents are caused by human error indicates the potential risks related to a ship's operation (Hwang and Lee, 1999; Kim et al., 2001; Yang, 2004) as well as potential damage to the marine environment. As the rate of marine accidents due to human error is so high, a good deal of research has been conducted on this topic, yet accurate classification and survey methods have not yet been adequate enough to get a clear insight into what is taking place. Diverse efforts need to be invested to reduce the rate of marine accidents from human error. Ergonomic factors are increasingly being considered for sailing equipment along with institutional supplementation and support system maintenance. The bridge of the ship is where a ship's officers actually work and functions as an information situation room and a navigation control room(Kemp P. 1994). Therefore, in the case of ship design, an ergonomic bridge design should be considered to allow the officer to conducttheir operations properly, be adaptable to each given situation by monitoring the sailing environment and physical function of the ship, as well as to minimize the effects of fatigue on officers due to the overburden of work (Lee et al., 2008; Ha et al., 2002).To propose an ergonomic layout on the bridge, therefore,this study conducted a usability evaluation on the on-bridgenavigational equipment of a college training ship currently inuse. In general, there are many layout-only programs thathave been developed to solve problems related to layouts. Infact, many new layout standards have been applied to thelayout of control panels, such as spatial compatibility, usagefrequency, importance and usage order principles.Composing the main console of control panels installed ona ship's bridge, navigational equipment varies in formdepending on each manufacturer, and ship owners' choices ofequipment are so variable that all kinds of products made bydifferent specialized manufacturers are mixed together duringthe purchasing process and are consequently used together.Therefore, although some ships may have the same designform, they do not necessarily have their navigationalconsoles in the same layout.Therefore,when a ship's bridge is built, unlike automobiles or airplanes, it should be designed after being prudently studied from a navigator's viewpoint sinceAn Ergonomic Approach for Optimized Layout of Training Ship’s Bridgenavigators are to be the end users (Lee, 1997).The existing research was mostly conducted on the panel layout, considering only the importance and usage frequency of on-bridge navigational equipment, but this study attempted to provide an optimized plan for on-bridge navigational equipment using a mathematical programming model (Cem Canel, 1996), based on ergonomic variables, such as the possibility of operational errors and readout errors and the effect of each item of navigational equipment on actual navigation.Besides that, this study suggested an optimized layout by considering the weighted value of each item of on-bridge equipment, on the basis of the manual for coping with emergency situations that can take place while a ship is being navigated (ICS, 2007).2. Method2.1 Usability EvaluationAlthough the required equipment varies according to the ship's type or size, generally between 40 to 80 control or display devices are necessary for ship navigation. The subject was the university training ship, and the survey asked its operators about the importance of, usage frequency of, influence on voyage of, possibility of error from, and error experience gained related to the operation of a total of 75 pieces of equipment from the model ship, including navigational equipment and control and display devices. After evaluating their usability and optimization, this study suggests an optimal ergonomic layout of on-Bridge navigational equipment.Fig. 1Bridge of training shipFig. 1 shows the inner view of the bridge of the navigation training ship used to optimize the layout of navigational equipment. The size of navigational equipment and the distance between each item were actually measured (Fig. 4).Table 1 shows the characteristics of each piece of equipment within the bridge of the training ship when selected as a model for an actual ship's layout evaluation. It lists the evaluation scores of the training ship navigators on the actual dimensions (width×height) of 75 pieces of navigational equipment and control/display devices, and their panel dimensions, importance factor of optimization performance, usage frequency of each item of equipment, influence of the equipment on a voyage, and the possibility of error are listed.The research also accounted for different types of emergency (collision, grounding, fire & explosion, sinking, and machinery damage) and each of these criteria was accordingly weighted in the final analysis.Table 2 describes the nature of marine accidents that occurred in Korea during the period 2003∼2007(Oh, 1992). It includes a list of the percentages(%) calculated of the sum of each type, and the results were given as weights for each piece of sailing equipment and control or display devices. For example, when values used to perform the optimization for this study were examined with No. 4-1 "No.2 Auto tel & Inmarsat-f tel" , the score calculated using the usability evaluation of navigational equipment was a total of 20, in combination of such items as importance (5.33), usage frequency (5.67), effect on actual navigation (3.33) and operational or readout errors (5.67), as shown in Table 1. As shown in Table 2, the weighted values were calculated by using the composition ratio (27.3%) in the collision item out of all the marine accidents.Table 3 indicates that 4-1 "No.2 Auto tel & Inmarsat-f tel" is a navigational device used for marine accidents such as fire, explosion, stranding and sinking, and the numbers in brackets mean the number of times this equipment has been used. When the weighted values of this equipment were calculated for "sinking" accidents, it was found that this equipment accounted for the third, fourth, and eleventh sinking accident-coping situations in the checklist of the Emergency Situation Manual (ICS, 2007), showing the accumulated number of times it has been used. That is, it was found that this equipment had been used a total of 3 times. Therefore, the values obtained by multiplying the composition ratio of accidents and the accumulated number of times the related equipment has been used for an emergency situation as well as the usability evaluation value for this equipment were used to perform the final optimization.Jun Hyuk Jang, Hongtae Kim, Joung Hoon Sim, Hea Sik KimTable 1Characteristics of each item of equipment in BridgeThe weighted values for other kinds of emergencysituations were calculated in the same way, and during thisprocess, the optimization of on-bridge navigationalequipments was carried out.The table above represents the investigation results on thenavigational equipment required by the manual for eachemergency, departure and arrival.The numbers here correspond to the order of thecounter-respond manual examples of Table 4. For example,the Chart(no.2) of Table 3 includes the numbering of each emergency situation and departure/arrival. "10" is assigned for the ship's "Sinking", and this number indicates that during an emergency situation in Table 4, which is the ship's sinking, this equipment will be used for the 10th order, "when needed, implementation of arbitrary sinking". In other words, the numbering of Table 3 for each item of equipment and each situation means that the equipment in the corresponding order will be used for each situation.The number of this system shows the cumulative usage frequency of each kind of equipment for the correspondingAn Ergonomic Approach for Optimized Layout of Training Ship’s BridgeTable 3 The usage frequency of each item of equipment in an emergencySearch light remote control panel Outside light control panel 8Sensor control panel Navigation light control panel 101612Signal light control panel 1016124X-band radar 5422,2524Table 4Emergency circumstance - Groundingsituation, and it is multiplied by the configuration rate ofmaritime accident statistics(Table 1) and used as a weight value of the corresponding equipment on the actual performance of optimization.Jun Hyuk Jang, Hongtae Kim, Joung Hoon Sim, Hea Sik Kim( = importance, usage frequency, voyage influence, possibility of operating and/or reading error)(1)2.2 OptimizationAmong the respondents, the officer group wassub-categorized based on their job position and then a layout for each position was analyzed. The weighting for accident type used for optimizations is listed in Table 2. The LINGO 8.0 Global Solver program was used in the optimization process to derive an ergonomic layout of navigational equipment. The following are the symbols and variables used in the test. Indexi : Equipment ⋯j : Equipment ⋯ : coordinate at datum position: coordinate at datum positionDecision Variables : linear distance from the datum position to the centroid of equipment : coordinate at lower right side of equipment: coordinate at lower right side of equipment: coordinate at upper left side of equipment: coordinate at upper left side of equipment: coordinate at the center of equipment: coordinate at the center of equipmentConstants : Width of equipment: Height of equipment: Attribute score on item of equipment: Width of bridge: Height of bridge: Constant of emergency weight for equipmentObjective Function⋅⋅Objective Function (1) is an equation to minimize thedistance from the datum position to the centroid of each itemof equipment, in consideration of each item's attribute scorefor a specific item. For the attribute function, importance,usage frequency, influence on voyage, and possibility of errorscores were obtained from the findings of the primitivesurvey on ship operators.Constraint∀ (2)Constraint (2) is an equation to calculate a linear distance from the datum position to the centroid of a specific item of equipment.∀ ∀∀∀ (3)(4)(5)(6)Equations (3) to (6) are designed to calculate coordinations at the centroid point of each item of equipment. × × × ∀ (7)Equation (7) above is a constraint to avoid any double count of any equipment. ≦∀ ≦ ∀ (8)Equation (8) is a constraint to ensure that the layout does not exceed the range of the bridge's width and height. ≧ ∀ ≧ ∀ ≧ ∀ ≧ ∀ (9)(10)(11)(12)Equation (9) to (12) are the constraints to ensure that the final suggestion of coordinate values are given in positive numbers.An Ergonomic Approach for Optimized Layout of Training Ship’s Bridge3. Result3.1 OptimizationThe values obtained by multiplying the composition ratioof accidents and the accumulated number of times the relatedequipment has been used for an emergency situation, as well as the usability evaluation value for this item of equipment, were used to perform the final optimization.The weighted values for other kinds of emergency situations were calculated in the same way, and during this process, the optimization of on-bridge navigational equipment was carried out.Fig. 2 shows the results of the optimization.Fig. 2 Result of Layout of optimization"A" shows the results of the optimization using scores based on the level of importance and usage frequency of the item of equipment.The bridge in the training ship consists of a total of 18 panels, and the numbering system indicated in the tables represents the actual layout of navigational equipment on the Bridge."B", the Bridge model was drawn based on the optimization result with the emergency weights applied to the scores of importance, usage frequency, influence on voyage, and the possibility of operation and readout errors.ECDIS(12) and VHF of GMDSS(9) were placed the nearest to the datum point, and this suggests that they need to be located nearest to the officer for operations.3.2 Comparison and AnalysisThe purpose here is to compare and analyze the final optimized layout (Fig. 4) and the existing layout. The movingdistances were calculated and the efficiencies based onthem were compared to predict the traffic lines of the operator when reading the control or display devices of the navigational equipment for ship operation. The Bridge model for the training ship has been used as the existing model in the research and the suggested optimal Bridge model after applying the ergonomic variables (Fig. 4) are shown. For comparison and evaluation, the distances between the items of navigational equipment were measured (Fig. 3).Fig. 3 The existing layoutFig. 4Proposal layoutTable 5 suggests a checklist upon the ship's departure. The indicated numbers of navigational equipment, control and display devices correspond to the order of the checklist. The listed numbers on Table 4 were used as the numbers of navigational equipments.For example, Navigational light control panel(7-3) and Signal light control panel(7-5) are the equipment to be used in the event of question 10, "the signal flags of departure is prepared and hoisted?". The Navigation light control panel and Signal light panel of the existing layout are located at panel 7. In this case, the calculated moving distance of the operator from datum point (▲) is 3,580mm. Therefore, according to the comparison and analysis between the existing and the optimized layouts, the operator's moving distance was reduced by 1,000mm and this is only 72.1% of the moving distance before optimization. Consequently, it shows an increased efficiency of 27.9%. If conducting the comparison and analysis on the navigational equipment with the same methods, the results show that the suggested layout in this research has reduced the moving distance of the operator by 18.5%.Table 6 suggests a checklist upon a ship's arrival. The indicated numbers of navigational equipment, control and display devices correspond to the order of the checklist. The listed numbers on Table 3 were used as the numbers of navigational equipment.For example, VHF(4-1) is the equipment to be used in theJun Hyuk Jang, Hongtae Kim, Joung Hoon Sim, Hea Sik Kim Table 5Checklist of departure sequenceDeparture sequence EquipmentDistance(mm)Rate(%) Existing Optimized1Navigation plan is established? 2,3,126,0605,59092.2 2Necessary watercourse map is prepared? 2,3,126,0605,59092.23Steering operation is checked?- main steering gear, sub auxiliary steering gear- check if rudder angle indicator matches- steering gear, joints, exterior status- alarm device at breakout1,15-1, 15-23,2204,0203,2204,250100105.74Test drive of main engine is conducted?9-1(M-E rpm indicator)205035017.15Navigational equipments are testes and ready for use?(RADAR, ECDIS, AIS, GPS, COURSE RECORDER, ECHO SOUNDER, WHISTLEand DOPPLER LOG)1, 3, 2, 5, 8, (9-11, 9-12, 9-17),(VHF, MF/HF, whistle), 11, 12, 1314,77518,230123.46Mooring facilities are test-driven and ready for use?7Tuning between MAIN GYROCOMPASS and REPEATER is checked?8Moving objects including boats, accommodation ladder, derrick and etc are secured at the ship body?9The close/open status of each storage door and watertight sliding door is checked? 4-2(Watertight sliding door)4,7102,06043.7 10Signal flag of departure is prepared and hoisted? 7-3, 7-53,5802,58072.1 11Take-off(departure) report is completed?9-11(VHF)2,05035017.1 12Clean water, fuel and other necessary consumables are supplied?13Repairs are finished for repair-necessary parts? 14All crew are on board? 15All persons to leave the ship are off the ship? 16Documents and certificates of main ship that were sent to the lang are collected? 17Vulnerable area is patrolled such as Bosun Store, (2 hours before the departure/1hour before the departure/immediately after departure) to prevent stowaways onboard?18Parallel running of STEERING MOTOR is taken care of? 16-21(S-G alarm panel)1,8852,860151.7 19Binocular and DAY LIGHT SIGNAL are ready? Total48,41045,08081.5 Table 6Checklist of arrival sequenceDeparture sequence EquipmentDistance(mm)Rate(%) Existing Optimized1Navigation plan is established? 2, 3, 126,0605,59092.2 2Necessary watercourse map is prepared?3Steering operation is checked?- main steering gear, sub auxiliary steering gear- check if rudder angle indicator matches- steering gear, joints, exterior status- alarm device at breakout1,15-1, 15-23,2204,0203,2204,250100105.74Test drive of main engine is conducted?telegraph 5Navigational equipments are testes and ready for use?(RADAR, ECDIS, AIS, GPS, COURSE RECORDER, ECHO SOUNDER,WHISTLE and DOPPLER LOG)6Mooring facilities are test-driven and ready for use? 7Communication devices such as Microphone and transceiver are tested?8Mooring line, heaving line, rat guard and accommodation ladder are readyfor use? (when necessary, prepare anchoring)9Tuning between MAIN GYROCOMPASS and REPEATER is checked?1(Gyro compass)32203220100 10Notified at engine room before leaving the port?11Signal flag of port entry is prepared and hoisted?12Port entry report is completed? 9-11 2,05035017.1 13Notified expected port entry time to inland agent?4-14,7102,06043.7 14Notification of ETA to PILOT STATION and PILOT LADDER is prepared? 15Tide tables are ready and the recent weather chart reception is completed? 16Port entry document is ready? 17Clocks at Bridge and engine room are matched and phone is tested?(4-1, 9-14)(Auto Telephone) 4,7102,41051.2 18Parallel running of STEERING MOTOR is taken care of? 16-21(SG alarm panel)1,5102,660176.2 19Binocular and DAY LIGHT SIGNAL are ready? Total29,50023,76080.5An Ergonomic Approach for Optimized Layout of Training Ship’s Bridgeevent of question 13, "Notified expected port entry time to inland agent". In this case, the calculated moving distance of the operator from the datum point (▲) is 4,710mm(Fig. 3). The VHF in the suggested layout is located at 2,060mm(Fig.4). Therefore, according to the comparison and analysis between the existing and the optimized layouts, the operator's moving distance was reduced by 2,650mm and this is only 43.7% of the moving distance before optimization. Consequently, it shows an increased efficiency of 56.3%. If conducting the comparison and analysis on the navigational equipment with the same methods, the results show that the suggested layout in this research has reduced the moving distance of the operator by about 18.99%.In conclusion, the suggested layout in this study has reduced the operator's moving distance by 19.5%, compared to the existing layout, and consequently, the reduced moving distance is expected to lower the possibility of human error among maritime accidents caused by the officer's reduced fatigue and rapid counter-responses.4. ConclusionThe purpose of this research was to develop an optimized layout of a ship's navigational equipment on-Bridge, differently arranged according to the ship's size or type, with the consideration of ergonomic factors.This research went further by suggesting a ship's ideal Bridge layout based on the importance and usage frequency of navigational equipment that was investigated during prior research, and conducted a usability evaluation on the importance and usage frequency of navigational equipment on the training ship's bridge, observing the degree of each piece of equipment's influence on navigation, and the possibility of operation or readout errors for the ship's officers. Optimization was performed using the Lingo program from the survey results. Optimization of on-Bridge navigational equipments was performed based on the primary questionnaire survey.According to the comparison and analysis between the existing and the optimized layouts, the operator's moving distance was reduced. Consequently, the results show that the suggested layout in this research has reduced the moving distance of the operator by about 19% on situations of departure and arrival. Also, with the consideration of emergency situations, an optimized suggestion for a layout was derived with emergency- weight applied. It was suggested that ECDIS, Radar and Conning Displays should be located nearest to the officer. Also, an optimized design for panels of control and display devices on the Bridge was suggested in this research.Also, the research proved through examples that the ergonomic layout, which was suggested by predicting the officer's traffic line upon a ship's departure or arrival, is more efficient than existing layouts.As a plan to relieve a ship navigator's burden, this study attempted to suggest an ergonomic layout for on-bridge navigational equipment that plays an essential role for ship operations, and it is expected that the results of this study will be used as basic data in providing ergonomic guidelines in the shipbuilding and maritime fields.ACKNOWLEDGEMENTThe contents of this paper are the results of the research project of Ministry of Land, Transport and Maritime Affairs of Korea(Development of prevention and management technology for human-related marine accident).References[1]ABS (2000), "Bridge Design and NavigationalEquipment/System", ABS.[2]Hwang,B.H. and Lee,J.I. (1999), "A Study on theConcept of Human Factors in Ship Operating System", Journal of the Korean Institute of Navigation, pp. 29-42 [3] Ha,W.J., Na,S.J., Kim,S.S., Lee,H.K. and Jong,J.Y. (2002),"A Study on the Conceptual Bridge Design based on the Ergonomic Background", Journal of the Korean Society of Marine Environment & Safety, pp. 53-59.[4] Ha,W.J. (2003), "A Study on the Design andArrangement of Coastal Ship's Bridge on the Bridge on the Basis of Ergonomics Concept", Graduate School of Korea Maritime University.[5] ICS (2007), "Bridge procedures guide", fourth edition,ICS.[6] Kim,H.T., Lim,J.M., Lee,D.K. and Park,J.H. (2001), "State of the Art of Human Factors for Ship Design and Navigation Safety", Journal of the Erogonomics Society of Korea, pp. 39-47.[7] Kemp P. (1994), "The Oxford Companion to Ships andthe Sea", Oxford University Press.[8] Lee,H.H., Kim,C.S. and Yang,W,J. (2008), "A Study onthe New Design of the Training Ship's Bridge", Journal of the Korean Society of Marine Environment & Safety, pp. 169-175.[9]Lee,D.S., Kim,D.J. and Cha,M.S. (1997), "OptimumJun Hyuk Jang, Hongtae Kim, Joung Hoon Sim, Hea Sik KimDesign of Ship's Bridge Layout for the AutomatedOcean Going Vessels", Journal of Korean Navigationand Port Research, pp. 1-9, (1997).[10] Oh,C.S.(1992), "Human Factors Aircraft Cockpit Designand Flying Qualities", Journal of the Ergonomics Societyof Korea, pp. 26-32.[11]Yang,C.S. (2004), "On the Contributing Factors ofMarine Casualties and Incidents : with a Focus onHuman Factors", pp. 99-111.[12] Yang,Y.H., Yang,C.S., Gong,I.Y. and Lee,B.Y. (2005),"The Basic Study On the Development of ErgonomicIntegrated Bridge Alarm Management System", Journalof the Korean Society of Marine Environment & Safety,pp. 17-22.Received7 February 2011Revised17 October 2011Accepted 5 December 2011。
On Calculating the Value of Improved Forecast AccuracyJ. Scott ArmstrongThe Wharton School, University of PennsylvaniaPhiladelphia, Pa. 19104Reprinted from Armstrong, J. Scott, Long-Range Forecasting. New York, John Wiley 1985.How valuable is it to develop a more accurate forecasting model? Numerous researchers have tried to answer this question. The value of improved accuracy was studied by Dunnette (1966) for personnel selection, (Dawes 1971) for the selection of graduate students, Lave (1963) for weather forecasts for an agricultural product, Buffa (1975) for inventory control, and Schnee (1977) and Marks (1980) for weather forecasting. This research supports the commonsense notion that the value of improved accuracy depends upon the situation.The evaluation of forecast accuracy for a firm is of general interest. It also represents one of the easiest situations to evaluate relatively speaking, because costs and benefits can be translated into dollars. Nelson (1973, pp. 11-14) and Daub (1974) examined the value of improved forecasting to a firm. An alternative approach to such evaluation is presented here. It is based upon two key questions:1.What decisions are affected by the forecast? (If no decisions are affected, improvedaccuracy will have no value.)2.What is the current level of accuracy? (To what extent is it possible to improve?)Thesequestions should be considered in evaluating accuracy in any situation. To evaluate theaccuracy of a firm’s forecasts, these questions call for an examination of the seven factorslisted in Exhibit 1.Exh ibit1: Fac t or s f or Ev al uat ionof I m pr ov ed Ac c ur acyA.Market sizeB.Costsmitment of costs (when?)D.Error reductionE.Loss function (how much?)F.Cost of capitalG.Tax rateUsing the framework provided below, I attempted to use reasonable estimates in order to obtain a rule of thumb on the value of accuracy to the firm. This rule suggested that large firms couldbudget no more than about one percent of their sales revenues for the forecasting effort. To judge from data in Wheelwright and Clarke’s (1976) survey, the actual forecasting budget ranges from about 2% in firms with $10 million sales, to 0.1% in firms with $1 billion sales.To estimate potential cost savings for your organization, rather than use this rule of thumb or the practices of other organizations, you should use your own estimates within the following framework.The primary purpose of assessing the value of improved accuracy for a firm is to establish a framework that can be applied in a specific situation. The appendix also provides a rough estimate of the value of improved accuracy in a firm’s sales forecast. If you do not like my assumptions, feel free to use your own.The value of improved accuracy is a function of a number of factors. These were the size of the firm, the costs incurred by the firm, the proportion of resources that the firm must commit now for a given year in the future, the extent to which accuracy can be improved, the dollar cost associated with forecast errors, the tax rate, and the time value of money. Subjective estimates are provided here for each of these items.A.Market Size. The firm in this example has sales of $1 million per year in the selectedproduct market.B.Costs. Costs are of interest because we are concerned about possible cost savings. Costswere estimated by selecting 100 firms at random from 25 industries in the Value Lineinvestment survey for 1977. Before-tax earnings were estimated to be 10%; thus costs are90%.mitment. The firm makes decisions now that affect costs in future years–for example,plant location, size of plant, equipment, ley personnel, long-term contracts with suppliersand distributors, bond issues, union contracts, and research and development. It wouldbe useful, but expensive, to examine each of these decision areas separately. Furthermore,these estimates would vary by industry and by firm.The proportion of the (eventual) costs to which the firm must commit itself is expected to drop continuously as the forecast horizon increases. For some items, no priorcommitment is necessary; for example, one can decide at the last minute to turn on theelectric lights. For other items, such as the location and size of a factory, priorcommitments may be made that cover 20 years or more.To determine a reasonable level of commitment for this typical firm, I have assumed that 25% of the total cost require no commitment. After that, commitment drops offsteadily until year 20, after which the commitments are negligible. The rate of decreaseis faster at first, but then slows. (This curve can also be viewed as representing theproportion of the costs incurred in year h as the result of decisions make in year h - t.)The resulting estimates are presented in column C of Exhibit 2. (Column letters in Exhibit2 correspond to the numbers given to the various factors examined in the text.)Exh ibit2Sav ings f r om I m pr ov ed For e c as t Ac c ur acy(f or h y pot h et ical f ir m w it h$1 m il l ion s al es)Forecast Year (h)CCommitmentDErrorReductionELossFunctionFCost of CapitalFactor @ 12%GPresent ValueDollar Savings*10.2200.0250.6440.8932,8472.174.038.568.7972,6943.137.050.510.7122,2344.107.062.460.6361,7395.083.074.420.5671,3136.065.086.387.5079927.050.099.367.4527428.038.113.353.4045499.029.126.343.36140710.022.140.335.32229811.017.153.328.28722112.013.167.322.25716213.010.180.317.22911814.008.193.313.2058915.007.206.310.1837416.006.219.309.1636017.005.232.307.1464718.004.244.305.1303519.003.257.303.1162420.002.270.301.10415Totals 1.00014,660* This column is the product of all columns here times $900,000 costs per year.D. Error Reduction. The extent to which the sales forecast error can be reduced depends upon the level of accuracy achieved by current forecasting practice. If current practice is extremely accurate, little gain is possible. Errors in very short-term forecasting result from inaccuracies in estimating current sales. These errors are expected to be about 6% on the average. The forecast error increases as the forecast horizon increases.The error curve starts at 6% for forecast horizon zero (i.e., for the average error in estimating current status). Many estimates were available for sales forecast errors for a one-year horizon. These include:Exh ibit3: Er r or s f or O ne-Year-Ah ead For e c as t sStudy MAPEStrong(1956)8.0Ricketts and Barrett (1973)7.3Dalrymple (1975) 6.9Baker and Tralins (1976)10.3Dalrymple (1985)9.9Schreuder and Klaassen (1984)7.3Average8.3It was difficult to find estimates or errors for longer horizons; Dalrymple (1985) mentioned a 14.2 MAPE for three-year horizons, and Mentzler and Cox (1984) reported errors of about 20% for “over two years.” An examination was made of Ascher (1978) to determine the size of errors as well as their growth over the forecast horizon. It was difficult to tell whether forecast errors increased faster or slower than the increase in the length of the horizon. These informal explorations impressed me with my ignorance. As a result, I thought it was best to depend on William of Occam and simplicity. My conclusion was to use the estimate of about six % for year zero, and eight % for year one; then to increase the error linearly by two % per year. This assumption is reflected in column 1 of Exhibit 4. Exhibit 4 is completed by simple arithmetic, yielding column D, which was then transferred to Exhibit 2.To what extent may these errors be reduced? I estimated that a 30% error reduction can be achieved in estimating current status.As the forecast horizon increases, more data become available and it is therefore possible to draw upon additional forecasting methods.Exh ibit4Es t im at es of Er r or Red uc t ionForecast Year (h)(1)CurrentForecastErrors (%)(2)PossibleReductionin Error (%)Percentage ErrorReduction *[(1) × (2)] ÷ 100Midpointof yearDErrorReduction**0630.0 1.81840.0 3.21 2.521044.0 4.42 3.831246.5 5.63 5.041448.3 6.84 6.251650.08.057.461851.69.368.672053.010.6710.082254.211.9811.392455.313.3912.6102656.314.61014.0112857.116.01115.3123057.917.41216.7133258.418.71318.0143458.820.01419.3153659.121.31520.6163859.322.51621.9174059.523.81723.2184259.725.11824.4194459.926.41925.7204660.027.62027.0* This column represents the reduction in the MAPE; for example, for year 2, the MAPE would be reduced by 3.2, from 8.0 to 4.8.** This column uses the midpoint for each year. These estimates were used in calculating the yearly savings.Thus the possibility for error reduction is expected to increase as the forecast horizon increases. My estimate of error reduction is presented in column of Exhibit 4. It was based on subjective assumptions of 40% for 1-year forecasts, 50% for five-year forecasts, and 60% for 20-year forecasts. A smooth curve was constructed through these points. These estimates appear conservative in light of the results in Armstrong and Grohman (1972). This translates into small improvements in the reduction of MAPEs for short-range forecasts (2.5% for one-year forecasts) and large improvements for long-range forecasts (27% for 20-year forecasts).E.Loss Function . Errors in the forecast will cause inefficiencies in the firm. The dollar value of these inefficiencies will be some percentage of the dollars committed because management can take corrective action. Unfortunately, such corrective actions seldom live up to the or promise because we pay great reverence to tradition. Sometimes these decisions,based upon poor forecasts, defy the law of sunk costs and lead to even poorer decisions to cover up. Hence the loss function here may be based on an optimistic assessment of our capability for taking corrective action.The loss function for the firm in this example is symmetrical. In other words, the cost of underestimating (which would cause wither an opportunity cost or inefficient production) is equal to the cost of overestimating the same percentage (leading to an excess investment in resources).Losses are high when the reaction time is short, for there is less time to find alternative uses for the excess resources or to make adjustments (such as scheduling overtime) if resources are short. For the day-to-day operations of the firm, little effective reaction is possible, and it was assumed that 70% of the dollars committed would be lost. As the time horizon lengthens, the loss function decreases. Errors in the five-year forecast will become apparent well before the year five arrives, so that changes can be made. Thus it was assumed that, for the one-year forecast, the reactions will be such that 60% will be lost; for the five-year forecast, 40% will be lost; and for the 20-year forecast, only 30% will be lost. A smooth curve was fitted to these points, and the value at the midpoint of each year was entered in column E of Exhibit 2.F.Cost of Capital. A cost of capital of 12% was used. This figure was selected on the basis of opinions of three Wharton faculty members who specialize in this area, and also on my own opinion, as I have been an expert witness on the cost of capital in antitrust cases. I also reviewed literature on the cost of capital. For example, the cost of capital was estimated to be 10.5% in a study by Stewart C. Myers, “What was A.T. and T.’s Cost of Capital in Early 1971?” (Sloan School of Management, M.I.T., 50 Memorial Drive, Cambridge, Mass.02139). The present worth of future revenues is discounted at 12% in column F of Exhibit2. For example, the value of a dollar received 20 years from now is only 10 cents.G.Present Value . Column G of Exhibit 2 calculates the before-taxes present value savings of improved accuracy, using assumptions A through F. This present value is equal toSavings =()()()()()()A B C D E F h h h h h h h =∑120where the letters correspond to the assumptions, and h designates the forecast horizon. For example, the savings for year 3 would be:Savings 3 = (1,000,000)(0.90)(0.137)(0.05)(0.510)(0.712) = $2,234The tax rate was not explicitly entered into my calculations, that is, I figured before-tax earnings. If, however, you adopt solely the viewpoint of the stockholders, you should compute after-tax earnings.The total present value savings for this hypothetical firm with sales of $1 million would be about $15, 000.The importance of the forecast horizon can be assessed from the last column of Exhibit 2. As expected, the greatest gains can be made in the first five years (74% of the total savings). About 1% of the savings are achieved for years 16 through 20. Gains beyond 20 years are zero in this example because I assumed no commitment beyond this point.The conclusion that there is little value to forecasts beyond 15 years bothers me. The analysis assumes that one is dealing with rational organizations, that management will learn from its poor forecasts, and that corrective action will be taken. These assumptions are naive. What seems to happen is that one error leads to another. Often we refuse to recognize the errors. Instead, we compound the felony by covering up the error, or by continuing to act as if there were no error. “For this such reverence is lent to well-established precedent.” We see around us mistakes in urban design, transportation, education, housing, and other fields that were made more than 20 years ago. Yet the suffering goes on. Sometimes it requires a sit-in, riot, or demonstration to get bureaucrats to consider corrective action. Conclusion: this analysis greatly underestimates the value of improved accuracy ion forecasting, especially in longer range forecasts.If firms were rational, we would expect them to recognize the potential savings from improved forecasts and to invest money to bring about such savings. Of course, firms do make substantial investments in forecasting. I analyzed data from Wheelwright and Clarke (1976), who had surveyed expenditures on forecasting by firms of various sizes. In rough terms, these expenditures were:Annual Sales of Company (× $1 million)Annual Budgetfor Forecasts($)Budget asPercentageof Sales1019,000 1.96032,0000.530056,0000.21,000137,0000.1Obviously, firms must spend something to achieve their current levels of accuracy. We asked how much might be saved from improvements in accuracy. It seems that firms could benefit from additional investments in forecasting methods. Expenditures of 1% of sales would not seem unreasonable.The above framework can be used to calculate potential savings for your organization. Those of you with a personal computer can test various assumptions on your spreadsheet.ReferencesArmstrong, J. Scott and Grohman, Michael C. (1972), “A Comparative Study of Methods for Long-Range Market Forecasting,” Management Science, 19, 211-221.Ascher, William (1978), Forecasting: An Appraisal for Policy-Makers and Planners. Baltimore, MD: Johns Hopkins University Press.Buffa, Frank P. (1975), “The Application of Dynamic Forecasting Model with Inventory Control Properties,” Decision Sciences, 6, 298-306.Dalrymple, Douglas J. (1985), “Sales forecasting practices on businesses: Results from a 1983 U.S. Survey,” working paper, Graduate School of Business, Indiana University, Bloomington, Indiana 47401.Dawes, Robyn M. (1971), “A Case Study of Graduate Admissions: Application of Three Principles of Human Decision Making,”American Psychologist, 26, 180-188.Dunnette, Marvin D. (1966), Personnel Selection and Placement, Belmont, Calif.: Wadsworth. Lave, Lester (1963), “The Value of Better Weather Information for the Raisin Industry,”Econometrica, 31, 151-164.Marks, Robert E., (1980), “The value of ‘almost’perfect weather information to the Australian tertiary sector,”Australian Journal of Management, vol. 5, pp. 67-85.Mentzer, John T. and Cox, James E., Jr. (1984), “Familiarity, application, and performance of sales forecasting techniques, “Journal of Forecasting, vol. 3, pp. 27-36.Nelson, Charles R. (1973) Applied Time Series Analysis for Managerial Forecasting. San Francisco: Holden-Day.Schnee, Jerome E. (1977), “Predicting the unpredictable: The impact of meterological satellites on weather forecasting,” Technological Forecasting and Social Change, vol. 10, pp. 299-307.Strong, Lydia (1956), “Sales Forecasting Comes of Age,” Management Review, 45, 687-701. Wheelwright, Steven C. and Clarke, Darral, J. (1976), “Corporate Forecasting Promise and Reality,”Harvard Business review, 76, November-December, 40 seq.。
Ergonomic evaluation of interior design of Shoka vehicle and proposingrecommendations for improvementAdel Mazloumi * , PhDDepartment of Occupational Health, School of PublicHealth, Tehran University of Medical Sciences, Tehran, IranMohammadreze Fallah, MScDepartment of Ergonomics, University of Social Welfare andRehabilitation Sciences, Tehran, IranHedayat Tavakoli, MScDepartment of Occupational Health, Polytechnic University of Chamran, School of Public Health, Tehran University of Medical Sciences, Tehran,IranObjectives: One of the applications of ergonomics disciplinary is designing driver workstation compatible to users’ characteristics. The aim of this study was evaluation of interior design of Shoka vehicle with respect to the accommodation for Iranian population and proposing suggestions for customizing design of the vehicle. Method: This study was a descriptive-analytical study conducted among thirty men from Iranian drivers population in 5, 50, 95 percentiles of the stature variable. Objective variables related to the occupant packaging and vehicle visual aspects including anthropometric variables, frontal, lateral, and side view and so on were investigated first. Then, subjective variables related to the driver mental workload and body comfort discomfort were studied using BMDMW and comfort questionnaires during 2-hour driving trial sessions.Results: Occupant packaging variables and hand-arm angle showed the least accommodation percent(%53).Seating angles showed low accommodation as well (%73). Among three percentile groups there were no significant differences between the mean values of mental Workload during two hours driving task. The mean value related to the comfort discomfort was 3.9 during driving sessions. Conclusion: Considering the findings in this study, it can be concluded that seating angles need correction and optimization. Taking mental workload results into account, it can be concluded that the interior design of the studied car had no influence on drivers’mental workload. From the aspect of comfort/discomfort, Shoka vehicle showed neutral state among drivers.Optimizing seating angles, decreasing vibration, correcting stiffness of seating pan are suggested for customization of the ergonomicsaspect of this vehicle.Keywords: Ergonomic evaluation, Vehicle, Shoka, Anthropometric, CompatibilityIntroductionSafety and comfort are of the most important criteria for both car manufacturing companies and for drivers as well as occupants from many points of views. Therefore considering the principles of ergonomics in automotive design will make remarkable benefits to many extents .Different researches have shown that fatigue played an important role in incidence of thirty percent of traffic accidents. One of the major causes of fatigue while driving is the car interior design. Features such as seats, steering wheel and pedals have shown a large impact on driver fatigue. The interior design may also cause human error .Inappropriate design of the vehicle may lead to some health problems among driverpopulation as well. The risk of lumbar discs deformation, pain in neck, back and shoulder tension,reduced blood circulation in the legs and buttocks are the instances of those problems .Moreover, interior design has impact on driver mental workload.Mental workload refers to engaging the mind while performing a task like driving task. Problems such as distraction,performance reduction and human error are somehow related to the drivers' mental workload.Several ergonomics indicators are used by car manufacturing companies for car design evaluation. These indicators are either qualitative (subjective) or quantitative indicators.Qualitative indicators are achieved with the use of questionnaires and special checklists among users population.Quantitative indicators such as anthropometric measures, access limits,forces and so on are also used to evaluate interior design and its compatibility with the user population in different percentiles.Several questionnaires, such as BMDMW and body comfort-discomfort questionnaire,have been published to evaluate the vehicle design. BMDMW questionnaire is a subjective measure based on driving behavioral patterns which evaluate drivers’mental workload and state.Body comfort discomfort questionnaire also evaluate whole body and comfort-discomfortrelated to body limbs.Quantitative indicators to assess ergonomics-related aspects of interior design are generally related to the design dimensions.Dimensions can indicate humanaccommodation;and many research studies are based on evaluating those dimensions in 5,50,95 percentiles.The greatest differences in anthropometric values that are more relevant to the driver accommodation are three variables:stature, sitting height and weight. But, some research studies have shown that stature is most important anthropometric variable involved in the automotive design .Occupant packaging can be used for ergonomic quantitative evaluation of car design when considering stature as an anthropometric variable.Occupant packaging is the interior design process of a vehicle to achieve a good level of accommodation, comfort and safety for passengers. The most relevant to the ergonomics variables when considering occupant packaging are: interior dimensions (SAE J1100), hand control reach (SAE J287), ellipse (SAE J941) and driver selected seat position (SAE J1517 . The optimal dimensions of the anthropometric data that utilized or ergonomic design are valuable; but it dose not comprehensively cover other aspects of vehicle design such as: ease of use, comfort, field of view and safety aspects . That is the reason why vehicles are evaluated after design process.Particularly, if the vehicle would be an imported car, its design features and compatibility with the users are of crucial importance. Car manufacturers that are not original designer and they just assemble car products need to do researches in the field of vehicle ergonomics evaluation. Methods and materialsTh-sectional case study to evaluate the Shoka vehicle. Itis a small truck car as a new model of the Nissan Z 24 which is manufactured in Zamyad Car Company in Iran.Main chassis as well asengine for this car are exactly for the Nis-san model Z 24. But, its body is completelya new design.There were two groups of variables to evaluate this car:1) Quantitative variables selected from occupant packaging and were related to SAE J1100 (interior dimensions),SAE J941(driver's eye location) SAE J1517 (accommodation) these variables and the comparison standard values are shown in Table 1. Tools used for measuring these variables included: goniometer, rulers in various sizes, tape and label as marker. Sitting pattern in the vehicle,definitions of the variablesand measuring methods and required adjustment of the vehicle during measurement activity was done according to the requirements of the SAE J1100 standard. Reference points of measurement were: seating reference point (SgRP), accelerator heel point (AHP) and Pedal Reference Point(PRP) that the other variables were deter-mined with respect to these points .Vision related quantitative variables included: nearest visible point on the ground from left sides and front that determined by a marker in meter Fig 1 shows the quantitative variables related to this study.2) Qualitative variables:Qualitative variables were investigated using BMDMW and body comfort-discomfort questionnaires during 2-hour driving sessions. TheBMDMW questionnaire, which evaluates driver mental workload, has been developed by Francesco Di Nasera in Rome University . Reliability and face validity of the BMDMW questionnaire was investigated and confirmed after translating into Persian language. BMDMW has six internal factors include: Disengagement, Vehicle Monitoring,Route Monitoring,Road Awareness,Control and Fatigue and has Likert scale that shows drivingevents from one" rarely" to five "high". Variables related to comfort-discomfort were collected during the two-hour road trial sessions by body comfort-discomfort questionnaire. This is one of the common tools for assessing comfort-discomfort that measures comfort-discomfort state by averaging twelve members state of the body during the driving.Validity and reliability of this questionnaire has been done in previous researches (4, 5). This questionnaire has seven point Likertscale that shows comfort-discomfort state from one “very comfortable” to seven veryuncomfortable” .Data for the occupant packaging and vision was collected first and then,after a two-hour road trial test the questionnaires were administered by selected drivers as participants. Statistical software of SPSS version 15 was conducted for data analysis. Data analysis included two parts: examine the aspects of the mean and standard deviation of the variables related to the descriptive statistics. Then from aspect of analytical statistics data normality was assessed by Kolmogorov-Smirnov test and normal data was used for the ANOV A test to examine mean differences.ConclusionConsidering the findings of this study it can be concluded that some changes in interior dimensions specially in sitting angles required in this vehicle. These changes should cover the entire percentile groups. The angles of trunk and hand-arm should beincreased and the angles related to the knee and foot should be decreased.This car has neutral conditions for Iranian drivers.Vehicle interior dimensions have no remarkable effects on whole body comfort-discomfort and other factors like: softness and rigidity of the seat, vibration of the vehicle and contact pressure will be effective on the d river’s comfort-discomfort.Tall drivers have more control over the vehicle.Fatigue, vehicle monitoring, route monitoring are same be-tween the drivers with different statures. Changes in H-Point, front hood bulge and wipers would be effective ways for improving driver view.人机工程学评价舒车辆内饰设计,并提出改善建议阿德尔Mazloumi *,博士职业卫生部门,公立学校健康,德黑兰大学医学科学院,伊朗德黑兰mohammadreze法拉赫,硕士人类工效学系,大学和社会福利康复治疗科学系,德黑兰,伊朗hedayat Tavakoli,硕士职业卫生署,理工大学chamran,德黑兰,伊朗德黑兰大学医学科学院,公共卫生学院,目标:人体工程学的应用学科之一,是兼容的驱动程序的工作站用户的特点设计。