A suite of multivariate test cases for anthropometry-based design

Transcription

1 A suite of multivariate test cases for anthropometry-based design G. NADADUR and M.B. PARKINSON May 8, 2013 Abstract This work presents a test suite of three design problems involving spatial fit-based interactions between users and the product. The test suite allows for multivariate analyses of fit-based accommodation using anthropometric data. This process of calculating and comparing accommodation percentages is proposed as a method for a three main purposes: assessing the relative merits of different anthropometry synthesis techniques, verifying the accuracy of synthesized datasets of body dimensions, and comparing datasets of multiple populations of interest. The test suite is demonstrated in the context of an actual and synthesized target user population which are compared for an assessment of the multidimensional fidelity of an anthropometry synthesis method. Keywords: Anthropometry synthesis; Anthropometric dataset accuracy 1 Introduction The field of anthropometry-based design is replete with populations of interest, datasets of body measures, human modeling packages, and anthropometry synthesis methods. Existing techniques to verify or compare these data, packages, and methods include descriptive statistics (e.g., mean, standard deviation, by-percentile values), regression parameters (e.g., coefficients, R 2 values), etc. This paper aims to add to these techniques through a test suite of applied, multivariate design problems. The problems are selected to comprise of commonly-used and widely-available body dimensions that represent measures of length, breadth, and circumference. 1.1 Anthropometry in Design Relevant body measures are known to be influential in determining numerous aspects of physical interactions between users and products (HFES 300 Committee, 2004; Strasser, 1995). These aspects include safety, comfort, ease of use, visibility, etc.; when these criteria are satisfied for a user, that individual is said to be accommodated by the design. One of the goals when designing for human variability (DfHV) is to achieve a cer- 1

2 tain percentage of accommodated users within the target population of interest. Accommodated percentages of 90% or 95% are common objectives in DfHV efforts. Information and data about anthropometric variation within the target population are crucial to this end. Examples of anthropometry-based design are widespread. Certain products are demonstrative of indirect yet implicit influences of body dimensions on the product. Some instances of these are: heights of counters and work surfaces; thickness of pens; placement of switches to electric lights, fans, and power outlets. If work surfaces were positioned too high, pens were to be made too thick for the hand, and switches and power outlets placed in inaccessible locations on walls, then these products would not function as intended for many users. In contrast to the aforementioned examples, there are numerous efforts across academia and industry that consist of scientific studies of anthropometry-based design. These studies are focused on assessing anthropometric variation for a population of interest, understanding its impact on user accommodation afforded by different product specifications, and helping to make suitable design decisions. For example, a survey consisting of 44 body dimensions of 1174 Bahraini school children in the 6-12 years age range was utilized in a school furniture design effort (Mokdad and Al-Ansari, 2009). Three-dimensional body scan data have been analyzed for improved fit-based accommodation within each size of apparel (Loker et al., 2005). Chaffin et al. (Chaffin et al., 2000) measured and modeled arm reach motions for more than 3,000 individuals in a diverse population, and discussed their use in digital human modeling of, among other things, vehicle interiors and industrial workplaces. Nadadur et al. (Nadadur et al., 2009) utilized a synthesized population of North American manufacturing workers in improving an assembly process. Many of these studies result in models of users interactions with products. For instance, Guan et al. (Guan et al., 2012) proposes a suite of multivariate models for truck cab design. Flannagan et al. (Flannagan et al., 1998) and Reed et al. (Reed et al., 2002) are two examples of automotive driving posture prediction models. Garneau and Parkinson (Garneau and Parkinson, 2009) presents a case study involving the development of a bicycle seat height estimation model, while Nadadur and Parkinson (Nadadur and Parkinson, 2012b) formulates a model to predict door size choice. All of these studies consisted of behavioral experiments to collect data about sample users interactions with prototypes of the products at hand. These data were the basis of the anthropometry-based models of user fit, preference, choice, etc. 1.2 Test Suites in Design Test suites consist of a set of standard problems for verifying or comparing datasets, methods, etc. While test suites may be unfamiliar in the context of anthropometry-based design, there are a number of examples of case studies being utilized to demonstrate and/or verify new methods and data. Additionally, test suites are common practices in certain other fields of research, with optimization being a prime example. The remainder of this section describes the purposes served by test suites and provides a few examples of their application in existing literature. The first purpose of test suites is to assess methods. New and existing methods can be utilized in the context of the design problems to compare their relative strengths in terms of prop- 2

3 erties such as time taken, effort involved, complexity, accuracy, etc. Deb et al. (2005) discusses a range of methods for developing test problem suites, and presents a suite of 9 problems for testing optimization algorithms. Baluja (1994) presents a new optimization tool that is a combination of the traditional genetic algorithm and simple competitive learning. The improved capabilities of this tool are compared with a standard genetic algorithm using 12 test problems that include a traditional genetic algorithm test suite. Storn and Price (1997) puts forth and appraises a new heuristic global optimization through a suite of 3 testbeds that demonstrate its faster convergence with greater certainty, improved robustness and usability, etc. The second purpose of test suites is to verify datasets. In a number of instances, required data are unavailable for a given design scenario, and must therefore be estimated or synthesized through an appropriate method. It is desirable but not always possible to statistically validate these estimated datasets. A more realistic goal is to satisfactorily verify the datasets in the context of their application. A suite of design problems can serve this purpose. In the context of anthropometry-based design, there are a number of statistical tools to compare and verify synthesized datasets. A widelyapplied process involves implementing the synthesis technique on a test population for which actual anthropometric data are available. These actual data allow for comparisons with the synthesized data through descriptive statistics (e.g., means, standard deviations, by-percentile values), average body proportions (Fromuth and Parkinson, 2008), and Mahalanobis distances (Mahalanobis, 1936). These methods are employed in verifying the underlying hypothesis and synthesized data from the method presented in Nadadur et al. (2013), for instance. Despite these different modes of verifying the accuracy of synthesized anthropometric datasets, there is no statistically robust means to thoroughly validate them. Accordingly, test suites of design problems can serve to supplement existing verification procedures and tools. Test suites of design problems can also be utilized for a third purpose: to compare datasets. Datasets containing large volumes of multidimensional information can be inputs for the design problems. The outputs of the problems (e.g., percentage accommodation, optimal design specifications) would then be indicators of underlying multivariate differences between the datasets. An example of work that applies this method to compare anthropometric datasets is Nadadur and Parkinson (2012a), which examines a diverse set of global anthropometric datasets for accommodation levels resulting in each population from different sustainable design decisions. This process helps to understand the implications of anthropometric variations on the development of appropriate design solutions. 1.3 Research Focus As described in the preceding subsection, test suites of design problems can serve three main functions: assessing methods, verifying datasets, and comparing datasets. These functions would be of value in the realm of anthropometry-based design. Accordingly, the focus of the work put forth in this paper was to propose a test suite of problems for multivariate anthropometry-based design. Section 2 introduces and describes the test suite, and Section 3 demonstrates the application of the test suite on a case study involving verification of a synthesized dataset. 3

4 2 Proposed Test Suite The proposed test suite was developed as a set of virtual fitting-based assessments of accommodation levels yielded by actual and synthesized datasets of a population. Each design problem in the suite entails the formulation of constraints involving selected product dimensions and their relevant body measures. Following this, an accommodation assessment based on anthropometric fit is performed on each individual in the actual and synthesized data. If an individual s body measures satisfy the preset constraints, then the individual is considered accommodated by the design. The percentage accommodation can thereby be calculated for the two datasets and with a range of product dimensions. If accommodation results are observed to be similar across the range of dimensions and for the entire test suite, then the actual and synthesized datasets could be verified as being statistically equivalent for design decision-making purposes. Three general guidelines are adhered to by the design problems in the test suite. First, the problems concern fit-based accommodation, since spatial fit is directly determined by relevant body dimensions. Preference, comfort, and other types of variability are explicitly excluded. Second, the body dimensions corresponding to design parameters are readily available in a variety of global surveys and datasets. This improves the availability and applicability of the suite. And third, the selected dimensions include measures of length, breadth, and circumference. The suite is intended to consider combinations of measures that are not strongly correlated with one another. The simultaneous assessment of measures of length, breadth, and circumference make the design problems a more comprehensive test of multivariate anthropometric accuracy. Figure 3: crutches. Shoulder-elbow length + Elbowwrist length Hand length The body dimensions involved in The resultant suite consists of three design problems cab geometry of work equipment, office chair, and crutches which involve a total of 11 body dimensions spanning measures of sitting, standing, and reach. These are recommended as being supplemented by descriptive statistical comparisons of stature-bmi data for the actual and synthesized populations. The first of the proposed design problems is focused on cab geometry of work equipment such as fork lifts and cranes (Figure 1). The second design problem entails office chairs (Figure 2), while the third involves crutches (Figure 3). Figures 1, 2, and 3 illustrate the anthropometry that are relevant to the test suite. Table 1 lists the product dimensions (P) that are incorporated into each design problem. The cab geometry problem consists of 4 dimensions: cab width, foot pedal length, control distance, and cab height. The office chair problem comprises of seat depth, arm rest height, seat width, and seat height, while the crutch problem involves grip circumference and supportgrip length. 4

5 Figure 1: The body dimensions involved in the cab geometry of work equipment (e.g., fork lift). Hip breadth (sitting) Elbow rest height (sitting) Buttock-popliteal length Popliteal height (sitting) Chair model: Figure 2: The body dimensions involved in office chairs. The chair in the figure is the Herman Miller Aeron. Each product dimension in the test suite is associated with relevant body measures (B). In the cab geometry problem, cab width is influenced by bideltoid breadth, which is a measure of 5

6 shoulder width. Control distance is determined by forward reach, which is a measure of the distance from the back of an operator to their grip location on their outstretched hand. Foot pedal length and cab height are a function of the operator s foot length and sitting height. In office chair design, the relevant body measures are buttockpopliteal length, elbow rest height (sitting), hip breadth (sitting), and popliteal height (sitting). For the crutch problem, grip circumference is considered to be a function of hand length, while the support-grip length is determined by the sum of elbow-wrist length and should-elbow length. Table 1 also contains suggested virtual fitting constraints between each product dimension and Table 1: The product dimensions involved in each of the three proposed design problems, along with their relevant body dimensions. Also shown are suggested constraints for use in the virtual fitting process of the test suite. Product dimension Body measure Suggested design (P) (B) constraints Cab geometry Cab width Bideltoid breadth B P; P= mm Foot pedal length Foot length B P; P= mm Control distance Forward reach B P; P= mm Cab height Sitting height B P; P= mm Office chair Seat depth Butt-popliteal length B P; P= mm Arm rest height Elbow rest height (sit) B P; P= mm Seat width Hip breadth (sit) B P; P= mm Seat height Popliteal height (sit) B P; P= m Crutch Grip circumference Hand length B P; P= mm Support-grip length Elbow-wrist length + Shoulder-elbow length B P; P= mm its relevant anthropometry. The and constraints show the necessary condition for the anthropometry to be accommodated on the basis of fit. For instance, in the office chair problem, one of the conditions for the product to fit the individual is: buttock-popliteal length seat depth; this constraint is also illustrated in Figure 2. Accordingly, for the individual to be considered accommodated in the three design problems, that individual s anthropometry must satisfy all the specified constraints. In addition to specifying the aforementioned constraints, Table 1 recommends ranges of values for the product specifications. These values can be used in defining the constraints. For example, the individual could be considered accommodated on seat depth if their buttockpopliteal length is less than or equal to any value in the range of 400 to 500mm. The product specification for each design problem can be defined by selecting values from the recommended range for each product dimension-body measure constraint. For each of the possible product specifications, the percentage accommodation can be calculated through quantitative virtual fitting for the actual and synthesized datasets. The variation of percentage accommodation for the two datasets can thereby be compared across the range of product specifications. The following section demonstrates the test suite by applying it in the context of an pair of actual and synthesized datasets. 3 Case Study The anthropometric database chosen for the purposes of this demonstrative case study consists mainly of civilian Japanese youth from the early 1990s (National Institute of Bioscience 6

7 and Human-Technology, 1996). The database is comprises of a comprehensive list of 255 body dimensions. Female subjects in the year age range were retained for the purpose of this case study, resulting in a total of 200 individuals in the dataset. This dataset is henceforth referred to as the actual dataset. For this population of young Japanese women, descriptive statistics were calculated for the 11 body dimensions that are part of the test suite. These statistics were the inputs to the percentile combinations dataset (PCD) method described in Nadadur et al. (Nadadur et al., 2013), which was the anthropometry synthesis technique selected for this case study. The percentiles of the 11 body measures for each individual in a reference database (Gordon et al., 1989) were compiled into a PCD. The aforementioned descriptive statistics for the Japanese population were used to generate a Gaussian curve-based anthropometry model for each body measure. These models were then sampled for the percentile values contained in the PCD, thereby producing anthropometry values for the synthesized Japanese population, termed the synthesized dataset. The proposed test suite was now utilized in comparing the actual and synthesized datasets. Quantitative virtual fitting was applied to calculate percentage accommodation for the two datasets across a range of product specifications. Table 2 contains a sample of the results of this process from the cab geometry design problem. The values of percentage accommodation calculated for the actual and synthesized female Japanese populations are shown to be equivalent, and this is true for a range of cab geometry specifications. Nadadur et al. (Nadadur et al., 2013) had utilized a variety of modes of comparison descriptive statistics, Mahalanobis distances, Wilcoxon tests to demonstrate the statistical equivalence of the actual and synthesized datasets. The equivalence of the data in terms of the percentage accommodation values listed in Table 2 is a final, multivariate mode of verification of the accuracy of the PCD method of anthropometry synthesis. This case study is a basic demonstration of one of the three proposed functionalities assessing methods, verifying datasets, and comparing datasets of the test suite. Future work will employ these design problems in broader contexts of anthropometry-based design. Accordingly, the cab geometry, office chair, and crutch design problems that constitute the test Table 2: A sample of the results of virtual fittingbased comparisons of the actual and synthesized ( Synth ) Japanese women. Percentage accommodation is shown for a range of product specifications for the cab geometry design problem in the proposed test suite. All dimensions are in mm. Cab Foot pedal Control Cab % accommodation width length distance height Actual Synth

8 suite are proposed as a useful addition to the existing digital human modeling toolkit. 4 Acknowledgements This research was partially funded by the National Science Foundation under Award No Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation. References Baluja, S. (1994). Population-based incremental learning: a method for integrating genetic search based function optimization and competitive learning. Technical Report CMU- CS , Computer Science Department, Carnegie Mellon University, Pittsburgh, PA, USA. Chaffin, D., Faraway, J., Zhang, X., and Woolley, C. (2000). Stature, age, and gender effects on reach motion postures. Human Factors, 42(3): Deb, K., Thiele, L., Laumanns, M., and Zitzler, E. (2005). Scalable test problems for evolutionary multiobjective optimization. Springer London. Flannagan, C. A. C., Manary, M. A., Schneider, L. W., and Reed, M. P. (1998). An improved seating accommodation model with application to different user populations. SAE International. Fromuth, R. and Parkinson, M. (2008). Predicting 5th and 95th percentile anthropometric segment lengths from population stature. In Proceedings of the ASME International Design Engineering Technical Conferences, number DETC , New York, NY. Garneau, C. and Parkinson, M. (2009). Including preference in anthropometry-driven models for design. ASME Journal of Mechanical Design, 131(10):6. Gordon, C. C., Churchill, T., Clauser, C. E., Bradtmiller, B., McConville, J. T., Tebbetts, I., and Walker, R. A. (1989) anthropometric survey of U.S. Army personnel: Methods and summary statistics, Final report. Technical Report NATICK/TR-89/027, U.S. Army Natick Research, Development and Engineering Center, Natick, MA. Guan, J., Hsiao, H., Bradtmiller, B., Kau, T.-Y., Reed, M. P., Jahns, S. K., Loczi, J., Hardee, H. L., and Piamonte, D. P. T. (2012). U.S. truck driver anthropometric study and multivariate anthropometric models for cab designs. Human Factors: The Journal of the Human Factors and Ergonomics Society, 54(5): HFES 300 Committee (2004). Guidelines for Using Anthropometric Data in Product Design. Human Factors and Ergonomics Society, Santa Monica, CA. Loker, S., Ashdown, S., and Schoenfelder, K. (2005). Size-specific analysis of body scan data to improve apparel fit. J. Textile and Apparel Technology Management, 4(3): Mahalanobis, P. (1936). On the generalised dis- 8

9 tance in statistics. Proceedings of the National Institute of Science, India, 2(1): Mokdad, M. and Al-Ansari, M. (2009). Anthropometrics for the design of bahraini school furniture. International Journal of Industrial Ergonomics, 39(5): Nadadur, G., Chiang, J., Parkinson, M., and Stephens, A. (2009). Anthropometry for a north american manufacturing population. In Proceedings of the SAE Digital Human Modeling for Design and Engineering Conference, number , Gothenburg, Sweden. Storn, R. and Price, K. (1997). Differential evolution - a simple and efficient heuristic for global optimization over continuous spaces. Journal of Global Optimization, 11(4): Strasser, H. (1995). Ergonomic efforts aiming at compatibility in work design for realizing preventive occupational health and safety. International Journal of Industrial Ergonomics, 16( ). Nadadur, G. and Parkinson, M. B. (2012a). The role of anthropometry in designing for sustainability. Ergonomics, Special Issue: Ergonomics & Sustainability, in review. Nadadur, G. and Parkinson, M. B. (2012b). A utility theory-based approach to reducing raw material usage in non-adjustable artifacts, tasks, and environments. Work: A Journal of Prevention, Assessment, and Rehabilitation, 41(1): Nadadur, G., Raschke, U., and Parkinson, M. B. (2013). An anthropometry synthesis technique for global user populations. International Journal of Industrial Ergonomics, in review. National Institute of Bioscience and Human- Technology (1996). AIST anthropometric database , Human body dimensions data for ergonomic design, Nippon Shuppan Service, Tokyo. Reed, M. P., Manary, M. A., Flannagan, C. A. C., and Schneider, L. W. (2002). A statistical method for predicting automobile driving posture. Human Factors, 44(4):