Robust parameters determination for ergonomical product design via computer musculoskeletal modeling and multi-objective optimization

Highlights

• Run computer software AnyBody with RSM to find muscle activity functions.

• Muscle activity functions are normalized to form multi-objective problem.

• Determine optimal parameter values via multi-objective optimization.

• Product design with determined optimal parameter values must hold fewer injuries.

Abstract

This research aims to determine the optimum parameter values for ergonomic product designs via computer musculoskeletal modeling (CMM) and multi-objective optimization (MOO). The multiple-muscle activities measured by the AnyBody (AB) Modeling System are used to develop the functional relationships with product design parameters via a statistical method, such as design of experiment (DOE). One DOE, Response Surface Methodology (RSM), is adopted in the present approach. Such functional relationships are considered as objective functions which will be further formulated via compromise programming (CP) for multi-objective optimization MOO) problems. A bike-frame design problem is chosen herein as an example to demonstrate the proposed method; it includes determining the lengths of the stem, head tube, fork, top tube, seat post, seat tube and pedal crank. Two cases related to the proposed approach are also introduced: one is a deterministic product design under certain conditions, and the other is a robust product design under uncertain conditions. Because the combination of computer musculoskeletal modeling, statistical method, and multi-objective optimization technique is realized in the proposed approach, ergonomic product designs for safety and efficiency under uncertainty can be achieved in the early stage of other product designs.

Introduction

It is well known that a poor ergonomic product design can lead to discomfort and even health disorders (e.g., back pain, neck and shoulder complaints, etc.) entailing major cost to society through missed work/reduced productivity, physiotherapy, and diminished work-effectiveness (Luttman et al., 2003, Major and Vezina, 2015, Niu, 2010). Thus, the ever-increasing demand by customers for products with improved performance and quality is complemented by an equally strong demand for products with improved comfort and health. These demands provide a possible way to differentiate products from those of competitors. Consequently, product design is forced to place additional emphasis on ergonomics for safety and health reasons in order to gain a competitive edge in the market (Azadeh & Sheikhalishahi, 2015).

Over the past decades, Computer-Aided Engineering (CAE) has revolutionized product development (Jeang, 2008, Jeang et al., 2008, Jeang, 2011, Pandian et al., 2013). The current available analysis facilities cover almost every conceivable technical property, and enable technical products to be developed via computer simulation before actual production. As CAE plays the main role, it has clearly become the requisite tool for modern industries. In addition to its contribution to the conventional applications in the manufacturing sector, CAE is being applied more rigorously to address the issue of product design for safety and health considerations. That is, the use of computer models for effectively integrating humans and products through their analyses highlights the process of product development; meanwhile, attention is also being paid to the ergonomic methods of product development (Damsgaard et al., 2006, Ma et al., 2009, Gragg et al., 2013). Ideally, in the early stages of product design, as aforementioned, a new product can be examined for its degrees of comfort and health by carrying out computer simulations of human interactions with the introduced product. To enhance a product’s competitiveness, computer simulation of the mechanics of the human body (the musculoskeletal system) has become a subject of importance in the essential research on the many relevant possible applications.

However, before these computer simulations become efficacious comfort and health assessment tools, the important problems of identifying comfort and health measures and establishing their relations to product design variables have to be solved. In taking product design bioinstrumentation measurements, the electromyogram (EMG), electroencephalogram (EEG), and electrocardiogram (ECG) are often employed to evaluate products ergonomically and quantitatively (Chaffin et al., 1999, Tsang and Vidulich, 2006, Petrone et al., 2015). Surface EMG is frequently used for evaluating physical stress and fatigue because of its noninvasive nature and ease of measurement (U.S. Department of Health & Human Services, 1992).

Currently, there are numerous commercialized human simulation tools available for job design and posture analysis, such as 3DSSPP (Chaffin et al., 1997), Jack (Badler, Phillips, & Webber, 1993), VSR (Yang, Marler, Kim, & Arora, 2004), AnyBody (Damsgaard et al., 2006, The AnyBody, xxxx). From the viewpoint of analysis, particularly in support of dynamic full body modeling, which involves quite a few muscles, the AnyBody Modeling System (AB) can be built for handling models that are substantially more intricate than other CAE technologies could handle. Additionally, an appropriate human simulation tool which can grasp the interaction between design parameters and the human body is essential to successfully perform ergonomic product designs. Until now, there has been no quantitative method like AB that is capable of effectively and efficiently analyzing the ergonomic quality of products. Furthermore, the continuous quantitative measurement score generated from AB makes the design of experiment (DOE) statistical method easily implementable. When the DOE method is implemented to provide the objective functions, optimization techniques such as MOO become usable. The above facts comprise the optimal solutions to be precisely and effectively obtained in the presented research. Thus, the AB Modeling System was adopted in this research. With the aid of AB, the body is perceived as a real multi-body mechanical system connected by joints and actuators in the form of a mechanism.

The usage of certain products involves complex and rhythmic activities, requiring the well-integrated efforts of brain, nerves, skeletal structure, and muscles (Cesar, Javier, & Ricardo, 2014). The coordinated action of various muscles is required to generate the synchronized forces needed to control their performance, so that the result is harmonious and efficient. The inherent assumption about rational coordination is that the CNS will try to smoothly and efficiently balance all the involved muscles in an optimal manner in order to minimize the total muscle activity, so that fatigue and injury can be reduced as much as possible. Because of the rationality assumption, the MOO technique will be introduced to formulate the multi-muscle activities for optimizing the product parameters design in order to achieve optimal health and efficiency. Because MOO is generally formulated with a set of objective functions for the study of related problems, a way to provide these multiple objective functions in advance becomes critically important in the presented approach.

A set of objective functions is unknown most of the time, particularly regarding the functions describing the dynamic interaction between the product design variables and the human musculoskeletal system under uncertain environments. One possible approach is to run physical experiments in building these functions. As is known, physical experiments are costly, inefficient, difficult, and impossible in some cases, so replacing physical experiments with computer simulation, as in the aforementioned discussion, is one of the best methods available for obtaining experimental data in most current engineering applications. These experimental data are then further analyzed by a statistical method to obtain the response functions used for formulating MOO. Thus, an integration of computer simulation and a suitable statistical method becomes crucial for achieving successful applications in the proposed approach (Jeang, 2008, Jeang, 2008, Jeang et al., 2008, Jeang, 2008, Jeang et al., 2008, Jeang, 2011). With physical stress and fatigue being important elements of concern in this study, the muscle activity measurements generated from the AB computer software are considered as the computer’s experimental data in the proposed approach. These experimental data will then be analyzed by a DOE statistical method, such as response surface methodology (RSM), to find the desired response functions for MOO (Myers, 1991, Myers et al., 2009).

In order to improve the ergonomics of product design and refine workplace tasks, the interaction between humans and product design has to be considered early and thoroughly in the lifecycle of products. Such interaction can be described by the known human body posture and the parameter values in product design. As aforementioned, posture can be measured by CAE, such as AB, which is capable of analyzing the musculoskeletal system of humans. Determining how to use the AB Modeling System to carry out an ergonomic product design efficiently becomes our concern. Without mentioning the disadvantages of adopting other CAE in the posture analysis, the conventional approach is performed by an exhaustive strategy, such as the trial and error method to create a set of alternative product designs. Subsequently, the related posture analyses through AB are carried out to evaluate these alternative designs correspondingly. Finally, the alternative design with the finest ergonomic evaluation score is selected from the creation alternatives, with the hope that the parameter values associated with the selected design candidate truly represent the optimal parameter decision for a product design; however, this strategy is very inaccurate, ineffective, and costly. In addition, product design is an evolutionary process requiring constant revision and modification, which usually entails a lot of sensitivity analyses over a period of time. Because of the aforementioned disadvantages, the existing methods prove very tedious, expensive, and time consuming in seeking the best candidates arbitrarily (and often in vain) to determine the truly optimal ergonomic product parameter values. In particular, the degrees of the above shortcomings become even worse for a product designed under an uncertain environment. For example, uncertainties arise from human body elements, such as sizes and weights; thus, the proposed approach adopts the DOE method, RSM, to efficiently create representative alternative designs and construct precise functional relationships between ergonomic product parameter values and ergonomic evaluation scores. With the aid of optimization techniques, these functional relationships can be promoted to formulate the design problem in relation to MOO to ensure that the determined product parameter values are the most representative and accurate optimal solutions. Owing to the reusable features of functional relationships, it turns out to be very convenient in re-conducting MOO for repeatedly revising product designs during the evolutionary design process. Based on the aforementioned discussions, the contributions of the proposed approach, which reflect the deficiencies in existing methods, are listed in the following six outlines.

• AB scores represent the equivalent EMG measurements which reflect the level of fatigue and the risk of injury for the combined consequences from the human body and the given product design.

• By applying CAE, AB software, the ergonomic design can be developed early, before the product is actualized.

• Due to the integration of the computer software AB system and statistical method RSM, ergonomic design functions (the functional relationships between the ergonomic evaluation scores and the product design parameters) can be constructed for design optimization. In addition, after the RSM analysis, the obtained ANOVA is capable of providing the suggestions for continuous product revision.

• MOO can be formulated from the determined ergonomic design functions. Because MOO is formulated via Mathematical Programming (MP), the derived ergonomic design functions can also be devised as constraints in design problem formulation for design sensitivity analysis. The design sensitivity analysis enables the exploration of the feasible design space, which provides critical information for product revision, particularly under uncertain environments or various applications. For example, by taking advantages of the MP formulation, a sensitivity analysis can be performed by varying the percentage upper levels for adapting to any particular working environment. Consequently, a customized product design for further safety and health improvement becomes possible.

• The obtained ergonomic design functions are reusable for further design improvement, and can also be integrated within other fields for advanced product design. The example here in employs ANSYS software for simultaneous ergonomic and mechanical designs (ANSYS Manual, 1997, Jeang et al., 2008).

• The random case introduced in the example demonstrates the possibility of product design under uncertain environments. For example, the uncertain human weight is plugged into AB software to generate evaluation scores for a robust product design under uncertainty.

• With the introduction of the proposed approach, musculoskeletal disorders (MSDs) can be avoided during the entire product lifecycle application. In addition to the cost reduction in product development, the cost savings to society are significant because of the reduction in physiological disorders.

This paper is divided as follows: Sections 2 Response surface method (RSM), 3 The problem in optimizing analytical functions or running CAE simulation under uncertainty, 4 Multiple objective optimization for multiple muscle activities contain relevant information regarding the presented approach. The information includes computer musculoskeletal modeling, optimization under uncertainty, computer experimental data examined through RSM for response functions, and normalized response functions for formulating the multi-objective optimization of multiple muscle activities. Section 5 introduces block diagrams for the proposed approach. Section 6 provides an application to demonstrate the proposed approach. A relevant discussion is offered in Section 7 for illuminating the proposed approach. Finally, the conclusion is given in Section 8.