Final Report Summary - DHERGO (Digital Humans for Ergonomic design of products)
Executive Summary:
Digital Mock-Ups (DMU) together with Digital Human Models (DHM), are more and more used in the early phase of product design in order to reduce the product development time and cost. The main objective of the DHErgo project is to develop anatomically and bio-mechanically sound human models for evaluating posture and motion related discomfort. The project started in September 2008 for 39 months and gathers 10 European partners. It is structured in eight work packages in charge of the technical work of the project and two work packages in charge of the project management and dissemination activities. Due to the complexity of modelling human body and data availability (especially for the detailed human musculoskeletal data), the project is structured in three overlapped cycles according to the body part to be modelled: 1) Lower limbs, 2) Upper limbs including the shoulder, 3) Whole body including the torso. Each cycle covers the whole core technical work, namely the data collecting / motion reconstruction - analysis - modelling - implementation - assessment sequence. A large amount of human data has been collected for developing realistic human models. New approaches for data collection were developed and highly valuable data were obtained for human functional data. For instance, data of joint strength and joint range of motion with consideration of adjacent joints were collected at TUM and IFSTTAR. In addition to detailed musculoskeletal data, ULB also investigated shoulder complex motion both in vitro and in vivo so as to find the mechanical relationships between the humerus position and orientation on one hand, and the related attitude of the scapula and clavicle at the same moment of time on the other hand. The algorithm combines corresponding fitting curves to evaluate motion prediction by quadric multiple regression. Accuracy of the reconstruction allows applying developed method in ergonomics and daily activity analysis. CEIT has proposed a dynamic motion simulation method which takes into account the constraints due to motion dynamics. The results show that the predicted motion resembles the real motions both in trajectories and in joint torque profiles. The tool has been validated against the database of reconstructed experimental motions. The reconstructed motions using a multi-body human model can now be exported into ESI’s musculoskeletal model including deformable soft tissues so as to estimate muscle forces as well as contact pressure. Three automotive application oriented case studies have been carried out at IFSTTAR in close collaboration with three car manufacturers as end-users in the project. These data are used for testing the motion and discomfort simulation methods. HS has implemented a new version of the DHErgo demonstrator which integrated the simulation tools on the existing human models RAMSIS and PAM Confort. BMW validated the demonstrator on two major aspects. The first one is the simulation and the deviation between simulated and recorded results, e.g. for motion trajectories or discomfort ratings. The second one is about the user interface and the handling of the demonstrator. Most of the requirements defined by the three end-users were achieved. ERT as project co-coordinator in charge of the dissemination activities and administrative management has organized the participation in key conferences. All the materials (newsletters, final leaflet, posters,…) have been widely distributed in the major DHM related events. All publications have been uploaded on the DHErgo website.
Project Context and Objectives:
Digital human simulation is believed to be one of the key technologies for proactive design. All components of a product can be digitalized using today’s CAD (Computer-Aided Design) technologies. Being able to evaluate the customer’s satisfaction through digital human simulation at a very early stage of product design becomes an absolute necessity for reducing design cycle. It can be easily understood that such an advanced digital human simulation tool can be applied in a large industrial domain where the designed system is addressed to a high number of users (clients/operators) and its physical mock-up is difficult to reproduce. Automotive industry is one of most active end-users of such a simulation tool because of high pressure for reducing time-to-market and cost. The final results of the project will mainly be demonstrated through case studies related to car design, thought they can be easily transferred to other industries.
In order to evaluate the physical aspects of a product or a workplace in its early design stage, a design engineer will insert a digital human representing one critical target user of the product in the CAD (Computer-Aided Design) environment at first, then simulate critical postures or movements and evaluate how the target user will judge it in terms of:
• Visibility. Can the target user see all necessary visual information?
• Clearance and accessibility. Does the product have the sufficient room so that the target user can carry out all necessary movements freely without collision?
• Ease of use. Can the target user reach and manipulate the controls easily within his/her physical capacity (strength, joint maximum range of motion, equilibrium keeping)?
Ideally, the design engineer would also like to be guided by the recommendations of the DHM simulation software for modifying the design and finding a compromise without degrading the ergonomic demand.
Since a product like an automobile can be used by thousands even millions of people, the design engineer has to consider the variability of user population. People vary not only in body dimension and shape (anthropometry), but also in physical capacity (body strength and joint mobility). Besides, the human body has in general more degrees of freedom than imposed by a task, implying that people can adopt different ways to execute the same task. The variability in postural and motion control has also to be considered in simulation. Often the design engineer has to answer the question what percentage of the whole target population (e.g. German car drivers) is accommodated for a specific task under consideration, like car ingress and egress for instance. Among the population of the drivers, there are more and more elderly people especially in developed countries. One has to take into account their specific needs due to aging.
However, currently available digital human simulation tools are still far from the designer’s expectations (see Chaffin, 2005 for an excellent review), especially in terms of motion simulation and evaluation of motion related discomfort. There are three main commercialized DHM packages used for ergonomic design of products:
• Jack, a software initially developed by University of Pennsylvania and now commercialized by UGS, http://www.ugs.com/products/tecnomatix/human_performance/jack/
• Safework, initially developed by a Canadian ergonomic consulting company, now purchased by Dassault Système and renamed as Human Builder in CATIA, http://www.3ds.com/
• Ramsis, developed and commercialized by Human Solutions with the support of German car makers, http://www.human-solutions.com
For these three human modelling packages, the human model behind is mainly a geometric and kinematic representation of the human body. Clearly, this is a first step towards a comprehensive ergonomic simulation tool for design engineers. Interdisciplinary research efforts have to be continued not only for realistic motion simulation but also for understanding and predicting posture and motion related discomfort. Relating the perceived discomfort with design parameters through human/product interaction simulation is a fundamental requirement for ergonomic applications.
The digital human models used for ergonomic design of products have to evolve towards dynamic and muscular ones. They must be able not only to have a realistic visual representation of human body and movement but also to evaluate the muscular efforts associated with a task for a better understanding of human performance and perceived discomfort. DHErgo mainly focused on the following scientific issues:
• Development of multi-body dynamic motion reconstruction methods in order to estimate joint motion and joint forces
• Integration of existing human anatomic data and development of detailed human musculoskeletal digital models
• Collecting necessary data for model developing and validation. Collecting of the basic human functional data such as joint strength and joint limit will be continued.
• Development of a hybrid optimisation/data based complex motion simulation approach. Optimisation criteria based on joint force and muscle performance will be explored.
• Development of a generic motion-based biomechanical discomfort criterion. Thanks to the musculoskeletal models to be developed, different biomechanical parameters including muscle forces and other soft tissues loads will be investigated.
• Elderly people. Basic biomechanical data of the elderly people will be collected. The effects of aging on movement and perceived discomfort will be studied.
Project Results:
DHErgo has fulfilled the most of the objectives defined at the beginning of the project. Here are the main achievements.
1-Human functional data for a more realistic representation of human physical capacity
To achieve a more realistic representation of human physical capacity extensive experimental data are obtained during the project. Both maximum joint range of motion (RoM) as well as maximum joint torque data are measured for all major joints of the human body (figure 1). To account for expected differences in age and gender 18 males and females are studied forming two age groups each: younger than 35 years and older than 65 years.
An important and new aspect in these measurements is the consideration of dependencies between different joint degrees of freedom of one joint as well as adjacent joints due to bi-articular muscles, which has not been done to this extent before. These relations are studied using specifically designed measurement devices and motions are recorded using a Vicon motion capture system. On the basis of anatomical considerations and experimental feasibility thoroughly planned experimental protocols are defined allowing for a very detailed analysis while limiting the experimental effort for the subjects to an acceptable level. The measurements are distributed on 20 sessions per subject lasting about two hours each.
The resulting data are analyzed using inverse kinematics and dynamics and statistical methods. One exemplary result for joint RoM is the relation of maximum hip flexion and knee flexion RoM which could be quantified and modeled. Considering maximum joint torque numerous relations were found. It could for instance be shown that elbow flexion torque is highly dependent on the elbow flexion angle as well as the shoulder flexion angle. Furthermore the expected decrease of joint torque with increasing age and the differences in gender could be confirmed and partially quantified.
The resulting mathematical models of joint torque – joint angle relations as well as inter-joint relations can now directly be applied to digital human models allowing for a more realistic simulation of physical capacity. F. Engstler & F. Guenzkofer [TUM]
2- Detailed human anatomic data collection
Musculoskeletal data collected from ULB past projects (e.g. VAKHUM and LHDL) were extended during DHErgo project by accurate full body soft tissue reconstruction, kinematical data collection for lower limb (e.g. daily motion and pedal clutching) and upper body (e.g. shoulder complex rhythm evaluation). These data were also reconstructed by ULB to obtain anatomically correct joint motion behavior for muscle force evaluation, taking into account electromyography of the selected muscles.
Bone and soft tissue (muscles, ligaments) 3D morphology reconstruction, including extraction of muscle and tendon fibre topology and attachments, contains 3D reconstruction of dissected soft tissues as alternative to high resolution MRI that shows some limitations when whole specimen body has to be scanned. ULB proposed a method using stereophotogrammetry on the basis of digital cameras allowing soft tissue data collection and maintaining anatomically-correct results. Prior to dissection, a specimen has been fully CT-scanned to obtain bone 3D models. During specimen dissection, pins with colored heads (CHs) were inserted in muscles and ligaments to characterize muscle and ligament fibre path, musculo-tendinous junctions, origins and insertions (Figure 2). CHs data registration to the 3D bone models occurred using technical frames including reflective markers and aluminium balls inserted into the bones for better accuracy. Once registered, CHs information was reconstructed in 3D together with the bone models (Figure 3). Muscle and tendon fiber length and pennation angles were evaluated after piece-wise linear approximation of the reconstructed points. The method has been applied on an entire human body. By V. Sholukha and S. Van Sint Jan [ULB]
3- Evaluation of the shoulder rhythm mechanism by quadric multiple regression
The human shoulder complex allows an important spatial displacement of the underlying bone segments thanks to the sequence of three intermediate joints connecting the humerus to the thorax. Due to anatomic constraints, their motions are coupled. The contribution of the scapular displacements, linked to the clavicle behavior, within the overall arm elevation follows a general pattern in which scapular motion is responsible for approximately one third of the total arm elevation. It is therefore to respect natural shoulder motion pattern (shoulder rhythm) when reconstructing and simulating human arm motion.
The main challenge of the underlying research is to find the mechanical relationships between the humerus position and orientation on the one hand, and the related attitude of the scapula and clavicle at the same moment of time at the other hand. This answers a practical problem of in-vivo motion analysis. Indeed, if data related to the humerus instantaneous spatial position is relatively easy to obtain, the same information is more difficult to collect for the clavicle and the scapula. Furthermore, a unique and straightforward mechanism like in the knee or the ankle joint is not possible, e.g. the humeral head shows similar displacement during shoulder elevation or scapula-humeral abduction due to the fact the vertical displacement of the humerus is closely linked to clavicle elevation. An algorithm that aims to estimate the clavicle and scapula pose compared to the humerus instantaneous posture, must first deal with the differentiation between a shoulder elevation due to either a humerus rotation or a translation, to only then combine correspondent fitting curves to evaluate motion prediction by quadric multiple regression.
In a study by ULB, data from three volunteers and two specimens were processed for a total of 51 different motions including abduction-adduction (frontal plane), flexion-extension (sagittal plane) and humeral head vertical ascent-descent. Typical example of the rhythm components fitting by linear and parabolic polynomial functions are presented in Figure 4 for three repetitions of flexion/extension. Figure 5 presents results of motion reconstruction and application of prediction-correction shoulder rhythm. Analysis of 15 different motions shows a mean accuracy for the clavicle and scapula rhythm reconstruction of 8.5 (SD=6.1) mm and 19.8 (SD=5.2) mm, respectively. Accuracy of the reconstruction allows applying developed method in ergonomics and daily activity analysis.
4-Inverse kinematic and inverse dynamic motion reconstruction methods
One of the major results produced by CEIT, as expert in dynamic multi-body modeling, is a method for the whole body motion reconstruction. This method has been implemented in a software tool that is used for reconstructing the motions recorded during the project experiments. A novel feature of the proposed method is that a generic human model is automatically scaled to the subject’s actual dimensions, based on palpated anatomic landmarks and anthropometric measures.
The motion reconstruction is a two step process. First is the kinematic motion reconstruction, which calculates the joint angles in function of time, using the recorded markers trajectories as input data. This process takes into account joint constraints and subject anthropometric dimensions, thus obtaining a consistent motion. Once the joint angles are known, the inverse dynamics reconstruction is performed. This second step allows calculating the reaction joint forces and torques as well as the motor forces that are needed to produce the recorded motion. These motor forces are equivalent to all the forces applied by the different muscles. In addition to the recoded motion, the inverse dynamics reconstruction requires as input data the external forces applied to the body, that are also provided by the experiments. Figure 6 shows an example of reconstructed motion for the lower limb.
5-Multi-Body dynamic motion simulation
One of the objectives of the DHErgo project is to predict a task-oriented motion. The motion prediction method developed in DHErgo is dynamic, in order to take into account the dynamic variables in the motion (such as joint torques, external contacts). Moreover, the developed motion prediction method is a hybrid data-and-knowledge based method, relying both on a database of captured motions as reference and on the definition of a motion control law to guide the predicted motion.
The motion prediction is carried out selecting from the database the motion which most suitably resembles the prediction conditions (subject and environment characteristics) and modifying it to meet the new goals in the motion. The predicted motion is obtained by solving a constrained optimization problem. The constraints to the motion are the fulfillment of the task as well the dynamic equilibrium of the DHM across the motion. Contact models have been employed to estimate the reaction forces of the environment due to its interaction with the DHM. On the other hand, the objective function seeks to resemble the joint angle profiles of the reference motion and to follow the motion control law defined as the resemblance with the joint power profiles of the reference motion.
The method has been applied to clutch-pedal depression motions (Figure 7), predicting the motion of a subject (which does not match the reference subject) in an environment (which does not match the reference environment), and validating it against actually performed motions in the predicted conditions. The results show that the predicted motion resembles the real motions both in trajectories (Figure 8) and in joint torque profiles (Figure 9).
6- Human musculo-skeletal motion reconstruction with consideration of the physical contact with a deformable environment
One of DHErgo objectives is to achieve a model for the reproduction of human musculo-skeletal (MS) motion, coupled with an accurate estimation of the contact between the MS model and its environment, by taking physical deformations of model and environment into account. The proposed methodology has been focused on the interaction between human occupant and driver seat in passenger cars in order to precisely define the seat contact force acting to the back, buttock and thigh segments and to model the compression of soft tissues. The methodology has been applied to pedal clutching motions in DHErgo. The MS model and environment are represented and analyzed by the PAM-Comfort™ software, commercialized by ESI Group for the industrial design of car seat structures.
A complete procedure from model scaling to MS motion reconstruction and estimation of the muscle force distribution has been developed. At first the MS model needs to be scaled according to a subject anthropometry. The bones are scaled based on anatomical landmark palpation, while the soft tissues are scaled from external dimension measurements like circumference (Figure 10). The muscle parameters are adjusted, in particular using their maximum isometric forces. In a second step, the motion of the scaled MS model is reconstructed from the captured motion performed by a human volunteer. This inverse kinematics motion provided by CEIT is converted into kinematic constraints to reproduce the motion with PAM-Comfort. The calculated human/seat interaction corresponds to a dynamic distributed loading at all skin nodes in contact with the seat. At the last step, at each selected time frame the muscle force distribution can be calculated with PAM-Muscle by taking into account the external forces (gravity, pedal force and human-seat interaction) and the inertia forces.
This process of MS motion reproduction was applied to an example of clutch pedal operation. After the scaling of the MS model to a specific subject, one of the clutch pedal motions performed by this subject was reproduced with the scaled model through PAM-Comfort in order to obtain the different steps of this motion as well as the estimation of the physical contact with the deformable seat (Figure 11). Based on the pedal force measured during the experiment and based on the calculated human-seat interaction, the joint moments as well as the muscle force distribution were calculated (Figure 12).
The different experimental data measured for this specific subject were analyzed by ULB. Relationships between muscle activity and expected muscle force could be obtained by taking into account muscle moment arms behavior. For the analyzed range of this pedal clutching motion mean value of the moment arm is ~0.06 m. By this value maximum agonist force values are expected about ~500 N for the group of Rectus Femoris and Vastus Intermedialis as well as for Vastus Lateralis and about ~300 N for Vastus Medialis. The maximum calculated forces of the three main muscles acting in knee extension are in agreement with the expected values defined by this analysis. While Vastus Lateralis and Vastus Medialis present a similar behaviour along the motion, Rectus Femoris + Vastus Intermedialis group is more activated at the beginning and the end of the motion. This can be explained by a high contribution of the hip joint at these stages and by the fact the Rectus Femoris muscle is bi-articular, acting on the hip as one of its main flexor. The correlation with the knee extension moment is better for Vastus Lateralis and Vastus Medialis than for the group of Rectus Femoris and Vastus Intermedialis. In fact Vastus Lateralis and Vastus Medialis act on the knee only and are its main extensors.
7- Motion related discomfort assessment and application to three case studies
The evaluation of motion related discomfort is one of the critical issues for digital human modelling. Existing ergonomic assessment methods were initially developed for the observation of working postures in the industry. They can be helpful in detecting the main risk factors at a workplace but can hardly be used for ergonomic evaluation of a product such as a vehicle. In DHErgo, an integrated approach for modelling both motion and discomfort has been adopted, based on the hypothesis that a better comfort can be obtained when people can make the appropriate adjustments by themselves. These less constrained motions can then be referred as reference data for comparing a proposed solution. This approach, illustrated in Figure 13 was investigated through three case studies which were performed at IFSTTAR (figure 14) with the active support of the three car manufacturers involved in the project:
• A lower limb task: clutch pedal operation
• An upper limb task: handbrake operation
• A whole body task: car ingress/egress
For each case study, the main critical design parameters that may affect the discomfort perception were identified and studied in an experiment with voluntary subjects and a well-planned experimental design. To manipulate the independent variables and to measure the dependent responses, a multi-adjustable experimental mock-up was used, with the necessary adjustments allowing the participants to easily choose their preferred car configuration. An optoelectronic motion capture system was also used to measure the trajectories of the markers attached to the body surface. Meanwhile, external contact forces were measured using force sensors. With these inputs and an individualized DHM, the movements were reconstructed using inverse kinematics and inverse dynamics procedures to access joint angles and torques during motions.
Then the basic idea of our approach is to compare the imposed and less constrained movements in order to identify relevant biomechanical parameters for defining objective discomfort indicators. Thanks to this generic approach, discomfort criteria have been identified for assessing clutching, hand brake operation and ingress movements. For instance, in case of the automotive clutching task, four indicators were defined and succeeded in differentiating the tested configurations in agreement with experimental observations.
8- DHErgo demonstrator and design oriented solutions using a digital human simulation tool
The design and implementation of the project demonstrator is driven by two main objectives. First the simulation tools, which are generated in the various research fields of the project, are integrated into a common software platform. On the one hand a DMU environment is established to simulate the design in a virtual way, and on the other hand, kinematic, dynamic and muscular motion and discomfort simulation tools are integrated into this environment based on the existing human models RAMSIS and PAM Comfort (Figure 15).
The second main objective is the demonstration of the feasibility of using the tools in design applications. In this context several design questions have been collected at the beginning of the project.
One of the main questions regards the investigation of critical product users for a specific design. These users cause the design to loose against a given benchmark and hence have to be analysed by designers in a detailed way. This analysis is supported by the demonstrator through a critical manikin identification feature (Figure 16).
The designer specifies a design evaluation score or rating coming from guidelines or a reference vehicle design and the demonstrator provides the percentage of the population being below the benchmark and the corresponding virtual manikins representing this population segment. The detailed investigation of the human design interaction of these manikins will bring up information to improve the design.
The second main design question deals with the identification of optimal designs by comparing different design alternatives. This task is supported by the demonstrator through a design comparison and optimization feature (Figure 17).
The designer specifies a population percentage for which the design should fit properly and the demonstrator provides the corresponding benchmark scores of different design alternatives, in particular the design with the optimal score. This analysis helps the designer to find optimal designs with regard to objective criteria and the target user population.
Both features are implemented into the design solution architecture and process developed in various work packages and displayed in Figure 17.
9- DHErgo Benefits for the end users
In the field of Ergonomics, digital Humans are well established within the automotive product development process. But of course the demands are growing with their successful application. At the beginning of the DHErgo project, end user requirements were defined by Renault, PCA and BMW. These can be structured in demands regarding the definition of manikins, digital mockup, motion simulation and evaluation of functionalities. Comparing these end user requirements with the project results by taking the experience with the pre-released Demonstrator and the presented forecast for the final Demonstrator into account, it can be stated that most of the requirements are or will be achieved.
Within the DHErgo Demonstrator it is now possible to create manikins based on experimentally measured anatomical landmarks. Besides other improvements - e.g. the calculation of joint centers - this approach is highly welcome because it simplifies the experimental workflow and increases the reliability for all calculated parameters based on anatomical issues. Of course the definition of manikins by the means of an anthropometric database is also offered within the DHErgo Demonstrator, and in the final version the database will include parameters for anthropometrics, range of motion (RoM), strength and age. Furthermore with this database the generation of a user defined population will be possible. This function is very appreciated since a specific customer group containing a huge amount of different manikins can be created and used for the design evaluation instead of some extreme percentiles.
The product development process of vehicles mainly takes places in virtual reality. Hence the integration of an interface for importing CAD data into the DHErgo Demonstrator was an indispensable part. The disadvantage of this approach is that the importing process has to be repeated by the end user every time when the vehicle geometry changes. As a solution the DHErgo Demonstrator offers another tool which is called “Parametric Scene”. The end user needs to define relevant dimensions for the task to be evaluated, e.g. pedal position and travel length. Once these dimensions are defined, the virtual mockup can be adjusted easily by just changing the values for the relevant dimensions and this makes the daily work with the tool quite efficient.
A main topic of the DHErgo project is the dynamic motion simulation including different motion strategies. Since the Demonstrator currently provided to the end users only includes the movement of the clutch pedal operation and some improvements have still to be implemented, information about the accuracy and reliability of this functionality cannot be stated. But it is obvious that the motion simulation leads to a better implementation of user demands, e.g. regarding the clearance between the manikin and the geometry done with a 3D motion silhouette (Figure 18) or taking pedal resistance forces and their effect on muscular activity into account.
Another DHErgo Demonstrator tool for a design evaluation with respect to a given benchmark is described in the chapter above. From the end user point of view this could be a powerful tool for making a decision between two designs, if the chosen evaluation score is reliable and meaningful. To assure this, the evaluation score is calculated in a very open and general manner. It is applicable to all motions and every biomechanical parameter can be included in the score calculation, meaning that every company can compose their own evaluation index. Furthermore the evaluation score gives information either for every chosen parameter or can be merged to one meaningful index value.
In summary, the big advantage of the DHErgo Demonstrator is that all functionalities are packed together in an integrated solution. Therefore in addition to the functional enhancements, the work efficiency is also highly improved.
Potential Impact:
The main impacts of DHErgo project can be summarized in three areas.
1- Improving scientific knowledge about human musculoskeletal structure, human functional capacities, motion simulation, discomfort perception
A large amount of human data has been collected in this project for developing realistic human models. For instance, data of joint strength and joint range of motion with consideration of adjacent joints are collected at TUM and IFSTTAR. In addition to detailed musculoskeletal data, ULB also investigated shoulder complex motion both in vitro and in vivo so as to find the mechanical relationships between the humerus position and orientation and the related attitude of the scapula and clavicle. It is now possible to define an individualized human model from manually palpated anatomic landmarks. An extensive human functional data base has been created, containing the joint maximum torque and ROM of all major joints for a same individual. A hybrid knowledge and data based dynamic motion simulation algorithm has been developed, including obstacle avoidance and contact force estimation. A musculoskeletal human model of the whole body has been refined based on accurate anatomic data. It is now possible to estimate muscle forces involved in a motion while taking into account human/seat interaction. Less-constrained motion concept has been successfully applied to three case studies for identifying motion-related discomfort assessment criteria. All these achievements improve the current scientific knowledge about human musculoskeletal structure, human functional capacities, motion simulation, discomfort perception.
2- Improving the competitiveness of human simulation tools
The DHErgo project opens an opportunity window not only for European companies to maintain the leadership in human-centred design for automotive manufacturers, but also for simulation tools providers. As direct examples we can mention, the RAMSIS human model developed by Human Solutions, which is the only ergonomic simulation software package in Europe. 70% worldwide car manufacturers make use of RAMSIS for car design. Also, ESI is well known for its software for crash simulation in safety application. In recent years, they have extended their simulation offer to comfort evaluation. DHErgo results will improve dramatically the position of both companies in the global market.
3- Improving the product ergonomics for end-users
In the traditional process of product developments, the human comes into play in a very late stage. When development is quite advanced, a physical mock-up is built and several tests with humans are performed on it. The test evaluation suggests design corrections which are fed back to the design. That requires a lot of resources regarding physical prototyping, test with real subjects and modifying the product design. Additionally the physical mock-ups have to be updated several times, but the work can hardly be re-used for other product variants. Comparing to physical tests, virtual human tests can be performed at each level of design process in CAD by one person running a DHM on a computer. The CAD product model is extracted from the company database and the evaluation is done from a virtual testing. This reduces the required time from days for the physical tests to hours on the computer. In addition, the virtual tests can be easily re-used for other product variants.
Thanks to the potential advantages of using DHMs, the DHM solutions have been becoming well installed in the standardized product development processes especially in automotive and aeronautic industries. Industrial use of DHMs for product design can be back to more than 15 years ago. Some companies even developed their own DHMs. For instance, Renault developed its own man model named “Man3D” in the late 1980s with help of INRETS and kept its use till to 1995 before switching to the commercialized software RAMSIS. Nowadays, more and more new design engineers are trained for using a digital human in product development. Industrial demands on virtual simulation based DHMs are becoming higher and higher. DHMs are used not only for static and geometric evaluation of product design regarding geometrical criteria such as reach capacity and visibility, but also for dynamic assessment taking motion and discomfort into account. Many tests, as for example ingress/egress of a car, have to be carried out on a physical mock-up currently. In order to reduce efforts, the companies have started to use existing static DHMs to simulate dynamic motions. Due to lack of functionalities of current DHMs, some companies conduct in-house studies to find their proper solutions. This background will cause no initial acceptance barriers of the new project technologies at potential end users. The current industrial demands for these technologies will rather easy and speed up their introduction into the design process.
The main outcome of the DHErgo project, the software demonstrator including all relevant research results, will show the feasibility of evaluating product design in a very early virtual design stage with respect to new applications. They mainly consist of simulating complex human motion behaviour in a virtual environment (for example ingress) and assessing the task induced discomfort. Current DHM modelling capabilities are therefore enhanced by dynamic and musculoskeletal simulation, making it possible to provide a more trustable virtual user replacing the real user.
In addition to these technical modelling benefits, the software demonstrator will provide design oriented solutions, which help the user to be more focused on design problems rather than the DHM use. In particular the software evaluates the design directly rather than assessing the human behaviours. For that, entire human populations including nation, age, gender and anthropometrics are considered in the background to assess a design variant. Moreover, the software will be able to optimize design proposals regarding user specific criteria. The evaluation and optimisation processes are set up by ergonomic experts, while these processes are applied in specific applications by usual design engineers requiring a minimum of time.
There is also a need to make compatible between different human models used for ergonomic and safety simulations so as to reduce product development time. This is particular true in automotive industries, as ergonomic and safety issues of a car design is in general treated in different services and the human models for ergonomic and safety simulations are completely incompatible. The DHErgo project has developed necessary interfaces so as that two types of human models well established in the automotive industry will be easily communicated, one ergonomic simulation man model RAMSIS from Human Solutions and the other PAM human model from ESI. This facilitates the trend towards the “virtual factory”, where effective and speedy numerical product design and optimization is carried out simultaneously by all design departments, which requires an easy exchange of models and data.
Summarizing, the DHErgo project will contribute to a higher competitiveness of product development by reducing development lead time and improving the ease of use of product.
Perspectives
In addition to the new scientific knowledge on human modelling, a new generation of digital human modelling software packages, capable of performing dynamic motion simulation and of providing objective discomfort indicators, will be available in the coming years. This will be facilitated by the two major software editors (Human Solutions and ESI) involved in the project. However, there are still many challenging issues in digital human modeling
- Difficulty for collecting all necessary data for human body modeling. Musculoskeletal modelling required various heterogeneous types of data (bone and joint morphology, muscle geometry, muscle physiological properties, joint and limb kinematics, etc). These data are usually gathered by researchers from various sources (literature, public repository, etc) which have no links with each other because collected during various data collection procedure on different specimens or volunteers. Objective and validated data registration is therefore difficult to achieve. During DHErgo, musculoskeletal data collected from ULB past projects (e.g. VAKHUM and LHDL) were extended by accurate full body soft tissue reconstruction, kinematical data collection for lower limb (e.g. daily motion and pedal clutching) and upper body (e.g. shoulder complex rhythm evaluation). Though some progresses have been achieved in anatomic data collection from one unique subject by ULB, such an effort should continue in the future.
- Difficulty of musculoskeletal model validation. Validation of musculoskeletal model is one of the most challenging aspects of the project. This kind of models includes a large numbers of variables and parameters that are difficult to evaluate independently, in group and in vivo. This is still a key issue in the related literature. Some work on this topic using “traditional” measurements tools (EMG, motion analysis systems, etc) has been performed in the project in order to validate the prediction tools developed by the consortium. However, the final models are be entirely validated because full validation would require new experimental tools that are yet to be developed. Such development is clearly outside the project scope and need to be investigated in the future.
- More human functional data are needed for objective discomfort assessment. The evaluation of the discomfort associated with a complex task-orientated motion is another challenging aspect of the project. During the project, two complementary approaches were undertaken, one focusing on discomfort modeling at joint level making use of maximum joint range of motion and joint strength, another focusing on identifying discomfort assessment indicators of a task oriented motion by comparing imposed and less-constraint motions. It revealed that functional data like joint maximum range of motion, joint strength, maximum hand and foot maximum exertion strength are reference data for objective comparison with actual values required for a task. Though a limited amount of data has been collected from 10 young and 10 old subjects for illustrating the feasibility of concepts and methods, functional data collection efforts should be continued covering a large range of population. Musculo-skeletal model based methods should be further developed for human functional data collecting.
- Lack of a generic human complex motion simulation method capable of taking motion variability into consideration. During DHErgo project, data based approach for motion simulation has been successfully extended to dynamics capable of taking into account motion dynamics constraints (equilibrium, external force contacts, etc…). But this approach is task specific and difficult to be applied to multi-sequential activities that are frequently encountered in a production assembly line. More generic motion methods are needed for simulating human multi-sequential complex motion.
Clearly, more human data and more collaboration between different scientific communities are required to improve DHMs for product design and manufacturing. New research issues related to DHM in transports can be summarized as follows:
- Elderly population is growing. Car manufacturers have to give greater attention to older consumers and workers in manufacturing process. Digital human models should be able to simulate ageing effects.
- Digital human modelling for virtual manufacturing
To reduce the risk of musculoskeletal disorders while improving productivity
To be able to simulate a high variety of human activities
- Integrated car design considering normal use, pre-crash and crash
To develop an integrated human simulation tool for car design engineers.
Contact: IFSTTAR
Dr Xuguang Wang
Laboratoire de Biomécanique et Mécanique des Chocs- LBMC
25, avenue François Mitterrand, Case 24
F-69675 Bron, France
Email: xuguang.wang@ifsttar.fr
Tel: +33 (0)4 72 14 24 51
Fax: +33 (0) 4 72 37 68 37
Digital Mock-Ups (DMU) together with Digital Human Models (DHM), are more and more used in the early phase of product design in order to reduce the product development time and cost. The main objective of the DHErgo project is to develop anatomically and bio-mechanically sound human models for evaluating posture and motion related discomfort. The project started in September 2008 for 39 months and gathers 10 European partners. It is structured in eight work packages in charge of the technical work of the project and two work packages in charge of the project management and dissemination activities. Due to the complexity of modelling human body and data availability (especially for the detailed human musculoskeletal data), the project is structured in three overlapped cycles according to the body part to be modelled: 1) Lower limbs, 2) Upper limbs including the shoulder, 3) Whole body including the torso. Each cycle covers the whole core technical work, namely the data collecting / motion reconstruction - analysis - modelling - implementation - assessment sequence. A large amount of human data has been collected for developing realistic human models. New approaches for data collection were developed and highly valuable data were obtained for human functional data. For instance, data of joint strength and joint range of motion with consideration of adjacent joints were collected at TUM and IFSTTAR. In addition to detailed musculoskeletal data, ULB also investigated shoulder complex motion both in vitro and in vivo so as to find the mechanical relationships between the humerus position and orientation on one hand, and the related attitude of the scapula and clavicle at the same moment of time on the other hand. The algorithm combines corresponding fitting curves to evaluate motion prediction by quadric multiple regression. Accuracy of the reconstruction allows applying developed method in ergonomics and daily activity analysis. CEIT has proposed a dynamic motion simulation method which takes into account the constraints due to motion dynamics. The results show that the predicted motion resembles the real motions both in trajectories and in joint torque profiles. The tool has been validated against the database of reconstructed experimental motions. The reconstructed motions using a multi-body human model can now be exported into ESI’s musculoskeletal model including deformable soft tissues so as to estimate muscle forces as well as contact pressure. Three automotive application oriented case studies have been carried out at IFSTTAR in close collaboration with three car manufacturers as end-users in the project. These data are used for testing the motion and discomfort simulation methods. HS has implemented a new version of the DHErgo demonstrator which integrated the simulation tools on the existing human models RAMSIS and PAM Confort. BMW validated the demonstrator on two major aspects. The first one is the simulation and the deviation between simulated and recorded results, e.g. for motion trajectories or discomfort ratings. The second one is about the user interface and the handling of the demonstrator. Most of the requirements defined by the three end-users were achieved. ERT as project co-coordinator in charge of the dissemination activities and administrative management has organized the participation in key conferences. All the materials (newsletters, final leaflet, posters,…) have been widely distributed in the major DHM related events. All publications have been uploaded on the DHErgo website.
Project Context and Objectives:
Digital human simulation is believed to be one of the key technologies for proactive design. All components of a product can be digitalized using today’s CAD (Computer-Aided Design) technologies. Being able to evaluate the customer’s satisfaction through digital human simulation at a very early stage of product design becomes an absolute necessity for reducing design cycle. It can be easily understood that such an advanced digital human simulation tool can be applied in a large industrial domain where the designed system is addressed to a high number of users (clients/operators) and its physical mock-up is difficult to reproduce. Automotive industry is one of most active end-users of such a simulation tool because of high pressure for reducing time-to-market and cost. The final results of the project will mainly be demonstrated through case studies related to car design, thought they can be easily transferred to other industries.
In order to evaluate the physical aspects of a product or a workplace in its early design stage, a design engineer will insert a digital human representing one critical target user of the product in the CAD (Computer-Aided Design) environment at first, then simulate critical postures or movements and evaluate how the target user will judge it in terms of:
• Visibility. Can the target user see all necessary visual information?
• Clearance and accessibility. Does the product have the sufficient room so that the target user can carry out all necessary movements freely without collision?
• Ease of use. Can the target user reach and manipulate the controls easily within his/her physical capacity (strength, joint maximum range of motion, equilibrium keeping)?
Ideally, the design engineer would also like to be guided by the recommendations of the DHM simulation software for modifying the design and finding a compromise without degrading the ergonomic demand.
Since a product like an automobile can be used by thousands even millions of people, the design engineer has to consider the variability of user population. People vary not only in body dimension and shape (anthropometry), but also in physical capacity (body strength and joint mobility). Besides, the human body has in general more degrees of freedom than imposed by a task, implying that people can adopt different ways to execute the same task. The variability in postural and motion control has also to be considered in simulation. Often the design engineer has to answer the question what percentage of the whole target population (e.g. German car drivers) is accommodated for a specific task under consideration, like car ingress and egress for instance. Among the population of the drivers, there are more and more elderly people especially in developed countries. One has to take into account their specific needs due to aging.
However, currently available digital human simulation tools are still far from the designer’s expectations (see Chaffin, 2005 for an excellent review), especially in terms of motion simulation and evaluation of motion related discomfort. There are three main commercialized DHM packages used for ergonomic design of products:
• Jack, a software initially developed by University of Pennsylvania and now commercialized by UGS, http://www.ugs.com/products/tecnomatix/human_performance/jack/
• Safework, initially developed by a Canadian ergonomic consulting company, now purchased by Dassault Système and renamed as Human Builder in CATIA, http://www.3ds.com/
• Ramsis, developed and commercialized by Human Solutions with the support of German car makers, http://www.human-solutions.com
For these three human modelling packages, the human model behind is mainly a geometric and kinematic representation of the human body. Clearly, this is a first step towards a comprehensive ergonomic simulation tool for design engineers. Interdisciplinary research efforts have to be continued not only for realistic motion simulation but also for understanding and predicting posture and motion related discomfort. Relating the perceived discomfort with design parameters through human/product interaction simulation is a fundamental requirement for ergonomic applications.
The digital human models used for ergonomic design of products have to evolve towards dynamic and muscular ones. They must be able not only to have a realistic visual representation of human body and movement but also to evaluate the muscular efforts associated with a task for a better understanding of human performance and perceived discomfort. DHErgo mainly focused on the following scientific issues:
• Development of multi-body dynamic motion reconstruction methods in order to estimate joint motion and joint forces
• Integration of existing human anatomic data and development of detailed human musculoskeletal digital models
• Collecting necessary data for model developing and validation. Collecting of the basic human functional data such as joint strength and joint limit will be continued.
• Development of a hybrid optimisation/data based complex motion simulation approach. Optimisation criteria based on joint force and muscle performance will be explored.
• Development of a generic motion-based biomechanical discomfort criterion. Thanks to the musculoskeletal models to be developed, different biomechanical parameters including muscle forces and other soft tissues loads will be investigated.
• Elderly people. Basic biomechanical data of the elderly people will be collected. The effects of aging on movement and perceived discomfort will be studied.
Project Results:
DHErgo has fulfilled the most of the objectives defined at the beginning of the project. Here are the main achievements.
1-Human functional data for a more realistic representation of human physical capacity
To achieve a more realistic representation of human physical capacity extensive experimental data are obtained during the project. Both maximum joint range of motion (RoM) as well as maximum joint torque data are measured for all major joints of the human body (figure 1). To account for expected differences in age and gender 18 males and females are studied forming two age groups each: younger than 35 years and older than 65 years.
An important and new aspect in these measurements is the consideration of dependencies between different joint degrees of freedom of one joint as well as adjacent joints due to bi-articular muscles, which has not been done to this extent before. These relations are studied using specifically designed measurement devices and motions are recorded using a Vicon motion capture system. On the basis of anatomical considerations and experimental feasibility thoroughly planned experimental protocols are defined allowing for a very detailed analysis while limiting the experimental effort for the subjects to an acceptable level. The measurements are distributed on 20 sessions per subject lasting about two hours each.
The resulting data are analyzed using inverse kinematics and dynamics and statistical methods. One exemplary result for joint RoM is the relation of maximum hip flexion and knee flexion RoM which could be quantified and modeled. Considering maximum joint torque numerous relations were found. It could for instance be shown that elbow flexion torque is highly dependent on the elbow flexion angle as well as the shoulder flexion angle. Furthermore the expected decrease of joint torque with increasing age and the differences in gender could be confirmed and partially quantified.
The resulting mathematical models of joint torque – joint angle relations as well as inter-joint relations can now directly be applied to digital human models allowing for a more realistic simulation of physical capacity. F. Engstler & F. Guenzkofer [TUM]
2- Detailed human anatomic data collection
Musculoskeletal data collected from ULB past projects (e.g. VAKHUM and LHDL) were extended during DHErgo project by accurate full body soft tissue reconstruction, kinematical data collection for lower limb (e.g. daily motion and pedal clutching) and upper body (e.g. shoulder complex rhythm evaluation). These data were also reconstructed by ULB to obtain anatomically correct joint motion behavior for muscle force evaluation, taking into account electromyography of the selected muscles.
Bone and soft tissue (muscles, ligaments) 3D morphology reconstruction, including extraction of muscle and tendon fibre topology and attachments, contains 3D reconstruction of dissected soft tissues as alternative to high resolution MRI that shows some limitations when whole specimen body has to be scanned. ULB proposed a method using stereophotogrammetry on the basis of digital cameras allowing soft tissue data collection and maintaining anatomically-correct results. Prior to dissection, a specimen has been fully CT-scanned to obtain bone 3D models. During specimen dissection, pins with colored heads (CHs) were inserted in muscles and ligaments to characterize muscle and ligament fibre path, musculo-tendinous junctions, origins and insertions (Figure 2). CHs data registration to the 3D bone models occurred using technical frames including reflective markers and aluminium balls inserted into the bones for better accuracy. Once registered, CHs information was reconstructed in 3D together with the bone models (Figure 3). Muscle and tendon fiber length and pennation angles were evaluated after piece-wise linear approximation of the reconstructed points. The method has been applied on an entire human body. By V. Sholukha and S. Van Sint Jan [ULB]
3- Evaluation of the shoulder rhythm mechanism by quadric multiple regression
The human shoulder complex allows an important spatial displacement of the underlying bone segments thanks to the sequence of three intermediate joints connecting the humerus to the thorax. Due to anatomic constraints, their motions are coupled. The contribution of the scapular displacements, linked to the clavicle behavior, within the overall arm elevation follows a general pattern in which scapular motion is responsible for approximately one third of the total arm elevation. It is therefore to respect natural shoulder motion pattern (shoulder rhythm) when reconstructing and simulating human arm motion.
The main challenge of the underlying research is to find the mechanical relationships between the humerus position and orientation on the one hand, and the related attitude of the scapula and clavicle at the same moment of time at the other hand. This answers a practical problem of in-vivo motion analysis. Indeed, if data related to the humerus instantaneous spatial position is relatively easy to obtain, the same information is more difficult to collect for the clavicle and the scapula. Furthermore, a unique and straightforward mechanism like in the knee or the ankle joint is not possible, e.g. the humeral head shows similar displacement during shoulder elevation or scapula-humeral abduction due to the fact the vertical displacement of the humerus is closely linked to clavicle elevation. An algorithm that aims to estimate the clavicle and scapula pose compared to the humerus instantaneous posture, must first deal with the differentiation between a shoulder elevation due to either a humerus rotation or a translation, to only then combine correspondent fitting curves to evaluate motion prediction by quadric multiple regression.
In a study by ULB, data from three volunteers and two specimens were processed for a total of 51 different motions including abduction-adduction (frontal plane), flexion-extension (sagittal plane) and humeral head vertical ascent-descent. Typical example of the rhythm components fitting by linear and parabolic polynomial functions are presented in Figure 4 for three repetitions of flexion/extension. Figure 5 presents results of motion reconstruction and application of prediction-correction shoulder rhythm. Analysis of 15 different motions shows a mean accuracy for the clavicle and scapula rhythm reconstruction of 8.5 (SD=6.1) mm and 19.8 (SD=5.2) mm, respectively. Accuracy of the reconstruction allows applying developed method in ergonomics and daily activity analysis.
4-Inverse kinematic and inverse dynamic motion reconstruction methods
One of the major results produced by CEIT, as expert in dynamic multi-body modeling, is a method for the whole body motion reconstruction. This method has been implemented in a software tool that is used for reconstructing the motions recorded during the project experiments. A novel feature of the proposed method is that a generic human model is automatically scaled to the subject’s actual dimensions, based on palpated anatomic landmarks and anthropometric measures.
The motion reconstruction is a two step process. First is the kinematic motion reconstruction, which calculates the joint angles in function of time, using the recorded markers trajectories as input data. This process takes into account joint constraints and subject anthropometric dimensions, thus obtaining a consistent motion. Once the joint angles are known, the inverse dynamics reconstruction is performed. This second step allows calculating the reaction joint forces and torques as well as the motor forces that are needed to produce the recorded motion. These motor forces are equivalent to all the forces applied by the different muscles. In addition to the recoded motion, the inverse dynamics reconstruction requires as input data the external forces applied to the body, that are also provided by the experiments. Figure 6 shows an example of reconstructed motion for the lower limb.
5-Multi-Body dynamic motion simulation
One of the objectives of the DHErgo project is to predict a task-oriented motion. The motion prediction method developed in DHErgo is dynamic, in order to take into account the dynamic variables in the motion (such as joint torques, external contacts). Moreover, the developed motion prediction method is a hybrid data-and-knowledge based method, relying both on a database of captured motions as reference and on the definition of a motion control law to guide the predicted motion.
The motion prediction is carried out selecting from the database the motion which most suitably resembles the prediction conditions (subject and environment characteristics) and modifying it to meet the new goals in the motion. The predicted motion is obtained by solving a constrained optimization problem. The constraints to the motion are the fulfillment of the task as well the dynamic equilibrium of the DHM across the motion. Contact models have been employed to estimate the reaction forces of the environment due to its interaction with the DHM. On the other hand, the objective function seeks to resemble the joint angle profiles of the reference motion and to follow the motion control law defined as the resemblance with the joint power profiles of the reference motion.
The method has been applied to clutch-pedal depression motions (Figure 7), predicting the motion of a subject (which does not match the reference subject) in an environment (which does not match the reference environment), and validating it against actually performed motions in the predicted conditions. The results show that the predicted motion resembles the real motions both in trajectories (Figure 8) and in joint torque profiles (Figure 9).
6- Human musculo-skeletal motion reconstruction with consideration of the physical contact with a deformable environment
One of DHErgo objectives is to achieve a model for the reproduction of human musculo-skeletal (MS) motion, coupled with an accurate estimation of the contact between the MS model and its environment, by taking physical deformations of model and environment into account. The proposed methodology has been focused on the interaction between human occupant and driver seat in passenger cars in order to precisely define the seat contact force acting to the back, buttock and thigh segments and to model the compression of soft tissues. The methodology has been applied to pedal clutching motions in DHErgo. The MS model and environment are represented and analyzed by the PAM-Comfort™ software, commercialized by ESI Group for the industrial design of car seat structures.
A complete procedure from model scaling to MS motion reconstruction and estimation of the muscle force distribution has been developed. At first the MS model needs to be scaled according to a subject anthropometry. The bones are scaled based on anatomical landmark palpation, while the soft tissues are scaled from external dimension measurements like circumference (Figure 10). The muscle parameters are adjusted, in particular using their maximum isometric forces. In a second step, the motion of the scaled MS model is reconstructed from the captured motion performed by a human volunteer. This inverse kinematics motion provided by CEIT is converted into kinematic constraints to reproduce the motion with PAM-Comfort. The calculated human/seat interaction corresponds to a dynamic distributed loading at all skin nodes in contact with the seat. At the last step, at each selected time frame the muscle force distribution can be calculated with PAM-Muscle by taking into account the external forces (gravity, pedal force and human-seat interaction) and the inertia forces.
This process of MS motion reproduction was applied to an example of clutch pedal operation. After the scaling of the MS model to a specific subject, one of the clutch pedal motions performed by this subject was reproduced with the scaled model through PAM-Comfort in order to obtain the different steps of this motion as well as the estimation of the physical contact with the deformable seat (Figure 11). Based on the pedal force measured during the experiment and based on the calculated human-seat interaction, the joint moments as well as the muscle force distribution were calculated (Figure 12).
The different experimental data measured for this specific subject were analyzed by ULB. Relationships between muscle activity and expected muscle force could be obtained by taking into account muscle moment arms behavior. For the analyzed range of this pedal clutching motion mean value of the moment arm is ~0.06 m. By this value maximum agonist force values are expected about ~500 N for the group of Rectus Femoris and Vastus Intermedialis as well as for Vastus Lateralis and about ~300 N for Vastus Medialis. The maximum calculated forces of the three main muscles acting in knee extension are in agreement with the expected values defined by this analysis. While Vastus Lateralis and Vastus Medialis present a similar behaviour along the motion, Rectus Femoris + Vastus Intermedialis group is more activated at the beginning and the end of the motion. This can be explained by a high contribution of the hip joint at these stages and by the fact the Rectus Femoris muscle is bi-articular, acting on the hip as one of its main flexor. The correlation with the knee extension moment is better for Vastus Lateralis and Vastus Medialis than for the group of Rectus Femoris and Vastus Intermedialis. In fact Vastus Lateralis and Vastus Medialis act on the knee only and are its main extensors.
7- Motion related discomfort assessment and application to three case studies
The evaluation of motion related discomfort is one of the critical issues for digital human modelling. Existing ergonomic assessment methods were initially developed for the observation of working postures in the industry. They can be helpful in detecting the main risk factors at a workplace but can hardly be used for ergonomic evaluation of a product such as a vehicle. In DHErgo, an integrated approach for modelling both motion and discomfort has been adopted, based on the hypothesis that a better comfort can be obtained when people can make the appropriate adjustments by themselves. These less constrained motions can then be referred as reference data for comparing a proposed solution. This approach, illustrated in Figure 13 was investigated through three case studies which were performed at IFSTTAR (figure 14) with the active support of the three car manufacturers involved in the project:
• A lower limb task: clutch pedal operation
• An upper limb task: handbrake operation
• A whole body task: car ingress/egress
For each case study, the main critical design parameters that may affect the discomfort perception were identified and studied in an experiment with voluntary subjects and a well-planned experimental design. To manipulate the independent variables and to measure the dependent responses, a multi-adjustable experimental mock-up was used, with the necessary adjustments allowing the participants to easily choose their preferred car configuration. An optoelectronic motion capture system was also used to measure the trajectories of the markers attached to the body surface. Meanwhile, external contact forces were measured using force sensors. With these inputs and an individualized DHM, the movements were reconstructed using inverse kinematics and inverse dynamics procedures to access joint angles and torques during motions.
Then the basic idea of our approach is to compare the imposed and less constrained movements in order to identify relevant biomechanical parameters for defining objective discomfort indicators. Thanks to this generic approach, discomfort criteria have been identified for assessing clutching, hand brake operation and ingress movements. For instance, in case of the automotive clutching task, four indicators were defined and succeeded in differentiating the tested configurations in agreement with experimental observations.
8- DHErgo demonstrator and design oriented solutions using a digital human simulation tool
The design and implementation of the project demonstrator is driven by two main objectives. First the simulation tools, which are generated in the various research fields of the project, are integrated into a common software platform. On the one hand a DMU environment is established to simulate the design in a virtual way, and on the other hand, kinematic, dynamic and muscular motion and discomfort simulation tools are integrated into this environment based on the existing human models RAMSIS and PAM Comfort (Figure 15).
The second main objective is the demonstration of the feasibility of using the tools in design applications. In this context several design questions have been collected at the beginning of the project.
One of the main questions regards the investigation of critical product users for a specific design. These users cause the design to loose against a given benchmark and hence have to be analysed by designers in a detailed way. This analysis is supported by the demonstrator through a critical manikin identification feature (Figure 16).
The designer specifies a design evaluation score or rating coming from guidelines or a reference vehicle design and the demonstrator provides the percentage of the population being below the benchmark and the corresponding virtual manikins representing this population segment. The detailed investigation of the human design interaction of these manikins will bring up information to improve the design.
The second main design question deals with the identification of optimal designs by comparing different design alternatives. This task is supported by the demonstrator through a design comparison and optimization feature (Figure 17).
The designer specifies a population percentage for which the design should fit properly and the demonstrator provides the corresponding benchmark scores of different design alternatives, in particular the design with the optimal score. This analysis helps the designer to find optimal designs with regard to objective criteria and the target user population.
Both features are implemented into the design solution architecture and process developed in various work packages and displayed in Figure 17.
9- DHErgo Benefits for the end users
In the field of Ergonomics, digital Humans are well established within the automotive product development process. But of course the demands are growing with their successful application. At the beginning of the DHErgo project, end user requirements were defined by Renault, PCA and BMW. These can be structured in demands regarding the definition of manikins, digital mockup, motion simulation and evaluation of functionalities. Comparing these end user requirements with the project results by taking the experience with the pre-released Demonstrator and the presented forecast for the final Demonstrator into account, it can be stated that most of the requirements are or will be achieved.
Within the DHErgo Demonstrator it is now possible to create manikins based on experimentally measured anatomical landmarks. Besides other improvements - e.g. the calculation of joint centers - this approach is highly welcome because it simplifies the experimental workflow and increases the reliability for all calculated parameters based on anatomical issues. Of course the definition of manikins by the means of an anthropometric database is also offered within the DHErgo Demonstrator, and in the final version the database will include parameters for anthropometrics, range of motion (RoM), strength and age. Furthermore with this database the generation of a user defined population will be possible. This function is very appreciated since a specific customer group containing a huge amount of different manikins can be created and used for the design evaluation instead of some extreme percentiles.
The product development process of vehicles mainly takes places in virtual reality. Hence the integration of an interface for importing CAD data into the DHErgo Demonstrator was an indispensable part. The disadvantage of this approach is that the importing process has to be repeated by the end user every time when the vehicle geometry changes. As a solution the DHErgo Demonstrator offers another tool which is called “Parametric Scene”. The end user needs to define relevant dimensions for the task to be evaluated, e.g. pedal position and travel length. Once these dimensions are defined, the virtual mockup can be adjusted easily by just changing the values for the relevant dimensions and this makes the daily work with the tool quite efficient.
A main topic of the DHErgo project is the dynamic motion simulation including different motion strategies. Since the Demonstrator currently provided to the end users only includes the movement of the clutch pedal operation and some improvements have still to be implemented, information about the accuracy and reliability of this functionality cannot be stated. But it is obvious that the motion simulation leads to a better implementation of user demands, e.g. regarding the clearance between the manikin and the geometry done with a 3D motion silhouette (Figure 18) or taking pedal resistance forces and their effect on muscular activity into account.
Another DHErgo Demonstrator tool for a design evaluation with respect to a given benchmark is described in the chapter above. From the end user point of view this could be a powerful tool for making a decision between two designs, if the chosen evaluation score is reliable and meaningful. To assure this, the evaluation score is calculated in a very open and general manner. It is applicable to all motions and every biomechanical parameter can be included in the score calculation, meaning that every company can compose their own evaluation index. Furthermore the evaluation score gives information either for every chosen parameter or can be merged to one meaningful index value.
In summary, the big advantage of the DHErgo Demonstrator is that all functionalities are packed together in an integrated solution. Therefore in addition to the functional enhancements, the work efficiency is also highly improved.
Potential Impact:
The main impacts of DHErgo project can be summarized in three areas.
1- Improving scientific knowledge about human musculoskeletal structure, human functional capacities, motion simulation, discomfort perception
A large amount of human data has been collected in this project for developing realistic human models. For instance, data of joint strength and joint range of motion with consideration of adjacent joints are collected at TUM and IFSTTAR. In addition to detailed musculoskeletal data, ULB also investigated shoulder complex motion both in vitro and in vivo so as to find the mechanical relationships between the humerus position and orientation and the related attitude of the scapula and clavicle. It is now possible to define an individualized human model from manually palpated anatomic landmarks. An extensive human functional data base has been created, containing the joint maximum torque and ROM of all major joints for a same individual. A hybrid knowledge and data based dynamic motion simulation algorithm has been developed, including obstacle avoidance and contact force estimation. A musculoskeletal human model of the whole body has been refined based on accurate anatomic data. It is now possible to estimate muscle forces involved in a motion while taking into account human/seat interaction. Less-constrained motion concept has been successfully applied to three case studies for identifying motion-related discomfort assessment criteria. All these achievements improve the current scientific knowledge about human musculoskeletal structure, human functional capacities, motion simulation, discomfort perception.
2- Improving the competitiveness of human simulation tools
The DHErgo project opens an opportunity window not only for European companies to maintain the leadership in human-centred design for automotive manufacturers, but also for simulation tools providers. As direct examples we can mention, the RAMSIS human model developed by Human Solutions, which is the only ergonomic simulation software package in Europe. 70% worldwide car manufacturers make use of RAMSIS for car design. Also, ESI is well known for its software for crash simulation in safety application. In recent years, they have extended their simulation offer to comfort evaluation. DHErgo results will improve dramatically the position of both companies in the global market.
3- Improving the product ergonomics for end-users
In the traditional process of product developments, the human comes into play in a very late stage. When development is quite advanced, a physical mock-up is built and several tests with humans are performed on it. The test evaluation suggests design corrections which are fed back to the design. That requires a lot of resources regarding physical prototyping, test with real subjects and modifying the product design. Additionally the physical mock-ups have to be updated several times, but the work can hardly be re-used for other product variants. Comparing to physical tests, virtual human tests can be performed at each level of design process in CAD by one person running a DHM on a computer. The CAD product model is extracted from the company database and the evaluation is done from a virtual testing. This reduces the required time from days for the physical tests to hours on the computer. In addition, the virtual tests can be easily re-used for other product variants.
Thanks to the potential advantages of using DHMs, the DHM solutions have been becoming well installed in the standardized product development processes especially in automotive and aeronautic industries. Industrial use of DHMs for product design can be back to more than 15 years ago. Some companies even developed their own DHMs. For instance, Renault developed its own man model named “Man3D” in the late 1980s with help of INRETS and kept its use till to 1995 before switching to the commercialized software RAMSIS. Nowadays, more and more new design engineers are trained for using a digital human in product development. Industrial demands on virtual simulation based DHMs are becoming higher and higher. DHMs are used not only for static and geometric evaluation of product design regarding geometrical criteria such as reach capacity and visibility, but also for dynamic assessment taking motion and discomfort into account. Many tests, as for example ingress/egress of a car, have to be carried out on a physical mock-up currently. In order to reduce efforts, the companies have started to use existing static DHMs to simulate dynamic motions. Due to lack of functionalities of current DHMs, some companies conduct in-house studies to find their proper solutions. This background will cause no initial acceptance barriers of the new project technologies at potential end users. The current industrial demands for these technologies will rather easy and speed up their introduction into the design process.
The main outcome of the DHErgo project, the software demonstrator including all relevant research results, will show the feasibility of evaluating product design in a very early virtual design stage with respect to new applications. They mainly consist of simulating complex human motion behaviour in a virtual environment (for example ingress) and assessing the task induced discomfort. Current DHM modelling capabilities are therefore enhanced by dynamic and musculoskeletal simulation, making it possible to provide a more trustable virtual user replacing the real user.
In addition to these technical modelling benefits, the software demonstrator will provide design oriented solutions, which help the user to be more focused on design problems rather than the DHM use. In particular the software evaluates the design directly rather than assessing the human behaviours. For that, entire human populations including nation, age, gender and anthropometrics are considered in the background to assess a design variant. Moreover, the software will be able to optimize design proposals regarding user specific criteria. The evaluation and optimisation processes are set up by ergonomic experts, while these processes are applied in specific applications by usual design engineers requiring a minimum of time.
There is also a need to make compatible between different human models used for ergonomic and safety simulations so as to reduce product development time. This is particular true in automotive industries, as ergonomic and safety issues of a car design is in general treated in different services and the human models for ergonomic and safety simulations are completely incompatible. The DHErgo project has developed necessary interfaces so as that two types of human models well established in the automotive industry will be easily communicated, one ergonomic simulation man model RAMSIS from Human Solutions and the other PAM human model from ESI. This facilitates the trend towards the “virtual factory”, where effective and speedy numerical product design and optimization is carried out simultaneously by all design departments, which requires an easy exchange of models and data.
Summarizing, the DHErgo project will contribute to a higher competitiveness of product development by reducing development lead time and improving the ease of use of product.
Perspectives
In addition to the new scientific knowledge on human modelling, a new generation of digital human modelling software packages, capable of performing dynamic motion simulation and of providing objective discomfort indicators, will be available in the coming years. This will be facilitated by the two major software editors (Human Solutions and ESI) involved in the project. However, there are still many challenging issues in digital human modeling
- Difficulty for collecting all necessary data for human body modeling. Musculoskeletal modelling required various heterogeneous types of data (bone and joint morphology, muscle geometry, muscle physiological properties, joint and limb kinematics, etc). These data are usually gathered by researchers from various sources (literature, public repository, etc) which have no links with each other because collected during various data collection procedure on different specimens or volunteers. Objective and validated data registration is therefore difficult to achieve. During DHErgo, musculoskeletal data collected from ULB past projects (e.g. VAKHUM and LHDL) were extended by accurate full body soft tissue reconstruction, kinematical data collection for lower limb (e.g. daily motion and pedal clutching) and upper body (e.g. shoulder complex rhythm evaluation). Though some progresses have been achieved in anatomic data collection from one unique subject by ULB, such an effort should continue in the future.
- Difficulty of musculoskeletal model validation. Validation of musculoskeletal model is one of the most challenging aspects of the project. This kind of models includes a large numbers of variables and parameters that are difficult to evaluate independently, in group and in vivo. This is still a key issue in the related literature. Some work on this topic using “traditional” measurements tools (EMG, motion analysis systems, etc) has been performed in the project in order to validate the prediction tools developed by the consortium. However, the final models are be entirely validated because full validation would require new experimental tools that are yet to be developed. Such development is clearly outside the project scope and need to be investigated in the future.
- More human functional data are needed for objective discomfort assessment. The evaluation of the discomfort associated with a complex task-orientated motion is another challenging aspect of the project. During the project, two complementary approaches were undertaken, one focusing on discomfort modeling at joint level making use of maximum joint range of motion and joint strength, another focusing on identifying discomfort assessment indicators of a task oriented motion by comparing imposed and less-constraint motions. It revealed that functional data like joint maximum range of motion, joint strength, maximum hand and foot maximum exertion strength are reference data for objective comparison with actual values required for a task. Though a limited amount of data has been collected from 10 young and 10 old subjects for illustrating the feasibility of concepts and methods, functional data collection efforts should be continued covering a large range of population. Musculo-skeletal model based methods should be further developed for human functional data collecting.
- Lack of a generic human complex motion simulation method capable of taking motion variability into consideration. During DHErgo project, data based approach for motion simulation has been successfully extended to dynamics capable of taking into account motion dynamics constraints (equilibrium, external force contacts, etc…). But this approach is task specific and difficult to be applied to multi-sequential activities that are frequently encountered in a production assembly line. More generic motion methods are needed for simulating human multi-sequential complex motion.
Clearly, more human data and more collaboration between different scientific communities are required to improve DHMs for product design and manufacturing. New research issues related to DHM in transports can be summarized as follows:
- Elderly population is growing. Car manufacturers have to give greater attention to older consumers and workers in manufacturing process. Digital human models should be able to simulate ageing effects.
- Digital human modelling for virtual manufacturing
To reduce the risk of musculoskeletal disorders while improving productivity
To be able to simulate a high variety of human activities
- Integrated car design considering normal use, pre-crash and crash
To develop an integrated human simulation tool for car design engineers.
Contact: IFSTTAR
Dr Xuguang Wang
Laboratoire de Biomécanique et Mécanique des Chocs- LBMC
25, avenue François Mitterrand, Case 24
F-69675 Bron, France
Email: xuguang.wang@ifsttar.fr
Tel: +33 (0)4 72 14 24 51
Fax: +33 (0) 4 72 37 68 37
