Research On Parametric Modeling And Simulation Of The Occupant Human Body Surface

With user’s comfort requirements more and more important today, ergonomics is gettingmore attention of researchers in the product design. As human is the key factor inergonomics, how to reflect occupants’ requirements in driving environment has become animportant problem in auto design, especially in auto body design. Building biomechanicshuman model is an effective way to solve this problem. But due to the diversity andcomplexity of human body, it makes building of digital human model which includes bodysurface and skeletal muscle system become a subject that need to be done for a long time.Taking it as a long-term goal, this thesis only studied on the establishment of a parameterizedmodel of human body surface, which close to the size and shape of real human body surface.As the relationships between anthropometric dimensions are the core content of the humanbody surface modeling, this thesis introduced the research in this field.Anthropometric dimensions and their relationships are very critical in human bodymodeling, so it must be considered in the first place. Taking the requirements of human bodysurface modeling and the present situation of anthropometric database to consideration,36anthropometric dimensions were selected for research. Because the shortage of data, usingthe numerical characteristics of USA’s ANSUR data a large sample data, which includes36human dimensions of30000samples, was generated. It laid the foundation of the latterresearches.In order to build a parameterized model of human body surface, getting the conversionrelationships between anthropometric dimensions is very important. In the field ofanthropometric dimensions prediction, some experts and scholars had put forward manymethods, such as the earliest proportionality constant method and some modified methodsbased on it, but there is a common problem that all the methods above can’t solved. Theproblem is that the differences between human body surface had not introduced into the prediction adequately, which will led to unsatisfactory on prediction accuracy andapplication effect. In order to solve this problem, a stepwise linear regression method wasselected in this thesis to predict anthropometric dimensions. To improve the accuracy ofprediction, we selected independent variables reasonably through correlation analysis andvariables classification firstly, and then got rid of correlation by creating new variables andprincipal component analysis. Finally, BMI and two main components of height and seatheight were selected as the main prediction factors, and prediction models were establishedfor the first layer variables. For some variables whose prediction effects are unsatisfactory,we made further prediction research by introducing the first level variables, which hadalready been predict in the last step, into their prediction models. Then the feasibility of thismethod was checked by comparing the deviations.Due to human body surface differ in thousands ways, if we predict the anthropometricdimensions using only one series of prediction models, it’s hard to avoid deviations causedby the differences between individuals, even if the prediction factors and methods areselected properly. In order to reduce the influence on prediction precision of anthropometricdimensions caused by the differences between human body, and improve the accuracy ofprediction, a method was put forward in this thesis that the large sample anthropometric datawere classified firstly, and then established the prediction models for each new samplesrespectively.The selection of classification variables is very important for finding a reasonableclassification scheme. In this thesis four variables were used for classification, which areBMI, Rss, Rwb, Rwh, obtained by constructing variables.All these four variables can be usedto describe the physical characteristic of human body. Aiming to improve the predictionaccuracy of anthropometric dimensions, the sample data was classified into6kinds of newsample using clustering analysis method combined by hierarchical cluster method andK-means cluster method. Then6discriminants were built to calculate the category ofindividuals, and the discriminants get accuracy of99.3%. Then prediction models had beenbuilt for each new samples respectively. In order to verify the effectiveness of this method, the predictive effect of the classified and unclassified predictive model was compared, whichshowed that the accuracy of anthropometric dimensions prediction can be improved bysomatotyping firstly.It is an effective way to describe human body using a parameterized human modelbased on anthropometric dimensions, and it is also a necessary and important supplementarymean for the design, valuation, analysis and research of man-machine system. To get sizesand shapes closed to the real human body surface, a parametric model of the human bodysurface was built using prediction model of the anthropometric dimensions and parametricmethod of knowledge engineering. This model includes two parts, casing ply and bodysurface. The casing ply has23parts, made up by points and lines, and it realized theparameterization of length and girth at the same time. The surface of this model wasestablished based on method of sections. There are15parts in the human body surfacemodel. Using images of human body slices in the process of modeling, the shape of thehuman body surface can be described more accurately. Then the parameterization ofanthropometric dimensions in girth was realized through building the scale parameters ofsections. The parametric model of the whole human body surface built in this thesis canrealize body size scaling and change human posture using the parameters. All that works laidthe foundation of establishing the parametric biomechanical human model.Finally, the parametric human model was used to accomplish a chest distanceevaluation of security for a car. Taking the three control variables as independent variables toget3principal components, and getting their values in5th,50thand95thpercentilerespectively,9human samples were selected as test models. Then put the models in the rightposition and posture, using the occupant posture prediction models established by Reed andothers. Eventually, the9human models were located in right place with reasonable postures.By measuring the minimum distance between the steering wheel and trunks of9models, wegot the evaluation result. The result shows that the distance could satisfy the requirementsfor safety.