Optimization Of Anti-Whiplash Seat Based On Neural Network And Genetic Algorithm

There are four kinds of impact form when automobile crashes happen. They are frontal impact, side impact, rear impact and rollover. Only in recent years, rear impact is getting increasingly attention from people. During the rear impact, occupant’s head and neck are injured due to the sudden change of force and torque. This kind of injury is called whiplash injury which may cause the occupant a lifelong pain. In 2012, the research center of Chinese Automobile Technology published a new C-NCAP rule which added an experiment of neck protection during the rear impact at low velocity (whiplash test). By now the rear impact is becoming an important part of the automobile passive safety. This paper uses the MADYMO software to build the constraint system of occupants with the safety seat. And by combining the neural network and the genetic algorithm a optimization is done to find out the best design parameters for the constraint system of occupants.First, this paper analyses research contents and overseas and domestic research status of rear impact. Second, explains occupant neck injury criteria in rear impact and rear crash regulations from country to country, and choose neck injury criteria NIC and Nkm as evaluate standards in this paper. Third, uses MADYMO to create a rear impact safety seat model, which includes the car body, safe-belt and BioRID II type dummy, definition of the contact characteristics, boundary conditions and initial conditions from which a calculation is done. After validating the simulation with real vehicle test to ensure its accuracy, export acceleration, force, torque, NIC and Nkm to analyze the dynamic response of rear impact. Forth, analyses the influence of various parameters on neck injury, and on the base of the sensitivity analysis, decides design variables by using sensitivity screen and range analysis. Then design orthogonal experiment of decided parameters, as input data in optimization on the one hand, finding primary and secondary order of the parameters on the other hand. Finally, combines BP neural network and genetic algorithm to optimize the parameters of the seat, and analyses optimization results with original set.There are several innovations in this paper. First, by classifying seat parameters according to type and characteristics, avoided the current research disadvantages which did not consider the influence between same type parameters. Second, by combining neural networks and genetic algorithms, an overall search for the optimal solution is done within avoiding the discrete defects of traditional multi-objective optimization. Third, with the help of the Matlab Toolbox and the programming skill, a combination of neural network and genetic algorithm is set up which can analyze the sample data and find out the best solution with multiple factors effectively and accurately.