Abstract: Based on dimension reduction algorithm, a novel car-body reliability optimization model is established, with minimizing car body mass as the optimization objective and the reliability degree of bending and torsional stiffness of car body as constraints. By using dimension reduction algorithm, a new model with n one-dimensional functions is established to approximate the original function. With the aid of Taylor series and statistical moment theory, the first four-orders of moment of the car body function are solved, and by using the fourth order moment method, the reliability degree of car-body is obtained. By means of improved adaptive genetic algorithm-based optimization scheme and reliability assessment, the nested optimization problem is decoupled into a single-layer optimization problem, the penalty function method is used to modify and improve the fitness function. Compared with traditional reliability-based optimization methods, with this method, neither using multiple integration to solve the statistical moment of functional function nor iteration for searching the most probable failure points are required, and can also avoid the defect of the population falling into local optimum. The results of numerical calculation show that the method proposed has higher accuracy and better adaptability.

Key words: car body lightweighting, reliability-based design optimization, dimension reduction method, fourth-order moment method, improved adaptive genetic algorithm