Research On Design Scheme Evalution And Structural Reliability-based Robust Design

Conceptual design has played a decisive role in the products design. As there are various design schemes produced from the conceptual design, how to select the best scheme out of a large number of design schemes to effectively avoid the effect of subjective factors on the assessment results in the evaluation process will have important guidance to practical engineering. The important role of design theory and method in the product development inspired people to the attention of the design, so that new theories and new methods of the reliability design are emerging. Because the reliability index of mechanical structure generality decreases with uncertainty factors, such as material properties, geometrical parameters, working conditions and so on, it is necessary to consider the reliability of structure, sensitivity and robust design. Therefore, establishing reliability-based robust design model is much closer to the actual working condition and has important theoretical significance and practical value to determine the reliability information of the mechanical structure.First of all, combining with axiomatic design, analytical hierarchy process and gray clustering theory, scheme evaluation model is constructed based on gray clustering and information axiom integration. In the model, qualitative fuzzy information indicators is expressed quantitatively based on grey clustering theory during the process of the information axiom evaluation; Evaluation index system is constructed in a reasonable manner using the analytic hierarchy process. Qualitative indicators is whitened using the triangle whiten function to solve the disadvantages that information axiom cannot evaluate the shortcomings of qualitative indicators. After improving the formula of information volume and extending the design scope and system scope, the best scheme is determined through comparing the total information volume of all alternatives. The method effectively overcomes the disadvantages of hard to computing in complex scheme of information axiom and avoids the effect of subjective assume on the evaluate results, so it is of greater applicability.Secondly, saddlepoint estimation techniques can be used to be infinitely close to the characteristic of linear limit state function of the probability distribution in the non-normal variables space characteristics and can effectively solve the problem that the distribution regularity is difficult to be determined because of less statistical data or the experimental data. Based on the mathematics model of reliability, reliability sensitivity formula based on saddle point estimation is deduced systematically and mechanical parts reliability sensitivity calculation problem is discussed based on saddlepoint estimation method with the combination of reliability design theory, sensitivity analysis technology and saddlepoint approximation theory. Based on saddlepoint estimation, the reliability sensitivity analysis method proposed in this paper lays a theoretic foundation for further analysis of reliability-based robust design of mechanical part.Thirdly, reliability optimum design model is constructed on the basis of the stochastic perturbation theory in consideration of multiple failure modes in practical engineering optimization problems. On the one hand, from the point of view of the sensitivity of various design parameters caused by reliability, we propose reliability-based robust optimization design method of machinery parts through controlling the sensitivity of the mean and variance caused by reliability. In this method, reliability analysis is by complete distribution information of limit state functions to avoid the mazy iterative optimization process and improve the efficiency and precision of the reliability-based robust design. On the other hand, through the sensitivity analysis of objective function and constraint function, additional items of sensitivity of objective function and constraint function are generated. Then we propose two kinds of reliability-based robust optimization design model on the basis of additional objective function of sensitivity. The model breaks the limitation of reliability-based robust optimization design of machine parts which are achieved through controlling sensitivity of the mean and variance caused by reliability. It provides a theoretical basis for the level of reliability and robustness of mechanical components.Finally, on the basis of random sampling in the big sample, mechanical structural strength reliability of numerical simulation is taken combining Monte-Carlo method with MATLAB. By the means of contrasting sensitivity analysis method based on the saddlepoint and the Monte Carlo method, the correctness and effectiveness of the sensitivity analysis technology based on the saddlepoint approximation is validated. Using MATLAB matrix calculation and image forming, the effects of the random variable change on the failure probability of the mechanical structural strength and the sensitive parameters whch have a great effect on the failure probability are visually displayed. Based on sensitivity analysis of the saddle point approximation method and the monte carlo method, probability density function of structural response under compressive stress is simulated to further illustrate the correctness and effectiveness of the reliability-based robust design method based on the saddlepoint approximation.