Mutli-quality Robust Optimization Design For Switching Mode Power Supply

With the requirement of higher reliability, quality dispersion of SMPS has become the key factor that restricts the reliability and life improvement of electronic systems in our country. Therefore, how to ensure quality consistency and reduce individual differences has become a technical problem for manufacturers. However, the switching mode power supply simlulation rarely consider the electro-thermal coupling effect of power semidonductor devices, the the simulation precision is not high enough, and there is no approximation modeling method that can be directly used in robust optimization design. Besides, the current optimization methods are only foucus on initial quality, and never consider the uncertainty of degradation process, so it can not radically solve the quality consistency problem of SMPS in the whole life cycle.On the basis of related research status of the world, the robust optimization design(RDO) method is developed for multiple quality characteristics of power supply. Through optimization of key design parameters and tolerance, the quality requirements of each component is allocated reasonably to reduce the s ensitivity of output characteristics and stresses on key component to various uncertain factors, and then to improve the inherent quality. Moreover, it can provide theoretical guidance for quality control of production process.Firstly, the electro thermal coupling modeling and simulation method for SMPS are studied based on i SIGHT. The equivalent circuit model of flyback transformer is built, and the parameters are extracted by finite element simulation. The electro thermal coupling models of power Schottky diode and MOSFET are established to understand the change in component performance affected by temperature. With circuit simulation model and steady thermal simulation model, the electro thermal coupling simulation process is integrated in the i SIGHT platform. It can accurately describe the relation between output responses and design variables, and provide an important means for the follow-up study.Secondly, the RDO method based on sequential approximation modeling is presented to solve the multi-quality fluctuation problem. The sensitive factors causing the multiple quality fluctuation are determined via sensitivity analysis based on orthogonal test. After that, the test samples for approximation modeling are constructed by optimal Latin hypercube sampling. The applicability of radial basis function(RBF) approximation model is studied using RMSE and R2 as evaluating index. The optimal basis function and width coefficient c are then determined. The approximation models of multiple quality characteristics are established by cross validation sampling to meet the precision requirement. Then, the mean and tolerance of key design variables are obtained simultaneously by PSO, and giving an improved design of initial quality consistency.Afterwards, according to the requirement of total lifecycle RDO, degradation model based on the identified sensitive parameters is developed respectively. The random effects model accounting for measurement errors is proposed for linear degradation paths. The unknown parameters are estimated by maximum likelihood method based on EM iterative. And, the degradation model of aluminum electrolytic capacitor is established using this method. For nonlinear degradat ion paths, the nonlinear Wiener process method is put forward. The model parameters are estimated by MCMC simulation. The power MOSFET is taken as an example to illustrate the degradation process modeling in detail.Finally, combined with the component degradation models, the multi-objective RDO method based on performance degradation is present ed. The continuous state system model and time-variant performance limit state functions are established to study the performance reliability calculation approach ba sed on set theory. According to the calculated performance reliability, based on life cycle quality lost criterion, the RDO model is established for output characteristics of power supply. And the RDO model is also established for power MOSFET based on tim e to degradation failure. The design variables and their tolerance are optimized to improve the performance robust and consistency in the product life cycle.This research can improve the quality consistency design, analysis and optimization ability of domestic power supplies, and has important theoretical significance and practical value. The related key technology and ideas can provide reference for quality improvement of other electronic system.