Research And Application Of Online Robust Parameter Design Based On Kriging Model

From the perspective of modern quality engineering,variation is the fundamental cause of poor product quality.How to reduce product fluctuation has become the focus of attention in all industries.In order to enhance product quality and reduce costs,variations around the target ought to be effectively controlled.An effective approach to control variation is quality design,the core of which is parameter design or called robust parameter design(RPD).Traditional robust parameter design(RPD)is a useful tool for offline quality improvement,which is mainly applied to physical experiment.If the optimal setting for controllable factors is not desirable,the offline RPD cannot utilize the new-obtained sample to update the optimal setting.In practice,however,it is impossible to do physical experiments under certain circumstances,for instance,the climate phenomenon.The booming computer industry in the 21 st century allows these complex physical experiments to be simulated by using computer models(or codes).With the help of the computer experiment,engineers can study various things,for instance,the effectiveness of input to output,so as to study and apply robust parameter design more effectively.In this paper,the robust parameter design based on metamodel model is taken as the research object.The off-line RPD is extended to online RPD,and the background of RPD is extended from physical experiments to computer experiments by combining system modeling,simulation examples and empirical analysis.The online response surface modeling based on Kriging and online robust parameter design based on Kriging model and their applications are systematically studied.The main research contents of this paper are summarized as follows:(1)Based on the IKriging model,a recursive Kriging(RKriging)model is proposed.With the idea of fixed-length sliding window,this model adds a forgetting mechanism to the incremental model.By adding new samples and deleting old samples in the window,the model updates the data and keeps the length of the data interval constant.A simulation experiment and an empirical experiment are used to verify the model.The RKriging model proposed in this paper has great fitting and prediction ability on the basis of high modeling efficiency,and is more suitable for modeling actual processes.(2)An online robust parameter design method based on IKriging model is proposed.Unlike the traditional robust parameter design method based on offline Kriging model,in the proposed IKriging based online RPD strategy,an appropriate algorithm is designed to numerically get the next optimal setting for the controllable factors,and a criterion is established to determine the next site for the noise factors to observe the response.By employing the IKriging model,the response surface can be online constructed efficiently,and furthermore the optimal setting for the controllable factor can be obtained immediately,meeting the requirement of online quality adjustment.The proposed methodology is confirmed using four examples,which illustrate that the underlying optimal settings can be found by utilizing the proposed IKriging based online RPD strategy for the computer experiment,and the IKriging based RPD method is preferable to existing online robust design methods.(3)On the basis of RKriging model and the idea of robust parameter design,an online robust parameter design method based on RKriging model is proposed.In this method,an appropriate algorithm is designed to numerically solve the controllable factors,and the judgment criterion of the next observation response position of noise factors and the criterion of eliminating the original samples are established.Using RKriging model,the response surface can be efficiently constructed online,and the optimal setting of controllable factors can be obtained immediately,which meets the requirements of online quality adjustment.The effectiveness of this method is verified by four experiments.The results show that the optimal setting can be found by using the online RPD strategy based on RKriging.(4)In this paper,in order to study the practicability of the two online robust parameter design methods,the online robust parameter design method based on IKriging model and the online robust parameter design method based on RKriging model are applied to the research of decoded signal quality in color TV signal transmission.This paper studies online response surface modeling based on Kriging,online robust parameter design based on Kriging model and its application,which provides a new idea and method for robust parameter design.Finally,this paper summarizes the above research results,points out some shortcomings in this paper,and also puts forward some problems that can be further discussed in the follow-up.