Robust Parameter Design Of Functional Responses

As an important method of quality improvement in modern manufacturing,robust parameter design is mainly applied in the design stage of products or processes.By adjusting appropriate process parameters and determining the level of controllable factors,the purpose of reducing system fluctuations is achieved.In the past,most of the robust parameter designs were for static response or multi-response.With the increasing complexity of the production process,in the manufacturing and processing of some complex products in the fields of engineering and medicine,the quality characteristics of the product or process usually show a certain quality.This kind of specific functional relationship,this kind of quality characteristics with functional relationship is called functional responses.At present,most of the research on functional responses stayed in the product quality control stage,but relatively few researches on the robust parameter design of functional responses.As a new type of response,functional responses are accompanied by measurement technology and complex product manufacturing With the rapid development of the industry,a new type of robust parameter design problem arises.In view of this kind of problem,if the traditional robust parameter design method is continued to solve,it may cause insufficient model fitting,only suboptimal results can be obtained,and the optimization results cannot be optimized.Because traditional robust parameter design methods generally assume that a single model structure or model parameters are known or can be accurately estimated,most studies ignore the impact of model uncertainty on robust optimization results.The research focus of this paper is to optimize the functional responses by using the optimization method in robust parameter design theory in the context of model uncertainty.The specific research content is as follows:(1)Aiming at the correlation of functional response and the change of response target value with time,this paper uses SUR regression method to estimate the model parameters of functional responses at different time points,and then uses MCMC method to analyze the data on the basis of sample data.Sampling iterations and statistical analysis of the quality characteristics of the functional responses using Bayesian methods.Finally,a method of simultaneously optimizing the process mean and variance of the functional response is used to obtain the best level combination of controllable factors.The research results show that: based on the SUR regression method,the Bayesian posterior introduced not only ensures that there is no strong correlation between the responses of the constructed models,but also it can maintain a good predict performance under the condition of increasing system fluctuations,thereby ensuring the accuracy and reliability of optimization results.(2)Aiming at the problem that the functional response is a linear contour,based on the linear contour layered model,this paper first builds a Bayesian generalized linear model of functional response from a limited sample data from the perspective of Bayesian statistics..Then based on the obtained model,comprehensively consider the robustness and optimality of the process,and adopt the robust optimization satisfaction function that assigns robustness and optimality weight as the optimization index,and optimize the parameters by genetic algorithm.The research results show that the proposed method can make the result of the functional response variable closer to the target value while considering the uncertainty of the model parameters.Finally,on the basis of summarizing the research results of this paper,the advantages and areas for improvement of the method proposed in this paper are expounded.At the same time,the next research ideas and expansion directions are prospected.