Parameters Optimization And Design Of Experiment For Complex Relationship Processes Considering Prior Knowledge

Design of experiments is the main method of parameter optimization.In recent years,there are more and more complex relationship processes because of the improvement of manufacturing technology.Traditional design of experiments for the process emerged many problems: the high cost or insufficient or lacking of representative.Therefore,it is urgent to improve the experimental design to make it suitable for complex relationship processes.In this way,we can optimize process parameters and improve the quality level.In fact,prior knowledge can provide information about the process.The knowledge can be used to guide distribution of sample points,so as to optimize design of experiment.On the other hand,it can be also used as the initial points of the optimization method.Based on this,the paper presents a method of design of experiments and parameter optimization considering prior knowledge.First of all,there are the brief introduction of the research background and significance,analysis of the current research methods and putting forward the innovation of the paper.The principles and steps of the method used in this paper are followed.Secondly,we imply prior knowledge into design of experiments.The D-S evidence theory based on the similarity coefficient of evidence is chosen to fuse the prior knowledge,and the result is used to guide the design of sample points' distribution.The method divides design of experiments' Parameter range into sub region and optimizes the traditional experimental design improving the effectiveness of the sample to the modeling.Thirdly,the characteristics of the complex relationship processes are analyzed corresponding to the requirements of modeling optimization method for complex relationship processes.Meanwhile,we compare the different methods.In the light of result,we introduce a method that suitable for the process modeling.A large number of studies have shown that the least squares support vector machine has the best ability of fitting and forecasting for small samples modeling.Genetic algorithm has a strong ability to search optimal values for complexrelationship processes.Finally,the method proposed in this paper conducts simulation and the empirical of injection molding.For the simulation,values of the predictive ability indexes Sep,MaxE and StdE of the model constructed by proposed sub-regional design of experiments and modeling method are average reduced by 47.7%,28.20% and71.80% respectively.For the empirical study,the proposed method is more accurate and suitable than the traditional method in optimization parameter result.The paper proposes the thinking,the realization methods,the procedure about the design of experiment and modeling optimization of complex relationship processes.The results show that the method is more suitable for the current process than the traditional method,which has significant theoretical significance and practical value for the improvement of the quality level.