Research On Evaluation Method Of Creation Design Scheme Based On Clustering And Deep Belief Network

Schemes evaluation is the key to determining the product design direction in the product design process.With the appearment of "Internet plus" concept,the creation design model has developed rapidly.Compared with the traditional design scheme,creative design scheme has the characteristics of large quantity,variety of expression forms and high similarity,which makes the evaluation of it very difficult.It is more and more important to quickly evaluate the design scheme of Zhongchuang.In this paper,the modeling method of creative design scheme is studied.The deep belief network model is applied to the creative design scheme evaluation to improve the evaluation efficiency in the scheme scoring process.Considering the distribution law of the data of creative design scheme,clustering-based evaluation method for creative design schemes,further improve the evaluation efficiency in the process of training set construction;developed a complex customized equipment creation design platform,and used the injection molding machine injection parts of the innovative design evaluation example to verify the effectiveness of the method.This article is divided into six chapters,the main contents are as follows:The first chapter introduces the development of product design and the research status of schemes evaluation at home and abroad,expounds the research background of this paper and gives the main research content and organizational framework of this paper.The second chapter analyzes the creative design schemes and gives the method of calculating the weight of the evaluation index.The performance form of creation design scheme is analyzed,and creation design scheme is modeled by evaluation index.The affiliation of evaluation index is considered,and the weight distribution of each evaluation index is calculated by analytic hierarchy process.Uing the 9-segment method optimizes the evaluation process,and proposes a weighted summation method for the scores of the creative design scheme to quantify the score results as the ranking rule of the creative design scheme.The third chapter constructs a deep belief network model for the evaluation of the creative design.The disparity standardization method is used to preprocess the dataset of creation design schemes,and the data dimension of each evaluation index is eliminated to conform to the input rules of the model.The number of hidden layers of the deep belief network and the number of hidden layer nodes are determined in the experimental flow.The weight initialization method and the learning rate momentum of the network structure are given.The model is pre-trained by the contrastive divergence method,and the backpropagation algorithm is used for tuning,and the optimal model structure is obtained.The fourth chapter proposes a method for evaluating the creative design scheme using clustering.The k,medoids clustering method is used to cluster the creative design schemes,and the optimal classification number k is determined according to the variation of sum of the squared error of the clustering;In the clustered results,k central point elements are extracted as the feature scheme.Carry out manual evaluation and obtain evaluation rules;analyze the value characteristics of evaluation indicators,construct the training set of deep belief network evaluation model by data simulation method,optimize the existing training set construction method,and improve the overall evaluation process of creation design scheme effectiveness.The fifth chapter introduces the complex design equipment based on the theory and method of this paper,and uses the injection molding machine injection parts as an example to verify the effectiveness of the theory and method.The sixth chapter summarizes the main research work and results of this paper,and on the basis of this,the future research content is prospected.