Research On Cost Rule Mining And Cost Forecasting In The Part Design Phase

With the maturity of data mining technology and the growth of manufacturing data,manufacturing companies rely more on mining and using knowledge from large amounts of data to assist design decisions.In the life cycle management of parts,the part design stage determines the cost of 75%of the full life cycle.Controlling the cost in the design stage has become the key to ensuring the competitiveness of manufacturing companies.This article focuses on the research of cost rule mining and cost prediction in the part design stage,and designs and develops the corresponding prototype system,which can analyze the cost in the part design stage,assist the designer to work,reduce the cost of the manufacturing enterprise,and improve the efficiency of the enterprise.First,to address the problem of incomplete cost data in a single-system database of an enterprise,a cost data warehouse was constructed at the part design stage.Clarify the design requirements and characteristics of the cost data warehouse,analyze the part design data in the information system of the manufacturing enterprise,use structured data extraction methods,obtain the cost data source of the design stage from each system,and design and build the cost data warehouse.Secondly,in view of the problem of outliers and missing values in the enterprise cost data,the source data in the cost data warehouse is cleaned.Isolate forest(iForest)outlier detection algorithm is used to remove the abnormal samples of cost data.At the same time,a nearest neighbor filling(KPKNN)algorithm based on K-prototypes clustering is proposed to fill in the missing data of cost and solve the problem of cost data cleaning in the part design stage The example proves that the algorithm can clean the data.Thirdly,for the problem of redundancy of cost rules in the part design stage,a cost rule mining algorithm based on improved FP-Growth in the part design stage is proposed.By researching the principle of FP-Growth algorithm,improving the generation process of conditional pattern base,constructing a mining model of cost rules in the part design stage based on improved FP-Growth to obtain cost rules in the part design stage.An example is used to prove that the algorithm in this paper can reduce the generation of redundant rules and shorten the running time.Then,for the problem of inaccurate cost prediction in the part design stage,a cost prediction model for the part design stage based on GA-BP neural network is constructed.Research the BP neural network based on genetic algorithm,construct the cost prediction model based on GA-BP neural network in the part design stage,and realize the part cost forecast.Case analysis shows that GA-BP neural network has higher prediction accuracy and is more suitable for cost prediction.Finally,based on the research content,this article uses Microsoft Visual Studio2019,SQL Server 2017,Anaconda3 to design and develop a prototype system.