Quality Prediction Of Car Body Production Process Based On CEEMDAN And XGBOOST

With the arrival of Made in China 2025,the scale of the digital workshop of the automobile manufacturing industry is gradually expanding.As the body-in-white is an important part of the automobile assembly,accidents caused by quality problems will inevitably attract the attention of the enterprise department,and its production process is also strictly required.The impact of the problem is increasing.The precise control of the body assembly in the multilevel manufacturing system is one of the problems that the automotive industry has not effectively solved for a long time.This article takes the production process quality management of manufacturing enterprises as the background.Aiming at the traditional machine learning methods in the automotive multilevel manufacturing system,the processing of multiple data samples is long and the accuracy is low.Based on big data analysis and combined with advanced machine learning technology,multilevel manufacturing The body assembly in the system is predicted,and the main research contents are as follows;Firstly,a feature extraction method based on CEEMDAN is proposed.On the basis of empirical mode decomposition(EMD),the construction principle and concept of complete empirical mode with adaptive noise are proposed.The data signal is affected by the external environment such as the size of fixture or parts,and the fault signal component will have obvious and unbalanced random performance.This paper analyzes the advantages of the complete empirical mode decomposition method based on adaptive noise in processing data signal,and provides a feasible means for processing such non-stationary data.Compared with EMD and EEMD,CEEMDAN is proved to be effective and superior in solving the problem of aliasing mode and over decomposition of data.Secondly,the intelligent prediction of quality in the production and assembly process of body-in-white size deviation is proposed.First,through the analysis of the multilevel assembly process of the car body,the preprocessing of many data samples,the establishment of the absolute correlation matrix of different characteristic elements based on the Spearman coefficient,and the use of all data mining analysis of the productionprocess,propose the data analysis process and data Based on the processing framework,an intelligent prediction model of car body size assembly quality based on XGBoost is established to realize precise control of car body size assembly,and accurately and quickly predict abnormal data in car body assembly.For the proposed model,an AUC＿ROC-based Evaluation plan,training and evaluation of the model.Starting from theoretical research and practical applications,the quality of the production process of the entire vehicle body is predicted,and the feasibility and efficiency of the XGBoost algorithm are verified through a comparative analysis with traditional machine learning algorithms.Finally,the case is simulated and analyzed.Starting from the case study,based on the vehicle body manufacturing quality data detection and management platform,the storage and operation analysis system of body manufacturing process dimension data is constructed.Based on the analysis of the key processes and mass data generated by the initial system in the body production process,the development environment conditions and overall structure of the initial system are used,and the case company is taken as an example to analyze the specific caseIn summary,this article proposes an intelligent body size prediction model based on CEEMDAN and XGBoost algorithms.On the one hand,it can solve the problems of large hierarchical data base,large difference data,data noise and instability in the automobile production process;on the other hand,It can realize the precise predictive control of the body size assembly and provide technical support for the stable operation of the automobile production line.