An Online Predicting Method for Product Quality Based on Energy Consumption Data Using Random Forest

An effective real-time product production quality assessment method is of great significance for production manufacturing workshops to increase production capacity and enhance enterprise competitiveness. In this study, a workshop industrial product production quality prediction method based on the DeepAR-Random Forest combined model is proposed based on industrial big data. This method first uses the autoregressive recurrent neural network (DeepAR) algorithm to predict real-time energy consumption of equipment, and then uses multi-source heterogeneous data fusion technology to model the production data of the manufacturing execution system (MES) and the predicted energy consumption data as data vectors. Based on Random Forest, the product quality condition is assessed online. Finally, based on data collected from the automotive engine valve production line, this study compares Random Forest and commonly used classification models like boosted regression trees and artificial neural networks, demonstrating that this design better solves the problem of product quality prediction, overcoming the limitations of human error and statistical lag in manually statistical product quality, and providing new ideas for the engineering application of subsequent artificial intelligence algorithms in production workshops.