A big data mining approach for environmental emissions prediction of die casting process

Abstract

As the key component of the manufacturing industry, die casting process suffers severe emission problems, which disobeys the prospects of Industry 4.0 and green manufacturing. Environmental emissions prediction of the die casting process is the foundation to optimize the environmental impact factors to achieve lower environmental pollution and health hazards of workers. However, the physical causal relationships between influential factors and environmental emissions are hard to be modeled. As a result, the environmental emissions prediction of the die casting process still faces a huge challenge. The big data mining approach can obtain the relationships disregarding the physical causal effect. This study firstly analyzes and demarcates nine impact factors that influence the four major environmental emissions PM2.5, PM10, noise, and temperature in the die casting process. Then, a big data mining approach integrating principal component analysis, particle swarm optimization, and back-propagation neural network is proposed for environmental emissions prediction. The effectiveness of the proposed approach is verified through a case study of die casting islands. The results show that the accuracy of the proposed method for the prediction of four environmental emissions and six types of die casting islands exceeds 90%. This proposed approach can strongly support the optimization of process parameters, layout of die casting island, etc. toward environmental emission reduction. It can also be easily applied to other processes such as plastic molding, machining, and extruding for emissions projection.