A Multi-sensor Big Data fusion Method in Quality Prediction of the Plasma Enhanced Chemical Vapor Deposition Process

Abstract

Plasma Enhanced Chemical Vapor Deposition (PECVD) is a critical process in the processing of solar cells. Large quantities of the process data are collected from different sensors during the PECVD process, which are high-dimensional and highly correlated. Most existing research only focus on the analysis of single sensor data instead of multi-sensor data. However, the information contained in single sensor data is incomplete. In this paper, the method of Convolutional Neural Networks (CNN) is adopted to analysis multi-sensor big data form PECVD process. The regression model between the multi-sensor data and the quality of solar cells is established for quality prediction. The impact of various types of hyper-parameters on the performance of the model is analyzed, and the predictive performance of the model is optimized by adjusting the hyper-parameters. The performance of the proposed method is compared with existing methods in a real-world case study.