The multisensor information fusion-based deep learning model for equipment health monitor integrating subject matter expert knowledge

Nowadays, the modern production machines are usually equipped with advanced sensors to collect the data which can be further analyzed because of the advent of Industry 4.0. This study proposes a novel deep learning (DL) information fusion-based framework collaborating convolutional neural network (CNN) architecture with subject matter expert (SME) for equipment health monitor. The author integrates the unsupervised learning with supervised learning strategies bringing several benefits. Unsupervised learning assists in identifying the underlying patterns and relation within data without the need for labeled data, while supervised learning trains the model by the labeled data to derive prediction results. Also, due to sensor data characteristics, this study develops the independent CNN-based backbone net to extract the features of multisonsor data and to allow the proposed architecture to flexibly adopt arbitrary number of sensors attached to the equipment. An empirical study is conducted to demonstrate the effectiveness and the practice viability of the proposed framework. The resulting outcomes show that the proposed algorithm has superior performance than other machine learning models. One could adopt the general framework to maintain the performance of the equipment.