A deep learning driven method for fault classification and degradation assessment in mechanical equipment

Mechanical degradation may cause equipment to break down with serious safety, environment, and economic impact. Since the equipment usually operates under a tough working environment, which makes it vulnerable and increases the complexity of fault diagnosis. Simultaneously, the requirement of manufacturing systems with reliable self-assessment has been increasingly raised with the trend of smart industry. The aim of this paper is to fill this gap by providing a deep learning driven method for fault classification and degradation assessment. Compared with conventional data-driven methods, deep neural network has the ability to learn multiple nonlinear transformation with high complexity through multiple hidden layers, which helps to capture the main variations and discover the discriminative information from the industrial data. During the experiment, to confirm the effectiveness of deep learning for fault classification and degradation assessment, similar popular data-driven methods, including support vector machine, deep belief network, back propagation neural network, and k-nearest neighbour classification are employed to present a comprehensive comparison in both fault classification and degradation assessment. According to the numerical results, the proposed method outperforms the other conventional approaches and demonstrate its superiority in degradation assessment for mechanical equipment.