Hybrid Degradation Equipment Remaining Useful Life Prediction Oriented Parallel Simulation considering Model Soft Switch

Abstract

Equipment parallel simulation is an emerging simulation technology in recent years, and equipment remaining useful life (RUL) prediction oriented parallel simulation is an important branch of parallel simulation. An important concept in equipment parallel simulation is the model evolution driven by real-time data, including model selection and model parameter evolution. The current research on equipment RUL prediction oriented parallel simulation mainly focuses on a single continuous degradation mode, such as linear degradation and nonlinear degradation. Under this degradation condition, the model parameter evolution methods in parallel simulation can effectively predict equipment RUL. However, in practice, most of the equipment degradation processes exhibit a mixture of continuous degradation and discrete shock. So this requires adaptive selection of simulation models based on real-time degradation data. In this paper, the hybrid degradation equipment RUL prediction oriented parallel simulation considering model soft switch is studied. Firstly, under the modeling framework of the state space model (SSM), two kinds of degradation simulation models are established using the Wiener process and Poisson effect. Driven by the real-time degradation data, the model probability is calculated by using the forward interactive multiple model filtering algorithm to realize the model soft switch and data assimilation. On the basis of model soft switch, the expectation maximization algorithm is utilized to achieve model parameter evolution. Through the iteration between model soft switch and model parameter evolution, the simulation fidelity can be effectively improved and the actual equipment degradation state is continuously approached. According to the full probability theorem and the concept of first hitting time, the simulated degradation state distribution is integrated into the inverse Gaussian distribution. Then the analytical expression of the RUL probability density function is obtained to achieve RUL real-time prediction. Finally, a case study was conducted by using a bearing degradation data. The results show that the parallel simulation can effectively model the hybrid degradation process of the bearing. Compared with the single-model method that only considers the model parameter evolution, the RUL obtained by the method proposed in this paper has higher prediction accuracy and smaller uncertainty.

A Novel Method for Remaining Useful Life Prediction of Roller Bearings Involving the Discrepancy and Similarity of Degradation Trajectories

“…To verify the performance of the proposed method, several state-of-the-art RUL prediction methods are compared, including Sutrisno's vibration frequency signature anomaly detection and survival time ratio [ 13 ], Hong's combinatorial feature extraction and self-organization mapping [ 17 ], Guo's recurrent neural-network-based health indicator [ 24 ], Singleton's extended Kalman filter-based method [ 46 ], Zhu's multiscale convolutional neural network-based method [ 22 ], Cheng's transferable convolutional neural network-based method [ 29 ], Mao's deep feature representation and transfer learning [ 30 ], and Li's deep adversarial neural networks-based method [ 33 ].…”

“…The main way to perform an AI-based framework for bearing RUL prediction is machine learning (ML) based prognostics, including artificial neural networks (ANN) [ 9 ], support vector machines (SVM) [ 10 ], random forests (RF) [ 11 ], and deep learning (DL) [ 12 ]. The conventional machine learning-based prognostic methods usually extract a single statistical feature in time or frequency domain from the original signal as a health index (HI), such as root mean square [ 13 ], Kurtosis [ 14 ], energy entropy [ 15 ], and so on. However, there are significant limitations in the characterization capability of such single statistical features.…”