Self-Tuning Intelligence Digital Twin for Bearing Pattern Recognition

Abstract

Induction motors are consumed around 80% of energy in heavy industries, that approximately 20% of this energy consumption is because ofmechanical failures. Moreover, the bearing failure with about 69% is the principal constituent of mechanical defects. In this study, the self-tuning intelligence digital twin is presented for bearing pattern recognition. The self-tuning digital twin is designed using the combination of knowledge-based models, physical models, and linear dynamic Bayesian estimator. In the physical-based approach, the mathematicalbased vibration signal modeling is used. To reduce the effect of unknown states, the knowledge-based technique based on an adaptive network-based fuzzy algorithm is recommended. To evaluate the signal estimation, the linear dynamic Bayesian estimation using the proportional-integral technique is endorsed. To classification, the bearing signals and pattern recognition the machine learning approach is recommended. Regarding the results, the bearing signals pattern recognition accuracy is 98%.