A Method of the Vibration Information Detection for Rotating Machinery Based on the Rolling-Shutter CMOS and Digital Image Processing

Abstract

It is crucial to develop precise, fast, cost-effective, and reliable technology for monitoring the condition of rotating machinery and diagnosing faults to ensure industrial production efficiency and safety. The study investigates non-contact vibration information acquisition and fault diagnosis methods. In response to the current reliance on high-speed camera hardware for visual vibration measurement, a method for visual vibration-based information acquisition and fault diagnosis is proposed. This method integrates the imaging characteristics of rolling shutter CMOS with digital image processing technology to significantly increase the sampling frequency of one-dimensional horizontal vibration information for low-cost cameras. A vibration information acquisition system is established to conduct multiple experiments on 42 vibration states. The results were consistent with theoretical values. The maximum relative error did not exceed 1.36%, and the relative standard deviation ranged from 0% to 0.6%, demonstrating the system’s accurate and stable acquisition results. Comparative analysis of the diagnostic results using the K-Nearest Neighbor, AdaBoost, CatBoost, and Random Forest algorithms revealed that the Random Forest algorithm achieved the highest diagnostic accuracy, exceeding 98%. The results identified the Random Forest algorithm as the most suitable for fault diagnosis based on vibration information in this work.

Keywords

• Visual vibration measurement
• rolling shutter CMOS
• digital image processing
• fault diagnosis
• machine learning