Damage degree prediction for rolling bearing

Rolling bearings are important components of pump and electromotor, which are widely used in nuclear plant system. Since the environment and condition changing in long term operation, the surface fatigue and lubrication gradually deteriorate, which would cause defect damage in the contact surface of bearing and affect the stability and safety of devices. The influence caused by early weak defect is limited, but the early weak defect would spread and deteriorate, which would cause serious hidden trouble. Therefore, predicting the trend of damage degree according to the early weak defect is extremely essential, and it is also conducive to the device maintenance in advance. In this study, a prediction method based on support vector regression is proposed for rolling bearing damage degree. Aiming at the life cycle data of bearing, several techniques are applied to evaluate the status information of bearing signal acquired from different periods of life cycle data. Then, the most sensitive feature information is adopted as the training data of support vector regression. The trained support vector regression can be applied to predict the trend of damage degree in near future with acceptable deviation. The prediction method can provide an early failure warning for bearings in operation conditions; it can make plenty of time to take maintenance measures before the fault deteriorates. (author)