Fault Diagnosis of Rolling Bearing Based on Knowledge Graph With Data Accumulation Strategy

The fault diagnosis of rolling bearing plays an important role in ensuring the safe and stable operation and maintenance of rotating machinery. Traditional bearing fault diagnosis methods fail to consider the correlation between faults and characteristics and do not take full advantage of the ever-increasing monitoring data. Thus, a bearing fault diagnosis framework based on knowledge graph (KG) and data accumulation strategy is proposed. First, the entities of the KG are defined based on multiple features extracted from the time domain, frequency domain, and time–frequency domain of bearing vibration data collected by the vibration sensors. Then, the feature-fault correlation as the edges is designed and calculated to establish a KG framework together with the entities. In addition, a weighted random forest algorithm is proposed as a reasoning algorithm for the KG, making full use of the feature-fault correlation to improve the accuracy of bearing fault classification. Finally, a data accumulation strategy is designed to continuously increase the size of the training dataset of KG. Relevant parameters are updated in the process to make the results produced by the inference algorithm more accurate. The advantage of the proposed method was demonstrated by a comparison with several models for the same circumstance. The test results showed that the proposed method was promising and it had good prediction accuracy and robustness for different working conditions.