Application of data mining-based affinity propagation clustering algorithm for diagnosis of mechanical equipment transmission system

Abstract

– Data mining techniques used to monitor and diagnose the faults of the transmission system of mechanical equipment, thereby promoting the development of big data analysis in the field of intelligent diagnosis. The Affinity Propagation (AP) clustering algorithm is commonly applied in cluster analysis techniques and used as a base to extract the fault information of helicopters. First, the data points are extracted randomly to build a dataset. Then, multiple feature definitions are utilized to extract fault information, which is imported into the AP algorithm for the construction of an early fault clustering effect diagram to judge the early faults of the transmission system and the characteristics of the traditional AP algorithm. Second, the Euclidean distance is weighed negatively; the similarity of the negative Euclidean distance metric is improved and applied to the traditional AP algorithm. The method mentioned above also randomly extracts data points to extract feature information, thereby building a multi-fault diagnosis method for helicopter transmission bearings. Finally, by integrating online and offline data analysis, a fault diagnosis treatment plan for the overall big data is proposed. The results show that, for the early diagnosis of slight changes in data, the application of the AP algorithm can reduce the data dimension and complete data division; however, the separation of data is not ideal, and data overlap occur. After improvement, the differences between the calculation dimension can be presented, while the original characteristics of the data are retained. The effectiveness and resolution of the AP algorithm are improved. The analysis of online and offline data is combined to make the final diagnosis result more accurate and reliable. Reduced number of calculations provide accumulation and support for the fault diagnosis of the transmission system in the future. © 2020, Cefin Publishing House. All rights reserved.