Application of machine learning algorithms on the multi-feature multi-classification problem - in the case of a hydraulic system

• Main Authors: Liang, Yun-Chia, Zhan, Xin
• Format: Conference Proceeding
• Language: English
• Subjects: Algorithms; Artificial neural networks; Classification; Decision trees; Fault diagnosis; Hydraulic equipment; Hydraulics; Machine learning; Neural networks; Support vector machines
• ISSN: 0094-243X; 1551-7616
• DOI: 10.1063/5.0106796

Nowadays, the complex machinery system is getting more complex due to its structure and components. Identifying the status of components so that the proper maintenance can be arranged in advance has become an essential and challenging issue. In order to find a more suitable algorithm to solve the multi-feature multi-classification problem, this paper takes the fault diagnosis of a complex hydraulic system as an example. Three Machine Learning algorithms, such as Decision Tree (DT), Support Vector Machine (SVM) and Artificial Neural Network (ANN), are used for fault diagnosis, and the results are compared and analyzed. There are many kinds of fault diagnosis problems in this cooling and filtering hydraulic system. The minimum characteristic number is 17, the maximum characteristic number is 1,326, the minimum classification number is 3, and the maximum classification number is 144. Among the three algorithms, the DT algorithm has the best overall performance and the most stable performance. In addition, artificial neural networks provide an excellent performance when fewer features are used.