Hydraulic Unit Vibration Monitoring Using Machine Learning

It is required to assess the hydraulic unit current state, if it is serviceable or unserviceable, as per its vibration monitoring data. This is the task of technical diagnostics, i.e. to determine the system state under the conditions of limited information, obtained during its operation. Machine learning can be utilized to solve this problem. During the investigation the cross-validation was applied for different variants of original samples drilling-down, plotting different classification models, and selection of significant exponents. Taking into account the observation data there was developed a program of technical objects automated diagnostics. The experiments showed that to assess the hydraulic unit serviceability one may utilize aggregated classifiers, which provide the best accuracy in comparison with classification base methods.