Hydraulic pump health data analysis methods that can detect failures using pump pressure signals were investigated. A reliable method could easily be incorporated into an on-line health monitoring system for hydraulic pumps. Two pump health data analysis methods were implemented and evaluated for efficiency and reliability. The two methodologies used in this research were fast Fourier transforms (FFT) and wavelet packet analysis. Each method was validated using two sets of pressure data from hydraulic piston pumps operating under normal conditions. One set of validation data was collected from healthy pumps while the other from defective pumps (pumps outputting only 90% of original flow rate). The models were evaluated on their overall accuracy, missed alarm rate, and false alarm rate. Wavelet analysis of hydraulic pump pressure data was 90% accurate; whereas, FFT was only 67% accurate. Wavelet analysis of pressure signals can be used effectively in an on-line health monitoring system for hydraulic pumps.
展开▼