The heart of the hydraulic system pump, the hydraulic system fault diagnosis is an important part of troubleshooting. Since the compressibility of the fluid, the pump source and servo systems and hydraulic fluid coupling itself has substantial natural mechanical vibration, making the failure mechanism of hydraulic pump complex, difficult to fault feature extraction, fault diagnosis fuzzy strong. A large number of pump diagnostic data show that exports through the pump source of the fault signal is often detected interference signal drowned, the single fault detection signal often has emerged in the ambiguity of strong, using conventional signal processing methods can not improve the fault characteristics effectively.
Piston loose ball for the failure mode, the export allocation through the pump and pressure sensor, vibration sensor fault detection, signal analysis by wavelet denoising, principal component analysis of effective integration of information extraction, an improved algorithm of BP neural network weak signal or pump failure to achieve a valid diagnosis.
1, the hydraulic pump failure mechanism of loose ball
Because of manufacturing errors or hydraulic pressure in the impact of the work process, often to the ball plunger ball and socket head and the concave deformation of the ball Shen ball and socket head and the gap increases, resulting loose ball plunger failure.
2, the wavelet signal denoising
Hydraulic pump generally poor work environment, the working conditions influenced by the environment, usually in the pump outlet to the detected signal contains a lot of noise. Tests showed that the export pump pressure signal is detected and the vibration signals reflect the following characteristics: ① distribution of a wide spectrum of the signal, waveform messy, bad laws; ② and non-stationary time-varying performance significantly. Therefore, based on the fault characteristics of these two signals is very difficult to extract, it is necessary to detect the signal denoising. Wavelet analysis is a more effective signal processing method, which can simultaneously in the time domain and frequency domain analysis of the signal can effectively distinguish the mutation in the part of the signal and noise, for signal denoising.
Pump vibration signal and the signal after the wavelet denoising, selected wavelet denoising threshold is 1.049 overall. Obviously, the detection signal includes a number of interference signals, it is difficult to simply use the vibration signal detected effective fault diagnosis. To eliminate the interference, after wavelet processing, can effectively eliminate the pump vibration signal contains noise, is conducive to fault feature extraction.
3, information fusion fault diagnosis method
Information fusion is to be intelligent synthesis of multi-source information to produce more accurate than a single source of information, fault tolerance and more robust estimates and judgments '2 '. Because the hydraulic pump to the information detected weak exports, easy to be overwhelmed by interference, it is difficult using a single sensor fault detection signal characteristics of weak effective diagnosis. Fault diagnosis using the information fusion process vibration signals and pressure signals about the wavelet denoising, and effective use of statistical analysis of feature information extracted by principal component analysis decouple the correlation between characteristics of the fault, reducing the dimension of fault characteristics, the improved BP neural network algorithm ball loose hydraulic pump fault diagnosis.
