Research On Valve Intelligent Fault Diagnosis System Of Reciprocating Compressor

Reciprocating compressors are widely used in petrochemical enterprises, which directly affect the state economic efficiency of enterprises. Therefore, carrying on the condition monitor and fault diagnosis for reciprocating compressors are seemed to be particularly important and meaningful.In this paper, the object of study is valve. Paper starts from the structure of reciprocating compressor, through valve vibration mechanical model, points out that valve impact lift or the seat are the main reasons to produce exciting force of the bonnet; vibration signals can reflect the valve working condition. Using wavelet packet analyses vibration signals, extract energy as feature vector, input support vector machines for intelligent diagnosis, using NI Compact RIO designs vibration signal data acquisition system of valve, and develops an intelligent fault detection system for reciprocating compressor valves using LabVIEW and Matlab.When the valve or valve spring failure, the vibration signals in the low-frequency or high frequency components will change, wavelet packet can refine the signals of different frequency bands to extract the fault characteristics of various frequency bands, and compared with the normal signal to determine the working state of valve to achieve valve fault diagnosis. Wavelet can decompose the signal to the scale domain, it is through multi-resolution decomposition, the time-frequency resolution at low frequencies is high, but at high frequencies is low. Wavelet packet technology improves the time-frequency resolution at high frequencies; solve the problem that signal frequency resolution is low at the high frequency.Acquisition signal of fault valves is difficult, number of fault signal samples is often small, while the SVM applied to a small sample of decision-making, coupled with SVM algorithm is simple, so widely used in signal intelligent fault diagnosis. After training valve's feature vector, SVM can analyse feature vector which input, determines the working state of valves to avoid the subjective judgments of engineers to achieve the vibration signals of intelligent fault diagnosis.Wavelet packet analysis has unique advantages in dealing with non-stationary signals, and support vector machine can effectively identify failure modes, experiments prove that wavelet packet analysis and SVM combination can effectively extract the fault feature vector, determine fault type.