Feature Extraction Based on Teager-Kaiser Energy Operation and Envelope Spectra for Fault Detection of a Reciprocating Compressor

Reciprocating compressors are widly used in petroleum industry. A fault in reciprocating compressor may cause serious issues in operation. In the present paper, signal analysis techniques based on Teager-Kaiser energy operation and envelope spectra for fault detection of discharge valve of a reciprocating compressor are proposed. It can identify the instantaneous frequency of the signal with the energy of the fault characteristic frequency hidden in the signal and the amplitude of the single component signal.

 The method can accurately identify existing fault of vibration signal features that it simulated by the synthetic signals. Firstly, different amplitude modulation and amplitude and frequency modulation simulating waveform signals are designed and to process with Teager-Kaiser energy operator. We carry out Teager-Kaiser energy operator to verify the feasibility of amplitude envelope and instantaneous frequency analysis. To plan the impact and periodic vibration signals generated when different compressor valves failing and perform signal measurement. The original time domain signal is directly converted into the frequency domain by the fast Fourier transform  and the original time domain signal is processed with Teager-Kaiser energy operator then converted into the frequency domain.

It has two main stages: signals simulation, which is based on the operation of a reciprocating compressor, and experiment design, which uses three condition. The first stage is to simulate the operation of the reciprocating compressor, which is to simulate a synthetic signal for the cycle and impact. The synthetic signal is composed of a noise, square wave, and pulse wave. In this study, the synthetic signal is signal-processed by the Teager-Kaiser energy operator and the envelope spectrum that they can effectively extract feature signal and the noise almost is eliminated. The second stage is applied to the signal processing technique proposed in the first stage. Experimental verification of experiment design by the different operating conditions of reciprocating compressor valves.

The experimental data is obtained by the compressor-shell acceleration signal during compressor operation. Through the above analysis technology, it is proved that the synthetic signal can be eliminated the background noise to obtain the feature signal. The feasibility of the proposed approach is verified by simulation results, the experiment is validated with the measurement signals from a six-cylinder reciprocating compressor under three actual different valve conditions including normal valve operation, loose valve release and valve leakage .The results demonstrate that the proposed technology can efficiently Diagnose potential faults in compressors. Simulations and experimental results support the proposed technology positively.