Air valve fault diagnosis method based on HHT and neural network

Abstract

The invention discloses a compressor air valve fault diagnosis method based on HHT (Hilbert-Huang Transformation) and an RBF (radial basis function) neural network. Based on preprocessing of a vibration signal through interceptive matrix singular value decomposition, HHT is utilized to perform decomposition and time-frequency analysis on the signal, and then the RBF network is utilized to train and recognize fault sample features. The method comprises the steps that 1, interceptive matrix singular value decomposition is utilized to perform denoising preprocessing on vibration signals generatedwhen an air valve is in a normal state and in a fault state; 2, an HHT algorithm is utilized to obtain EMD (Empirical Mode Decomposition) results and Hilbert marginal spectra under various states after denoising; 3, based on the EMD results and the Hilbert marginal spectra, feature vectors of the air valve under all the operating states are extracted, and normalization processing is performed; and 4, the RBF network is utilized to train feature samples under all the states. Test samples are recognized, and the effectiveness of the method on air valve fault diagnosis is verified.

Description

technical field [0001] The invention relates to a gas valve fault diagnosis technology of a reciprocating compressor, in particular to a gas valve fault diagnosis method based on HHT and RBF neural network. Background technique [0002] The reciprocating mechanical structure is complex, and its pistons, connecting rods, air valves and other components are prone to failures, among which the failures of the air valves are frequent and there are many types of failures. Taking a reciprocating compressor as an example, the healthy operation of the valve directly affects the displacement, power consumption and reliability of the compressor. The air valve can be divided into suction valve and exhaust valve. Every time the piston reciprocates up and down, the suction valve and the exhaust valve are opened and closed once, so that the compressor can complete the working process of suction, compression, and exhaust. Because the air valve moves a lot and bears complex force during the...

AI Technical Summary

Problems solved by technology

However, FFT is only suitable for the analysis of stationary signals, and it is difficult to obtain time-frequency information, frequency and amplitude information, including some instantaneous frequency information, when analyzing non-stationary signals.

But in fact, transient vibration is very common in the vibration test signals of reciprocating compressors, so the traditional time-frequency analysis method cannot analyze video characteristics such as instantaneous amplitude and instantaneous frequency, making it difficult to carry out effective fault diagnosis

Therefore, when performing condition monitoring and fault diagnosis for reciprocating compressors, it is difficult to extract signal fault features.

Traditional signal analysis techniques, such as spectrum analysis and wavelet analysis based on Fourier transform, are difficult to reflect the time-varying and non-stationary characteristics of the signal.

Images