Condition Detection And Fault Diagnosis Of Centrifugal Compressor And Its Application

Centrifugal compressor is one of the main equipments in the petrochemical and power plants, it has the important influence on the safe and economy operation. Many companies are trying to monitor the running state of the compressor, and to predict the trend of compressor running, and advance to find the corresponding countermeasures.This paper firstly analyzes the working principle and common faults of centrifugal compressor, and propound a a more comprehensive on-line monitoring target. For denoising the measured signal, Fourier transform, wavelet transform, wavelet packet transform and other methods are compared; and the wavelet packet transform is the most suitable denoising method for non-stationary signal.On this foundation, this paper elaborates the intelligent recognition method of operation condition of the compressor, mainly researches the spectrum analysis method to fault diagnosis method: theory based on BP neural networks. Based on extracting the fault signal feature set as training samples, this paper constructs the fault diagnosis model of three layers BP neural network, and through an example to prove the effectiveness and accuracy of this system.Finally this paper introduces the on-line monitoring and fault diagnosis system of centrifugal compressor, such as the overall structure, function of software and hardware configuration, principle of selection, and through the system in practical application cases, show that the system reaches the expected design goal, and improves the reliability of the production.