An early fault feature extraction method for compressor surge based on enhanced entropy weight

Abstract

The invention discloses a compressor early surge fault feature extraction method based on enhanced entropy weight and dimension reduction, which belongs to the technical field of mechanical equipment fault diagnosis. The method can more accurately and effectively extract early surge fault sensitive features. The method includes the following steps: collecting normal vibration data samples of m groups of compressors to be monitored and fault vibration data samples of early surge failures. Surge early failure refers to a rotating stall failure. From the normal vibration data samples and the fault vibration data samples, n characteristic parameters in the time domain and frequency domain are extracted respectively, and the sample characteristic matrix is ​​constructed. The sample feature matrix is ​​normalized, and the enhanced information entropy value of n feature parameters is calculated. Based on the enhanced information entropy value of characteristic parameters, the weight coefficients of n characteristic parameters that characterize the early failure capability of surge are measured. According to the relationship between the weight coefficients, the characteristic parameters whose weight coefficients exceed the set threshold are extracted as the sensitive characteristics of the early surge fault to be monitored.

Description

technical field [0001] The invention relates to the technical field of fault diagnosis of mechanical equipment, in particular to a method for extracting early fault features of compressor surge based on enhanced entropy weight. Background technique [0002] Compressors are widely used in petrochemical, smelting, machinery, power and other fields. In the actual operation process, it is very easy to deviate from the normal working conditions due to improper adjustment of working conditions, environmental changes and other factors, resulting in surge faults. Once the surge fault occurs, it will often cause damage to important components such as the compressor main shaft, impeller, bearing, seal, etc., which seriously threatens the safe and stable operation of the equipment. Therefore, how to quickly and accurately identify surge faults has become a key factor in maintaining the stable operation of the unit and ensuring the continuous production of the enterprise. [0003] Bas...

AI Technical Summary

Problems solved by technology

[0004] There is overlap of fault information between multi-dimensional feature parameters, resulting in redundant information, which not only fails to improve the accuracy of fault identification, but also increases the complexity of the algorithm, and brings about the problem that the features cannot be visualized

At present, the commonly used feature selection method can play an ideal effect when there are obvious differences between the characteristics of different types of data, but the characteristics of the surge signal are weak in the early stage, coupled with the interference of vibration and noise, resulting in a gap between normal and fault states. The distinction degree of the characteristic parameter values ​​between is extremely small, and it is difficult to accurately select sensitive features by using the above feature selection method, which cannot be used for the extraction of early features of surge faults.

[0005] Therefore, there are still some deficiencies in the research on the early fault characteristics of compressor surge, and the problem of how to effectively extract the fault characteristics that accurately and timely reflect the compressor surge state needs to be solved urgently

Images