Reciprocating compressor intelligent diagnosis method based on EMD-PCA

Abstract

The invention discloses a reciprocating compressor intelligent diagnosis method based on EMD-PCA. The method comprises the steps that reciprocating compressor working condition features are extractedbased on empirical mode decomposition; dimension reduction is conducted, specifically, dimension reduction is conducted on a 30-dimensional feature vector extracted after empirical mode decompositionis conducted through a PCA method, the 30-dimensional feature vector is reduced into a 3-dimensional feature vector, and three feature parameters are obtained; according to the reciprocating compressor intelligent diagnosis method based on the EMD-PCA, an EMD method is used for decomposing non-stationary signals to obtain an intrinsic mode function (IMF), and the energy sum of different frequencyranges of each order of IMF is calculated to serve as a signal feature; main feature components in the features are extracted by using a PCA method to achieve the purpose of dimension reduction; and an SVM method identifies and classifies the high-dimensional fault features after dimension reduction. Typical faults of a reciprocating compressor can be accurately recognized, and the fault diagnosisaccuracy is high.

Description

technical field [0001] The invention relates to an intelligent diagnosis method for a reciprocating compressor. In particular, it relates to an intelligent diagnosis method for reciprocating compressors based on EMD-PCA. Background technique [0002] The reciprocating compressor is the key equipment in many process industries. Its structure is complex and there are many sources of vibration excitation. Once a failure occurs, it will easily bring huge losses to the enterprise. Therefore, it is very necessary to study on-line monitoring and intelligent fault diagnosis of reciprocating compressors. Condition monitoring of reciprocating compressors is generally realized by monitoring physical quantities such as cylinder vibration, valve cover temperature, and dynamic pressure. Monitoring and analyzing the cylinder acceleration signal of the reciprocating compressor is an important means of fault diagnosis of the reciprocating compressor. Through the monitoring of the cylinder ...

AI Technical Summary

Problems solved by technology

When identifying high-dimensional features, there is a phenomenon of dimensionality disaster. This phenomenon will inevitably lead to inaccurate fault feature identification. Therefore, on the premise of ensuring that the main information of the features is not lost, it is necessary to reduce the dimensionality of the features, and then perform feature analysis. identify

Images