An abnormal detection method for reciprocating machinery based on fault-free vibration signals

Abstract

The invention relates to a reciprocating machine abnormality detection method based on a fault-free vibration signal, comprising the following steps: 1) collecting the vibration signal of the cylinder head of the reciprocating machine and the key phase signal of the crankcase; 2) intercepting the full-period vibration signal in the angle domain; 3) Transform the vibration signal into a frequency band energy percentage feature sequence; 4) Build and train the variational autoencoder network model; 5) Use the sample reconstruction probability as an indicator to detect anomalies. The research innovation and characteristics of the present invention lie in the accurate interception and preprocessing of the vibration signal of the whole cycle in the angular domain of the diesel engine and the use of variational self-encoding network to model the potential space corresponding to the normal state of the reciprocating machinery, which makes up for the dependence on fault data in the traditional fault monitoring method It avoids the design and extraction of artificial features, and is especially suitable for fault monitoring of rotating and reciprocating power machinery.

Description

technical field [0001] The invention belongs to the technical field of abnormal detection of reciprocating mechanical equipment used in armored vehicles, ship power and construction machinery, and in particular relates to a method for detecting abnormalities of reciprocating machinery based on fault-free vibration signals, which can provide early warning for abnormalities of reciprocating machinery in the absence of fault data . Background technique [0002] Reciprocating machinery is the core power equipment of armored vehicles, ship power and construction machinery, and its safe and reliable operation plays an important role in the smooth completion of military or production tasks. The vibration signal of reciprocating machinery contains a wealth of state information. In order to give early warning to the abnormality of reciprocating machinery in time, prevent further deterioration of the abnormality, and reduce the serious loss caused by malignant faults, scholars at home...

AI Technical Summary

Problems solved by technology

[0004] In order to solve the problem of abnormal early warning in the absence of fault data, the present invention provides a reciprocating mechanical abnormal detection method based on non-fault vibration signals that does not require any abnormal data, has high efficiency and good accuracy

The purpose of the present invention is to solve the problems of difficulty in obtaining abnormal data and small quantity, etc., to provide a potential space model corresponding to the normal state of reciprocating machinery based on the variational self-encoding network based on the frequency band energy percentage characteristic sequence of the full-period signal in the angle domain, An anomaly detection method based on the probability that the sample obeys the latent space distribution is used as an index to realize early warning of abnormal states such as excessive clearance of reciprocating mechanical valves

Images