Mechanical equipment abnormal sound detection method based on self-supervised feature extraction

Abstract

The invention discloses a mechanical equipment abnormal sound detection method based on self-supervised feature extraction, solves the problems of large influence of human factors, weak universality of an artificial extraction method, and the like, which are caused by the fact that mechanical equipment abnormal sound detection mainly extracts sound signal features based on an artificial construction algorithm, and then performs abnormal detection based on the features, and improves the abnormal sound detection accuracy of the mechanical equipment. The method comprises the following implementation steps: collecting sound of the mechanical equipment; preprocessing data; manufacturing an abnormal sound sample; using a normal sound sample and the manufactured abnormal sound sample to train a convolutional network as a self-supervised feature extractor; training an auto-encoder network by using extracted normal sound sample features; judging whether the sound of the mechanical equipment to be detected is abnormal or not by using a trained self-supervised feature extractor and a self-encoder network. According to the method, results show that the features extracted based on self-supervised learning are more suitable for anomaly detection, the accuracy is high, and the method can be applied to real factory mechanical equipment anomaly detection tasks.

Description

technical field [0001] The invention relates to the technical field of abnormal sound detection, in particular to a method for detecting abnormal sound of mechanical equipment based on self-supervised feature extraction. Background technique [0002] With the rapid development of industrial production automation, the normal operation of mechanical equipment in factories plays an important role in industrial production, but the high complexity of industrial equipment makes it difficult for humans to detect early equipment failures, resulting in major economic losses. It is of great significance to study the abnormal diagnosis. At present, abnormal detection of mechanical equipment based on vibration signals has been widely used. Similar to vibration signals, sound signals are an important source of information to reflect the operating status of equipment, and sound signals have the advantages of convenient collection, non-contact measurement, and fast processing speed. The ...

AI Technical Summary

Problems solved by technology

[0004] In view of this, in order to solve the problem in the prior art that when the training data only has normal sound samples, the manual extraction of sound features may be poorly correlated with the next step of abnormal detection, resulting in low accuracy of abnormal sound detection of mechanical equipment. The invention proposes a method for detecting abnormal sounds of mechanical equipment based on self-supervised feature extraction

Images