Mel frequency cepstrum coefficient-based motor abnormal sound detection method

Abstract

The invention discloses a Mel frequency cepstrum coefficient-based motor abnormal sound detection method. The method comprises the following steps that: a sampling frequency and a sampling duration tare set; framing and windowing processing is performed on audio signals, the frame length L of each frame is set to be L, and the overlap length (frame shift) of two adjacent frames is set to be M, the audio signals are divided into N frames of signals; audio features are extracted through using the Mel frequency cepstrum coefficient; and the mfcc parameter of each frame of current audio signals is calculated, and a change curve of each mfcc parameter changing with the number of frames (time) is obtained. The method of the invention has many advantages. With the method adopted, the abnormal sound recognition of a motor can be performed; detection efficiency can be improved; the outgoing quality of products can be ensured; the overall production efficiency of an enterprise can be improved;the manufacturing cost of the enterprise can be reduced; the physical health of workers can be protected; the problem of the unsteady state of the audio signals of the motor can be effectively solved;an unsteady motor abnormal sound fault can be effectively detected; and recognition accuracy is high.

Description

technical field [0001] The invention relates to the field of motor fault detection, in particular to a method for detecting abnormal noise of a motor. Background technique [0002] my country is the main producer of electric motors for various small household appliances such as washing machines, household air conditioners, refrigerators, and electric fans. The annual output of household air conditioner motors alone exceeds billions. [0003] In the production line of small motors, manual listening is generally used to distinguish good and defective products before the products go off-line, that is, workers listen to the sound of the motor in sequence with their ears in the soundproof room, and judge whether the motor is faulty based on the personal experience of the workers. . [0004] Due to the need for people to make subjective judgments, it has long been difficult to be replaced by automated devices. Moreover, it is impossible to establish a unified evaluation standard...

AI Technical Summary

Problems solved by technology

The disadvantages of this detection method for abnormal noise of the motor are: 1. The audio signal is a quasi-stable signal, that is, short-term stable; Carry out signal feature extraction; however, in abnormal noise faults of motors, there are a large number of fault samples of instantaneous unsteady signals; for these unsteady signals, Fourier transform is powerless

2. Judging the abnormal sound by judging whether there is a waveform beyond the specified maximum value, it does not have the ability to adapt, and the versatility is not high: because for different types of motor products, technicians need to reset the threshold; and through sensor collection For audio signals, the distance between the sensor and the sound source will also affect the threshold setting

3. Whether it is setting the maximum value or comparing waveforms, it requires professionals to judge, and automatic identification cannot be realized

Images