CN112729529A

CN112729529A - Motor defect detection method

Info

Publication number: CN112729529A
Application number: CN202011498701.4A
Authority: CN
Inventors: 陈立国; 张雯霞; 薛立伟; 王在臣
Original assignee: Suzhou University
Current assignee: Suzhou University
Priority date: 2020-12-17
Filing date: 2020-12-17
Publication date: 2021-04-30
Anticipated expiration: 2040-12-17
Also published as: CN112729529B

Abstract

The embodiment of the invention discloses a motor defect detection method. The method comprises the steps of starting the motor under the condition that the motor is in no-load condition, so that the motor idles; collecting idle running noise of the motor by adopting an acoustic sensor, and defining the noise as initial noise; performing environmental noise reduction on the initial noise to obtain pure noise; and limiting the frequency band of the pure noise, reserving the sound segment within the range of the audible frequency band of the human ear, dividing the sound segment into a plurality of sound frequency segments, analyzing the plurality of sound frequency segments by adopting a two-receptive field network, extracting defect characteristics, and judging the defect condition of the motor according to the defect characteristics. The method realizes automatic analysis of motor noise.

Description

Motor defect detection method

Technical Field

The invention relates to the technical field of industrial detection, in particular to a motor defect detection method.

Background

After the motor is assembled, a factory needs to perform defect detection on the performance of the motor before leaving the factory. Since the housing of the motor radiates almost no vibration noise when the motor is idling, it is possible to detect whether the motor is faulty or not using the noise. In particular, in the field of household appliances such as washing machines, there is a unique requirement for the noise of the motor, i.e. the appliance used in the household environment should follow the principle of silence as much as possible. When the motor of the household appliance works, the noise emitted by the motor has good roundness so as to ensure that the motor does not emit excessively sharp noise to cause discomfort. Therefore, in the field of household appliances, it is essential to detect motor defects by means of motor noise.

At present, domestic factories generally adopt a mode of ear auscultation of staff to detect motor noise. Because the types of the motor defects of the household appliances are more, and the noise difference between different defects is smaller, the method for auscultation by human ears is easy to cause misjudgment and missed judgment of the motor defects. Further, since the auscultation method of human ears depends on subjective judgment of human, a unified evaluation standard cannot be established, and it has been difficult to adopt an automatic device for replacement for a long time. In the process of mass production of household appliances, the procedure of detecting the defects of the motor by auscultation of human ears consumes a large amount of labor cost, and the repeated and monotonous listening operation easily causes fatigue of workers and irreversible damage to the hearing of the workers.

Therefore, in order to solve the above-mentioned technical problem, it is necessary to provide a method for detecting a motor defect by a motor noise method with an automatic performance.

Disclosure of Invention

In view of the above, an object of the embodiments of the present invention is to provide a method for detecting a motor defect in a motor noise manner with automatic performance. The motor defect detection method provided by the embodiment of the invention still detects the motor defects through the motor noise and realizes automatic analysis of the motor noise through the neural network model.

In order to achieve the above object, an embodiment of the present invention provides the following technical solutions: a motor defect detection method comprises the steps of S1, turning on a motor under the condition that the motor is unloaded, so that the motor idles; step S2: collecting idle running noise of the motor by adopting an acoustic sensor, and defining the noise as initial noise; step S3: performing environmental noise reduction on the initial noise to obtain pure noise; step S4: and limiting the frequency band of the pure noise, reserving the sound segment within the range of the audible frequency band of the human ear, dividing the sound segment into a plurality of sound frequency segments, analyzing the plurality of sound frequency segments by adopting a two-receptive field network, extracting defect characteristics, and judging the defect condition of the motor according to the defect characteristics.

As a further improvement of the invention, the motor is arranged in the silent box.

As a further improvement of the invention, the environmental noise is reduced to be the audio frequency with smooth change filtered by a filter.

As a further improvement of the present invention, the band limitation mode employs a 4 th order butterworth filter to filter the clean noise.

As a further development of the invention, the audio time durations of the plurality of audio segments are equal.

As a further improvement of the present invention, the step S4 includes the steps of: analyzing the plurality of audio segments by adopting a network model with a large receptive field size and extracting first defect characteristics; analyzing the plurality of audio segments by adopting a network model with a small receptive field size and extracting a second defect characteristic; performing feature fusion of different weights on the first defect feature and the second defect feature to obtain a fusion feature; and performing characteristic matching on the fusion characteristics and each defect characteristic, and judging the defect condition of the motor.

As a further refinement of the invention, the individual defects are characterized by pre-stored information.

As a further improvement of the present invention, the fusion features are feature-matched with each defect feature, and the defect feature corresponding to the maximum value of the sample classification probability is used as the defect of the motor.

The invention has the following advantages:

the motor defect detection method provided by the embodiment of the invention meets the process that the motor idle noise detection defect needs to be adopted in the field of household appliances, realizes automatic analysis of motor noise through a neural network model, and effectively solves the problems that the process depends on people and cannot be objectively evaluated in the prior art. Furthermore, in the process of automatically analyzing the motor noise, the motor noise is sliced, so that the difficulty of data processing is effectively reduced, and the processing time is saved.

Drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments described in the present invention, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.

Fig. 1 is a schematic flow chart of a method for detecting a motor defect according to an embodiment of the present invention;

fig. 2 is a schematic diagram of audio slice preservation in the embodiment shown in fig. 1.

Detailed Description

In order to make those skilled in the art better understand the technical solution of the present invention, the technical solution in the embodiment of the present invention will be clearly and completely described below with reference to the drawings in the embodiment of the present invention, and it is obvious that the described embodiment is only a part of the embodiment of the present invention, and not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

As shown in fig. 1, a flow chart of a method for detecting a motor defect according to an embodiment of the present invention is schematically shown. In this embodiment, the method for implementing the real-time operating system residing under the embedded architecture includes four steps, and the specific content of each step is as follows.

And step S1, in the case of no load of the motor, the motor is started to idle. In the embodiment of the invention, a household washing machine in household appliances is taken as a specific application scene. In particular, in the case of an unloaded washing machine, the household washing machine is turned on, so that the motor is idle.

Step S2: and collecting the idle noise of the motor by adopting an acoustic sensor, and defining the noise as initial noise. Preferably, the motor is placed in a silent box. The mute box can well filter noise in the environment and furthest reserve the audio information of motor noise.

Step S3: and carrying out environmental noise reduction on the initial noise to obtain pure noise. In the preferred embodiment, since the motor is disposed in the silent box, the ambient noise will have a tendency to change smoothly and slowly even if the ambient noise exists during the process of collecting the noise by the acoustic sensor. In this step, the ambient noise is reduced to an audio frequency with a gentle change filtered out by a filter. The environmental noise is filtered from the initial noise to obtain clean motor noise, i.e. clean noise. Further eliminate the interference of environmental noise, provide effectual noise foundation for the detection of motor defect.

Step S4: and limiting the frequency band of the pure noise, reserving the sound segment within the range of the audible frequency band of the human ear, dividing the sound segment into a plurality of sound frequency segments, analyzing the plurality of sound frequency segments by adopting a two-receptive field network, extracting defect characteristics, and judging the defect condition of the motor according to the defect characteristics.

In this embodiment, the band limitation method employs a 4-step butterworth filter to filter the clean noise, so as to filter out a noise band outside the audible range of human ears, and only reserve a band including motor defect characteristics, thereby providing efficiency of motor defect detection.

In a preferred embodiment, the audio durations of the plurality of audio segments are equal. As shown in fig. 2, the band-limited audio is divided into a plurality of audio segments with equal duration, such as 400 consecutive audio slices with equal length. By adopting a frequency band segmentation mode, the speed of subsequent neural network feature extraction and analysis can be improved, and the time length of the motor defect detection process is reduced. And respectively storing each audio segment with the same length in a defect type folder, scanning each audio segment, analyzing and extracting defect characteristics by adopting a two-receptive field network, and judging the defect condition of the motor according to the defect characteristics. And each audio segment is subjected to characteristic scanning, so that scanning calculation of the whole audio is avoided, the network calculation amount is reduced, the computer calculation force is saved, and the detection instantaneity is improved.

Specifically, the specific processing steps for each audio segment include:

step S41: and analyzing the plurality of audio segments by adopting a network model with a large receptive field size and extracting first defect characteristics. Wherein, the receptive field size is generated by automatic training of the neural network model, and is not set artificially. Taking a household washing machine as a specific application scene as an example, the network model with the large receptive field size corresponds to the noise of the washing machine when the motor rotates at the spin-drying speed.

Step S42: and analyzing the plurality of audio segments by adopting a network model with a small receptive field size and extracting a second defect characteristic. Taking a household washing machine as a specific application scene as an example, the network model with a small receptive field size corresponds to the noise of the washing machine when the motor dehydrates the rotating speed. The neural network model branches can firstly carry out local synchronous extraction on the motor noise characteristics, and as the layer number of the neural network is deepened, the network model extracts more complex and deeper signal characteristics to obtain a larger receptive field range on the basis of the extracted characteristics.

Step S43: and performing feature fusion of different weights on the first defect feature and the second defect feature to obtain a fusion feature. The method and the system have the advantages that certain weights are respectively taken for the characteristics of the large receptive field size network and the small receptive field size acquisition characteristics for characteristic fusion, and due to complementarity among different network models, the method and the system can improve the performance of the network models to a certain extent. And performing feature matching on the fusion features and each defect feature, and taking the defect feature corresponding to the maximum value of the sample classification probability as the defect of the motor. The calculation formula of the sample classification probability is shown as formula 1:

equation 1 is the probability of classifying a sample i into a class j, and is also the output after the classification process, where k is the motor defect characteristic. By the method, the defect of low convergence speed in the back propagation process is avoided, and the detection precision of the model is improved.

Step S44: and performing characteristic matching on the fusion characteristics and each defect characteristic, and judging the defect condition of the motor. In this embodiment, each defect feature is pre-stored information.

In this process, the network model analysis of the large receptive field size and the network model analysis of the small receptive field size are collectively referred to as two receptive field network analysis.

It will be evident to those skilled in the art that the invention is not limited to the details of the foregoing illustrative embodiments, and that the present invention may be embodied in other specific forms without departing from the spirit or essential attributes thereof. The present embodiments are therefore to be considered in all respects as illustrative and not restrictive, the scope of the invention being indicated by the appended claims rather than by the foregoing description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein. Any reference sign in a claim should not be construed as limiting the claim concerned.

Furthermore, it should be understood that although the present description refers to embodiments, not every embodiment may contain only a single embodiment, and such description is for clarity only, and those skilled in the art should integrate the description, and the embodiments may be combined as appropriate to form other embodiments understood by those skilled in the art.

Claims

1. A method of detecting a defect in an electric machine, the method comprising:

step S1, under the condition that the motor is unloaded, the motor is started to idle;

step S2: collecting idle running noise of the motor by adopting an acoustic sensor, and defining the noise as initial noise;

step S3: performing environmental noise reduction on the initial noise to obtain pure noise;

2. The method of claim 1, wherein the motor is placed in a quiet box.

3. The method as claimed in claim 1, wherein the noise reduction of the ambient noise is performed by filtering out a gently changing audio frequency with a filter.

4. The method of claim 1, wherein the band limiting is performed by filtering the clean noise with a 4 th order Butterworth filter.

5. The method as claimed in claim 1, wherein the audio frequency durations of the audio frequency segments are equal.

6. The method for detecting defects of a motor according to claim 5, wherein the step S4 includes the steps of:

analyzing the plurality of audio segments by adopting a network model with a large receptive field size and extracting first defect characteristics;

analyzing the plurality of audio segments by adopting a network model with a small receptive field size and extracting a second defect characteristic;

performing feature fusion of different weights on the first defect feature and the second defect feature to obtain a fusion feature;

and performing characteristic matching on the fusion characteristics and each defect characteristic, and judging the defect condition of the motor.

7. The method of claim 5, wherein each defect characteristic is pre-stored information.

8. The method as claimed in claim 5, wherein the fused features are matched with the features of each defect, and the defect feature corresponding to the maximum value of the sample classification probability is used as the defect of the motor.