CN112365901A - Mechanical audio fault detection method and device - Google Patents
Mechanical audio fault detection method and device Download PDFInfo
- Publication number
- CN112365901A CN112365901A CN202011207952.2A CN202011207952A CN112365901A CN 112365901 A CN112365901 A CN 112365901A CN 202011207952 A CN202011207952 A CN 202011207952A CN 112365901 A CN112365901 A CN 112365901A
- Authority
- CN
- China
- Prior art keywords
- data
- mechanical
- audio
- mechanical audio
- audio data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 139
- 230000009467 reduction Effects 0.000 claims abstract description 76
- 238000005520 cutting process Methods 0.000 claims abstract description 70
- 238000000034 method Methods 0.000 claims abstract description 29
- 238000012549 training Methods 0.000 claims description 57
- 238000012545 processing Methods 0.000 claims description 46
- 238000005457 optimization Methods 0.000 claims description 30
- 238000012360 testing method Methods 0.000 claims description 17
- 238000004422 calculation algorithm Methods 0.000 claims description 14
- 230000008569 process Effects 0.000 claims description 13
- 238000007637 random forest analysis Methods 0.000 claims description 12
- 238000000605 extraction Methods 0.000 claims description 11
- 238000012795 verification Methods 0.000 claims description 11
- 238000004140 cleaning Methods 0.000 claims description 5
- 238000012216 screening Methods 0.000 claims description 4
- 230000011218 segmentation Effects 0.000 claims 2
- 230000006870 function Effects 0.000 description 5
- 238000003860 storage Methods 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 4
- 238000005070 sampling Methods 0.000 description 4
- 230000002159 abnormal effect Effects 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 230000009471 action Effects 0.000 description 2
- 238000003066 decision tree Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000010365 information processing Effects 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000012952 Resampling Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 239000012634 fragment Substances 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
- 230000002194 synthesizing effect Effects 0.000 description 1
- 238000013024 troubleshooting Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS OR SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING; SPEECH OR AUDIO CODING OR DECODING
- G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
- G10L25/48—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 specially adapted for particular use
- G10L25/51—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 specially adapted for particular use for comparison or discrimination
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS OR SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING; SPEECH OR AUDIO CODING OR DECODING
- G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
- G10L25/03—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters
Abstract
The invention provides a mechanical audio fault detection method and a device, wherein the method comprises the following steps: obtaining a plurality of original mechanical audio data from a preset electronic mechanical machine, and respectively carrying out dimensionality reduction on the plurality of original mechanical audio data to obtain a plurality of mechanical audio dimensionality reduction data; respectively cutting the data samples of the multiple pieces of mechanical audio dimension reduction data to obtain multiple pieces of mechanical audio cutting data; and respectively extracting the features of the plurality of mechanical audio cutting data to obtain a mechanical audio data feature set. The method can screen out data containing missing values, acquire the characteristic points more accurately, analyze more useful information data and information with larger influence on identification and detection, is convenient to obtain a detection model with higher identification and detection accuracy, can effectively improve the reliability and stability of audio data detection, further improves the detection accuracy of the audio data, better solves the problem of identification and detection of audio faults, and realizes fault detection from the aspect of audio.
Description
Technical Field
The invention mainly relates to the technical field of audio identification, in particular to a mechanical audio fault detection method and device.
Background
With the continuous development of computer technology, network technology and communication technology, multimedia data such as images, videos and audios become important information media in the field of information processing, wherein audio information plays an important role. How to process, organize, analyze and utilize massive audio information is an important topic in the information processing field, and audio identification is one of the key technologies.
Today's society is a digital age, and many people have made continuous efforts to seek excellent sound quality in order to obtain audio information more comprehensively. Audio is an important media in multimedia, and the range of audio frequencies we can hear is 60Hz to 20kHz, with speech distributed approximately within 300Hz to 4000Hz, and music and other natural sounds are distributed over the full range. Sound is recorded or reproduced by analog equipment as analog audio, which is digitized as digital audio. The sampling rate during digitization must be higher than twice the signal bandwidth to recover the signal correctly, and samples can be represented by 8 or 16 bits. Nowadays, much attention is paid to voice recognition in the aspect of audio detection, research on voice-free audio is not comprehensive and mature enough, the characteristic feature of sound waves is not available, and detection and recognition of mechanical audio features are worth of research, such as mechanical audio troubleshooting, noise detection and the like.
Disclosure of Invention
The invention aims to solve the technical problem of the prior art and provides a mechanical audio fault detection method and device.
The technical scheme for solving the technical problems is as follows: a mechanical audio fault detection method comprises the following steps:
obtaining a plurality of original mechanical audio data from a preset electronic mechanical machine, and respectively performing dimensionality reduction processing on the plurality of original mechanical audio data to obtain a plurality of mechanical audio dimensionality reduction data;
respectively carrying out data sample cutting on the plurality of mechanical audio dimension reduction data to obtain a plurality of mechanical audio cutting data;
respectively carrying out feature extraction on the plurality of mechanical audio cutting data to obtain a plurality of mechanical audio data feature points, and collecting the plurality of mechanical audio data feature points to obtain a mechanical audio data feature set;
constructing a training model, and training the training model according to the mechanical audio data feature set to obtain an audio detection model;
optimizing the audio detection model according to the mechanical audio data feature set to obtain an audio detection optimization model;
and detecting and processing the mechanical audio data to be detected according to the audio detection optimization model to obtain a detection result of the mechanical audio data.
Another technical solution of the present invention for solving the above technical problems is as follows: a mechanical audio fault detection device, comprising:
the system comprises a dimension reduction processing module, a dimension reduction processing module and a dimension analysis module, wherein the dimension reduction processing module is used for obtaining a plurality of original mechanical audio data from a preset electronic mechanical machine and respectively carrying out dimension reduction processing on the plurality of original mechanical audio data to obtain a plurality of mechanical audio dimension reduction data;
the data sample cutting module is used for respectively cutting the data samples of the plurality of mechanical audio dimension reduction data to obtain a plurality of mechanical audio cutting data;
the characteristic extraction module is used for respectively carrying out characteristic extraction on the plurality of mechanical audio cutting data to obtain a plurality of mechanical audio data characteristic points, and collecting the plurality of mechanical audio data characteristic points to obtain a mechanical audio data characteristic set;
the model training module is used for constructing a training model and training the training model according to the mechanical audio data feature set to obtain an audio detection model;
the optimization processing module is used for optimizing the audio detection model according to the mechanical audio data feature set to obtain an audio detection optimization model;
and the detection result obtaining module is used for detecting and processing the mechanical audio data to be detected according to the audio detection optimization model to obtain the detection result of the mechanical audio data.
The invention has the beneficial effects that: the method comprises the steps of obtaining a plurality of mechanical audio frequency dimension reduction data through dimension reduction processing on a plurality of original mechanical audio frequency data, screening out data containing missing values, cutting data samples of the plurality of mechanical audio frequency dimension reduction data to obtain a plurality of mechanical audio frequency cutting data, extracting features of the plurality of mechanical audio frequency cutting data to obtain a plurality of mechanical audio frequency data feature points, obtaining the feature points more accurately, analyzing data of more useful information and information with larger influence on identification and detection, conveniently obtaining a detection model with higher identification and detection accuracy, obtaining an audio frequency detection model according to training of a mechanical audio frequency data feature set on the training model, effectively improving reliability and stability of audio frequency data detection, obtaining an audio frequency detection optimization model according to optimization processing of the mechanical audio frequency data feature set on the audio frequency detection model, and obtaining detection of the mechanical audio frequency data according to detection processing of the mechanical audio frequency data to be detected by the audio frequency detection optimization model And the detection result further improves the detection accuracy of the audio data, better solves the discrimination and detection of audio faults and realizes the fault detection from the aspect of audio.
Drawings
Fig. 1 is a schematic flow chart of a mechanical audio fault detection method according to an embodiment of the present invention;
fig. 2 is a block diagram of a mechanical audio fault detection apparatus according to an embodiment of the present invention.
Detailed Description
The principles and features of this invention are described below in conjunction with the following drawings, which are set forth by way of illustration only and are not intended to limit the scope of the invention.
Fig. 1 is a schematic flow chart of a mechanical audio fault detection method according to an embodiment of the present invention.
As shown in fig. 1, a mechanical audio fault detection method includes the following steps:
obtaining a plurality of original mechanical audio data from a preset electronic mechanical machine, and respectively performing dimensionality reduction processing on the plurality of original mechanical audio data to obtain a plurality of mechanical audio dimensionality reduction data;
respectively carrying out data sample cutting on the plurality of mechanical audio dimension reduction data to obtain a plurality of mechanical audio cutting data;
respectively carrying out feature extraction on the plurality of mechanical audio cutting data to obtain a plurality of mechanical audio data feature points, and collecting the plurality of mechanical audio data feature points to obtain a mechanical audio data feature set;
constructing a training model, and training the training model according to the mechanical audio data feature set to obtain an audio detection model;
optimizing the audio detection model according to the mechanical audio data feature set to obtain an audio detection optimization model;
and detecting and processing the mechanical audio data to be detected according to the audio detection optimization model to obtain a detection result of the mechanical audio data.
It should be understood that dimension reduction processing is performed on one of the original mechanical audio data to obtain one mechanical audio dimension reduction data.
It should be understood that, the data sample cutting is performed on one piece of the mechanical audio dimension reduction data, and a plurality of pieces of mechanical audio cutting data are obtained.
In the above embodiment, the multiple pieces of mechanical audio dimension reduction data are obtained by performing dimension reduction processing on the multiple pieces of original mechanical audio data, data containing missing values can be screened out, the multiple pieces of mechanical audio cutting data are obtained by cutting data samples of the multiple pieces of mechanical audio dimension reduction data, the multiple pieces of mechanical audio data feature points are obtained by extracting features of the multiple pieces of mechanical audio cutting data, the feature points can be obtained more accurately, data with more useful information and information with greater influence on identification and detection can be analyzed, a detection model with higher identification and detection accuracy can be obtained conveniently, an audio detection model is obtained according to training of a mechanical audio data feature set on the training model, reliability and stability of audio data detection can be effectively improved, and an audio detection optimization model is obtained according to optimization processing of the mechanical audio data feature set on the audio detection model, the detection result of the mechanical audio data is obtained by detecting and processing the mechanical audio data to be detected according to the audio detection optimization model, so that the detection accuracy of the audio data is further improved, the discrimination and detection of audio faults are better solved, and the fault detection from the aspect of audio is realized.
Optionally, as an embodiment of the present invention, the process of performing dimension reduction processing on the multiple pieces of original mechanical audio data to obtain multiple pieces of mechanical audio dimension reduction data includes:
respectively carrying out data cleaning on the plurality of original mechanical audio data to obtain a plurality of cleaned original mechanical audio data;
and respectively carrying out data dimension reduction on the cleaned original mechanical audio data by utilizing a local linear embedding algorithm to obtain a plurality of mechanical audio dimension reduction data.
It should be understood that data cleansing refers to the last procedure to find and correct recognizable errors in a data file, including checking data consistency, processing invalid and missing values, and the like.
It should be understood that a Local Linear Embedding (LLE) is a nonlinear dimension reduction algorithm, which can make the data after dimension reduction better keep the original popular structure; LLE attempts to preserve the local nature of the original high-dimensional data by assuming that the original data lies approximately on a hyperplane so that certain data of the locality can be linearly represented by its neighborhood data, and LLE considers that each data point can be constructed from a linear weighted combination of its neighbors.
It should be understood that in the field of machine learning and statistics, dimensionality reduction refers to the process of reducing the number of random variables under certain defined conditions to yield a set of "uncorrelated" principal variables. The dimension reduction of the data can save the storage space of a computer on one hand, and can eliminate the noise in the data and improve the performance of a machine learning algorithm on the other hand; the root of data dimension reduction: the data dimensionality is reduced, and the data after dimensionality reduction can represent the original data as much as possible.
Specifically, the WAV data of the original audio including the normal audio and the abnormal audio are obtained 1317 in total, namely, the raw machine audio data, where 667 normal audio data and 650 abnormal audio data are included in the 1317 pieces of raw machine audio data, the 1317 pieces of raw machine audio data are described in a data format (1317,250), where 1317 represents 1317 of the raw machine audio data, 250 represents the dimensions of each of the raw machine audio data, i.e. the size of each of said original mechanical audio data is 250 x 1, where 1 represents single channel sampling, 250 is the total number of points of single channel sampling, for WAV data of any one of said original audio, the characteristic values of some dimensions of the method can generate a plurality of columns of missing values in the column direction, so in the process of data cleaning, and deleting all columns of missing values in the characteristic values to obtain corresponding mechanical audio dimension reduction data.
In the above embodiment, the data of the plurality of original mechanical audio data are respectively cleaned to obtain the plurality of cleaned original mechanical audio data, data with poor quality and missing values can be screened out, so as to obtain mechanical audio dimension reduction data with better quality, the data of the plurality of cleaned original mechanical audio data are respectively dimension reduced by using a local linear embedding algorithm to obtain the plurality of mechanical audio dimension reduction data, so that main characteristics with greater influence on detection of the mechanical audio data are obtained, and the computation amount of subsequent steps is reduced.
Optionally, as an embodiment of the present invention, the process of respectively performing data sample cutting on the multiple pieces of mechanical audio dimension reduction data to obtain multiple pieces of mechanical audio cutting data includes:
establishing a window according to a preset window proportion, sliding the window in the direction from the starting point to the ending point of each mechanical audio dimensionality reduction data by taking half of the window length as a sliding step length, and cutting a sample when sliding is completed each time, so as to obtain a plurality of mechanical audio cutting data.
It should be understood that, the data sample cutting is performed on a plurality of pieces of the mechanical audio dimension reduction data in a window sliding mode, so that the sample size is enlarged, and a plurality of pieces of the mechanical audio cutting data are obtained.
It should be understood that, in general, the audio detection is monitored at intervals and only for a short time, and the fault audio is not necessarily at the beginning of the mechanical audio dimension reduction data, so that sample cutting is required.
Specifically, the mechanical audio dimension reduction data are respectively subjected to sample cutting by taking the window proportion as the size of a window sample and taking half of the window length as a sliding step length according to a preset window proportion.
It should be understood that, according to the proportional length of the mechanical audio dimension reduction data, a window with a fixed proportional length may be set, the window is moved by a step of half the window length from the beginning of the data, the complete sample is divided into the mechanical audio cutting data, the data size is expanded from 1316 to 30268, the feature information is refined, and the position of the feature point can be better determined.
Specifically, the dimension reduction data is divided into 1: and 12, as a fixed window proportion, using 1/24 of the single mechanical audio dimension reduction data as a sliding step length, dividing each mechanical audio dimension reduction data into 23 mechanical audio cutting data, sliding until the last audio is discarded, not affecting the accuracy of the data, increasing the number of samples under the action of window sliding, and providing more support for model training.
In the above embodiment, the window is established according to the preset window proportion, half of the window length is used as the sliding step length, the window is slid from the starting point to the ending point of each mechanical audio dimensionality reduction data, the sample is cut when the sliding is completed each time, so that a plurality of mechanical audio cutting data are obtained, the sample size can be enlarged, the feature information can be refined, the position of the feature point can be better determined, meanwhile, the number of samples is increased under the action of the sliding of the window, and more support is provided for model training.
Optionally, as an embodiment of the present invention, the process of respectively performing feature extraction on the multiple pieces of mechanical audio cutting data to obtain multiple pieces of mechanical audio data feature points includes:
marking each piece of mechanical audio cutting data to obtain a plurality of marked mechanical audio cutting data;
and respectively identifying the characteristic points of the marked mechanical audio cutting data by using a preset spectrogram to obtain a plurality of mechanical audio data characteristic points.
Specifically, the mechanical audio cutting data with important features are clustered by respectively identifying feature points of the marked mechanical audio cutting data by using a preset spectrogram.
It should be understood that identifying characteristic points of 30268 mechanical audio cutting data clusters important characteristic data, and 28854 mechanical audio data characteristic points are reserved; the spectrogram is obtained by calling a spectrogram function, wherein the spectrogram function also has the self-contained sliding detection effect, can more finely process a cut sample, and obtains a signal by using short-time Fourier transform.
It should be understood that spectrograms are the primary tools for analyzing vibration parameters, and are used in mechanical fault diagnosis systems to answer questions about the location, type, extent, etc. of a fault; the problem of audio fault maintenance can be solved better by using the spectrogram to extract features, the amplitude and frequency signals of audio are analyzed by using the spectrogram of a spectral function in MATLAB, the mechanical audio data feature points are obtained from the audio, 28854 mechanical audio data feature points are identified by the spectrogram, the spectrogram has the effect of self-contained sliding detection, and a cut sample can be processed more finely.
In the above embodiment, the mechanical audio cutting data after a plurality of marks are obtained for the marks of each mechanical audio cutting data respectively, a plurality of mechanical audio data feature points are obtained by utilizing the preset spectrogram to identify the feature points of the mechanical audio cutting data after the marks respectively, the feature points can be obtained more accurately, more useful information data and information with larger influence on identification and detection can be analyzed, a detection model with higher identification and detection accuracy can be obtained conveniently, the detection accuracy of the audio data is further improved, the audio fault identification and detection are better solved, and the fault detection from the aspect of audio is realized.
Optionally, as an embodiment of the present invention, after the feature point identification, the method further includes the following steps:
and storing the unidentified mechanical audio cutting data without the mechanical audio data characteristic points.
Specifically, unidentified mechanical audio cutting data without the characteristic points of the mechanical audio data are stored, the positions of the audio fault sections can be located, and fault detection not only aims at detecting which audio section is in problem, but also can detect where the audio fault occurs.
In the above embodiment, the storage of unidentified mechanical audio cutting data without the mechanical audio data feature points can locate the segment position where the audio fault occurs.
Optionally, as an embodiment of the present invention, the constructing a training model, and training the training model according to the mechanical audio data feature set to obtain an audio detection model includes:
s1: dividing the mechanical audio data feature set according to a preset proportion to obtain a mechanical audio data feature training set, a mechanical audio data feature testing set and a mechanical audio data feature verification set;
s2: constructing a model based on a random forest algorithm to obtain a random forest structure;
s3: inputting the mechanical audio data feature training set and the mechanical audio data feature testing set into the random forest structure for detection processing to obtain a training model;
s4: and carrying out model screening processing on the training model according to the mechanical audio data feature training set and the mechanical audio data feature testing set to obtain an audio detection model.
It should be understood that, since each division of the mechanical audio data feature set into the mechanical audio data feature training set, the mechanical audio data feature test set and the mechanical audio data feature verification set is a random division, random proportions of each division are different, and a train _ test _ split function can be called for the random division.
It should be understood that each decision tree in the random forest structure is utilized to detect the mechanical audio data feature training set and the mechanical audio data feature testing set respectively, so as to obtain the training model.
Specifically, in the process of synthesizing 28854 mechanical audio data feature points into the mechanical audio data feature set, since 28854 mechanical audio data feature points include 13928 normal mechanical audio data and 14936 abnormal mechanical audio data, when a sample is selected, random sampling is performed, and the mechanical audio data feature training set, the mechanical audio data feature test set, and the mechanical audio data feature verification set are obtained for performing experiments.
It should be understood that Random Forest (RF) is an ensemble learning algorithm, which uses a Random resampling technique bootstrap and a node Random splitting technique to construct a plurality of decision trees, and obtains a final detection result by voting; the RF has the capability of analyzing complex interaction detection characteristics, has good robustness on noise data and data with missing values, has fast learning depth, integrates the advantages of various machine detection algorithms, and has high performance and strong stability.
It should be understood that model optimization of random forests is done from two aspects: 1. performing the feature point concentration of the mechanical audio data, and distinguishing the mechanical audio data feature points from the mechanical audio cutting data without the mechanical audio data feature points, wherein the audio fault location is also positioned to enable a model to only train a sample with the mechanical audio data feature points, so that a better training sample can be obtained, namely, a fragment without a feature effect is screened out, and the sample is concentrated on an effective sample; 2. and (6) optimizing parameters.
In the above embodiment, the mechanical audio data feature training set, the mechanical audio data feature testing set and the mechanical audio data feature verifying set are obtained by dividing the mechanical audio data feature set according to the preset proportion, so that the objectivity of data can be ensured, human factors can be reduced, the accuracy of a subsequent detection model can be effectively improved, the mechanical audio data feature training set and the mechanical audio data feature testing set are input into the random forest structure for detection processing to obtain the training model, the higher detection accuracy is ensured, and the detection model meeting the expectation is obtained, the audio detection model is obtained by screening the training model according to the mechanical audio data feature training set and the mechanical audio data feature testing set, so that the accuracy of recognition and detection can be kept at a higher level all the time, and the stability and reliability of audio data recognition are improved.
Optionally, as an embodiment of the present invention, the process of performing optimization processing on the audio detection model according to the mechanical audio data feature set to obtain an audio detection optimization model includes:
s5, inputting the mechanical audio data feature training set and the mechanical audio data feature testing set into the audio detection model by using an automatic parameter searching algorithm to perform parameter optimization processing to obtain optimized parameters;
s6: replacing parameters in the audio detection model according to the optimized parameters to obtain an audio detection model to be verified;
s7: verifying the audio detection model to be verified according to the mechanical audio data feature verification set, and if the verification is passed, taking the audio detection model to be verified as an audio detection optimization model; if the verification is not passed, the process returns to step S1.
Specifically, the mechanical audio data feature training set and the mechanical audio data feature testing set are input into the corresponding audio detection model, a parameter range of the audio detection model is input, an automatic parameter adjusting function is used for performing parameter adjustment on the audio detection model, and an optimal parameter is selected as a parameter value of the audio detection model to obtain the audio detection model to be verified.
In the above embodiment, the automatic parameter-finding algorithm is used to input the mechanical audio data feature training set and the mechanical audio data feature testing set into the audio detection model for parameter optimization processing to obtain optimized parameters, the audio detection model to be verified is obtained by replacing the parameters in the audio detection model according to the optimized parameters, and the audio detection optimized model is obtained by verifying the audio detection model to be verified according to the mechanical audio data feature verification set.
Fig. 2 is a block diagram of a mechanical audio fault detection apparatus according to an embodiment of the present invention.
Alternatively, as another embodiment of the present invention, as shown in fig. 2, a mechanical audio fault detection apparatus includes:
the system comprises a dimension reduction processing module, a dimension reduction processing module and a dimension analysis module, wherein the dimension reduction processing module is used for obtaining a plurality of original mechanical audio data from a preset electronic mechanical machine and respectively carrying out dimension reduction processing on the plurality of original mechanical audio data to obtain a plurality of mechanical audio dimension reduction data;
the data sample cutting module is used for respectively cutting the data samples of the plurality of mechanical audio dimension reduction data to obtain a plurality of mechanical audio cutting data;
the characteristic extraction module is used for respectively carrying out characteristic extraction on the plurality of mechanical audio cutting data to obtain a plurality of mechanical audio data characteristic points, and collecting the plurality of mechanical audio data characteristic points to obtain a mechanical audio data characteristic set;
the model training module is used for constructing a training model and training the training model according to the mechanical audio data feature set to obtain an audio detection model;
the optimization processing module is used for optimizing the audio detection model according to the mechanical audio data feature set to obtain an audio detection optimization model;
and the detection result obtaining module is used for detecting and processing the mechanical audio data to be detected according to the audio detection optimization model to obtain the detection result of the mechanical audio data.
Optionally, as an embodiment of the present invention, the dimension reduction processing module is specifically configured to:
respectively carrying out data cleaning on the plurality of original mechanical audio data to obtain a plurality of cleaned original mechanical audio data;
and respectively carrying out data dimension reduction on the cleaned original mechanical audio data by utilizing a local linear embedding algorithm to obtain a plurality of mechanical audio dimension reduction data.
Optionally, as an embodiment of the present invention, the data sample cutting module is specifically configured to:
establishing a window according to a preset window proportion, sliding the window in the direction from the starting point to the ending point of each mechanical audio dimensionality reduction data by taking half of the window length as a sliding step length, and cutting a sample when sliding is completed each time, so as to obtain a plurality of mechanical audio cutting data.
It is clear to those skilled in the art that, for convenience and brevity of description, the specific working processes of the above-described apparatuses and units may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.
In the several embodiments provided in the present application, it should be understood that the disclosed apparatus and method may be implemented in other ways. For example, the above-described apparatus embodiments are merely illustrative, and for example, a division of a unit is merely a logical division, and an actual implementation may have another division, for example, a plurality of units or components may be combined or integrated into another system, or some features may be omitted, or not executed.
Units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment of the present invention.
In addition, functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.
The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. It will be understood that the technical solution of the present invention essentially contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product stored in a storage medium and including instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.
While the invention has been described with reference to specific embodiments, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (10)
1. A mechanical audio fault detection method is characterized by comprising the following steps:
obtaining a plurality of original mechanical audio data from a preset electronic mechanical machine, and respectively performing dimensionality reduction processing on the plurality of original mechanical audio data to obtain a plurality of mechanical audio dimensionality reduction data;
respectively carrying out data sample cutting on the plurality of mechanical audio dimension reduction data to obtain a plurality of mechanical audio cutting data;
respectively carrying out feature extraction on the plurality of mechanical audio cutting data to obtain a plurality of mechanical audio data feature points, and collecting the plurality of mechanical audio data feature points to obtain a mechanical audio data feature set;
constructing a training model, and training the training model according to the mechanical audio data feature set to obtain an audio detection model;
optimizing the audio detection model according to the mechanical audio data feature set to obtain an audio detection optimization model;
and detecting and processing the mechanical audio data to be detected according to the audio detection optimization model to obtain a detection result of the mechanical audio data.
2. The method according to claim 1, wherein the step of performing dimension reduction processing on the plurality of original mechanical audio data to obtain a plurality of mechanical audio dimension reduction data comprises:
respectively carrying out data cleaning on the plurality of original mechanical audio data to obtain a plurality of cleaned original mechanical audio data;
and respectively carrying out data dimension reduction on the cleaned original mechanical audio data by utilizing a local linear embedding algorithm to obtain a plurality of mechanical audio dimension reduction data.
3. The method according to claim 1, wherein the step of performing data sample segmentation on the plurality of pieces of mechanical audio dimension reduction data to obtain a plurality of pieces of mechanical audio segmentation data comprises:
establishing a window according to a preset window proportion, sliding the window in the direction from the starting point to the ending point of each mechanical audio dimensionality reduction data by taking half of the window length as a sliding step length, and cutting a sample when sliding is completed each time, so as to obtain a plurality of mechanical audio cutting data.
4. The method according to claim 1, wherein the step of performing feature extraction on the plurality of pieces of mechanical audio cutting data to obtain a plurality of mechanical audio data feature points comprises:
marking each piece of mechanical audio cutting data to obtain a plurality of marked mechanical audio cutting data;
and respectively identifying the characteristic points of the marked mechanical audio cutting data by using a preset spectrogram to obtain a plurality of mechanical audio data characteristic points.
5. The mechanical audio fault detection method of claim 4, further comprising, after feature point identification, the steps of:
and storing the unidentified mechanical audio cutting data without the mechanical audio data characteristic points.
6. The method according to claim 1, wherein the constructing a training model and training the training model according to the mechanical audio data feature set to obtain an audio detection model comprises:
s1: dividing the mechanical audio data feature set according to a preset proportion to obtain a mechanical audio data feature training set, a mechanical audio data feature testing set and a mechanical audio data feature verification set;
s2: constructing a model based on a random forest algorithm to obtain a random forest structure;
s3: inputting the mechanical audio data feature training set and the mechanical audio data feature testing set into the random forest structure for detection processing to obtain a training model;
s4: and carrying out model screening processing on the training model according to the mechanical audio data feature training set and the mechanical audio data feature testing set to obtain an audio detection model.
7. The method according to claim 6, wherein the optimizing the audio detection model according to the mechanical audio data feature set to obtain an audio detection optimization model comprises:
s5, inputting the mechanical audio data feature training set and the mechanical audio data feature testing set into the audio detection model by using an automatic parameter searching algorithm to perform parameter optimization processing to obtain optimized parameters;
s6: replacing parameters in the audio detection model according to the optimized parameters to obtain an audio detection model to be verified;
s7: verifying the audio detection model to be verified according to the mechanical audio data feature verification set, and if the verification is passed, taking the audio detection model to be verified as an audio detection optimization model; if the verification is not passed, the process returns to step S1.
8. A mechanical audio fault detection device, comprising:
the system comprises a dimension reduction processing module, a dimension reduction processing module and a dimension analysis module, wherein the dimension reduction processing module is used for obtaining a plurality of original mechanical audio data from a preset electronic mechanical machine and respectively carrying out dimension reduction processing on the plurality of original mechanical audio data to obtain a plurality of mechanical audio dimension reduction data;
the data sample cutting module is used for respectively cutting the data samples of the plurality of mechanical audio dimension reduction data to obtain a plurality of mechanical audio cutting data;
the characteristic extraction module is used for respectively carrying out characteristic extraction on the plurality of mechanical audio cutting data to obtain a plurality of mechanical audio data characteristic points, and collecting the plurality of mechanical audio data characteristic points to obtain a mechanical audio data characteristic set;
the model training module is used for constructing a training model and training the training model according to the mechanical audio data feature set to obtain an audio detection model;
the optimization processing module is used for optimizing the audio detection model according to the mechanical audio data feature set to obtain an audio detection optimization model;
and the detection result obtaining module is used for detecting and processing the mechanical audio data to be detected according to the audio detection optimization model to obtain the detection result of the mechanical audio data.
9. The mechanical audio fault detection device of claim 8, wherein the dimension reduction processing module is specifically configured to:
respectively carrying out data cleaning on the plurality of original mechanical audio data to obtain a plurality of cleaned original mechanical audio data;
and respectively carrying out data dimension reduction on the cleaned original mechanical audio data by utilizing a local linear embedding algorithm to obtain a plurality of mechanical audio dimension reduction data.
10. The mechanical audio fault detection device of claim 8, wherein the data sample cutting module is specifically configured to:
establishing a window according to a preset window proportion, sliding the window in the direction from the starting point to the ending point of each mechanical audio dimensionality reduction data by taking half of the window length as a sliding step length, and cutting a sample when sliding is completed each time, so as to obtain a plurality of mechanical audio cutting data.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011207952.2A CN112365901A (en) | 2020-11-03 | 2020-11-03 | Mechanical audio fault detection method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011207952.2A CN112365901A (en) | 2020-11-03 | 2020-11-03 | Mechanical audio fault detection method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112365901A true CN112365901A (en) | 2021-02-12 |
Family
ID=74514049
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011207952.2A Pending CN112365901A (en) | 2020-11-03 | 2020-11-03 | Mechanical audio fault detection method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112365901A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114485496A (en) * | 2021-12-30 | 2022-05-13 | 安徽中科昊音智能科技有限公司 | Clean ball position system for monitoring oil pipe |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108593282A (en) * | 2018-07-05 | 2018-09-28 | 国网安徽省电力有限公司培训中心 | A kind of breaker on-line monitoring and fault diagonosing device and its working method |
| CN108806718A (en) * | 2018-06-06 | 2018-11-13 | 华中师范大学 | Based on the audio authentication method to ENF phase spectrums and instantaneous frequency spectrum analysis |
| CN111134664A (en) * | 2019-12-18 | 2020-05-12 | 武汉工程大学 | Epileptic discharge identification method and system based on capsule network and storage medium |
| CN111341343A (en) * | 2020-03-02 | 2020-06-26 | 乐鑫信息科技（上海）股份有限公司 | Online updating system and method for abnormal sound detection |
| CN111508480A (en) * | 2020-04-20 | 2020-08-07 | 网易（杭州）网络有限公司 | Training method of audio recognition model, audio recognition method, device and equipment |
| CN111768761A (en) * | 2019-03-14 | 2020-10-13 | 京东数字科技控股有限公司 | Training method and device of voice recognition model |
-
2020
- 2020-11-03 CN CN202011207952.2A patent/CN112365901A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108806718A (en) * | 2018-06-06 | 2018-11-13 | 华中师范大学 | Based on the audio authentication method to ENF phase spectrums and instantaneous frequency spectrum analysis |
| CN108593282A (en) * | 2018-07-05 | 2018-09-28 | 国网安徽省电力有限公司培训中心 | A kind of breaker on-line monitoring and fault diagonosing device and its working method |
| CN111768761A (en) * | 2019-03-14 | 2020-10-13 | 京东数字科技控股有限公司 | Training method and device of voice recognition model |
| CN111134664A (en) * | 2019-12-18 | 2020-05-12 | 武汉工程大学 | Epileptic discharge identification method and system based on capsule network and storage medium |
| CN111341343A (en) * | 2020-03-02 | 2020-06-26 | 乐鑫信息科技（上海）股份有限公司 | Online updating system and method for abnormal sound detection |
| CN111508480A (en) * | 2020-04-20 | 2020-08-07 | 网易（杭州）网络有限公司 | Training method of audio recognition model, audio recognition method, device and equipment |
Non-Patent Citations (2)
| Title |
|---|
| 姚勇: "基于声学信号的典型机械设备故障智能诊断方法研究", 《中国优秀博硕士学位论文全文数据库》 * |
| 郑树泉: "《工业智能技术与应用》", 31 December 2018, 上海科学技术出版社 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114485496A (en) * | 2021-12-30 | 2022-05-13 | 安徽中科昊音智能科技有限公司 | Clean ball position system for monitoring oil pipe |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US8140331B2 (en) | Feature extraction for identification and classification of audio signals | |
| Mekyska et al. | Robust and complex approach of pathological speech signal analysis | |
| CN110782920B (en) | Audio recognition method and device and data processing equipment | |
| CN107274916B (en) | Method and device for operating audio/video file based on voiceprint information | |
| JP4797342B2 (en) | Method and apparatus for automatically recognizing audio data | |
| US9058384B2 (en) | System and method for identification of highly-variable vocalizations | |
| CN105513598A (en) | Playback voice detection method based on distribution of information quantity in frequency domain | |
| US11735202B2 (en) | Systems and methods for pre-filtering audio content based on prominence of frequency content | |
| JP4267463B2 (en) | Method for identifying audio content, method and system for forming a feature for identifying a portion of a recording of an audio signal, a method for determining whether an audio stream includes at least a portion of a known recording of an audio signal, a computer program , A system for identifying the recording of audio signals | |
| CN110689885B (en) | Machine synthesized voice recognition method, device, storage medium and electronic equipment | |
| CN112365901A (en) | Mechanical audio fault detection method and device | |
| Heise et al. | Acoustic detection of bees in the field using CASA with focal templates | |
| CN109271501A (en) | A kind of management method and system of audio database | |
| Huang et al. | Fast diagnosis of bowel activities | |
| CN116741200A (en) | Locomotive fan fault detection method and device | |
| CN107993666B (en) | Speech recognition method, speech recognition device, computer equipment and readable storage medium | |
| Frusque et al. | Canonical polyadic decomposition and deep learning for machine fault detection | |
| CN114996489A (en) | Method, device and equipment for detecting violation of news data and storage medium | |
| CN117351988B (en) | Remote audio information processing method and system based on data analysis | |
| CN115132183B (en) | Training method, device, equipment, medium and program product of audio recognition model | |
| CN116612746B (en) | Speech coding recognition method in acoustic library based on artificial intelligence | |
| Barbedo et al. | Empirical methods to determine the number of sources in single-channel musical signals | |
| CN112634942B (en) | Method for identifying originality of mobile phone recording, storage medium and equipment | |
| CN117409761B (en) | Method, device, equipment and storage medium for synthesizing voice based on frequency modulation | |
| Murphy et al. | Application of audio fingerprinting to Neonatal EEG |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210212 |