CN115758237A - Bearing fault classification method and system based on intelligent inspection robot - Google Patents
Bearing fault classification method and system based on intelligent inspection robot Download PDFInfo
- Publication number
- CN115758237A CN115758237A CN202211367420.4A CN202211367420A CN115758237A CN 115758237 A CN115758237 A CN 115758237A CN 202211367420 A CN202211367420 A CN 202211367420A CN 115758237 A CN115758237 A CN 115758237A
- Authority
- CN
- China
- Prior art keywords
- bearing
- feature
- classification
- inspection robot
- 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
Images
Landscapes
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The invention discloses a bearing fault classification method and a system based on an intelligent inspection robot, belonging to the technical field of sensing technology and digital signal processing, aiming at solving the technical problem of classifying bearing faults according to sound by using the intelligent inspection robot so as to accurately acquire the fault parts of the bearings, and adopting the technical scheme that: s1, acquiring acoustic signals: the method comprises the steps that fault sounds of a bearing inner ring raceway, a bearing outer ring raceway, a bearing rolling body and a bearing assembly are collected based on a collection card and a microphone of an intelligent inspection robot, and sound information of normal bearing operation is collected; s2, constructing a time-frequency signal: constructing a time-frequency signal to perform data dimension-increasing signal feature expression; s3, feature extraction: extracting features through a self-coding network; s4, feature classification: carrying out feature classification through a convolutional neural network; s5, optimizing and classifying: and training and testing by using an Adam optimization algorithm, and continuously adjusting parameter settings in the network so as to obtain a more accurate classification result.
Description
Technical Field
The invention relates to the technical field of sensing technology and digital signal processing, in particular to a bearing fault classification method and system based on an intelligent inspection robot.
Background
In the study of mechanical failure diagnosis, vibration signals are generally the main study objects. The purpose of fault analysis without disassembling the machine is realized by utilizing the relation between a fault source and vibration of a bearing or a motor and the like. However, in actual operation, a position needs to be reserved for an acceleration sensor and the like, and strict specifications are needed to be used, otherwise the sensor can fly out during the operation of the bearing to cause danger.
Through analysis, the sound is essentially a mechanical wave which is outwards propagated through an air medium by the vibration of an object, and the energy generated by the vibration of a sound source pushes a nearby elastic medium to diffuse and propagate towards the periphery to form a sound wave. Thus, the acoustic signal and the vibration signal are identical in nature, except for differences in the form of representation. The sound signal is collected by a microphone, and the sound signal is overlapped and reflected in the air, but the superposition and the reflection tend to enhance certain characteristics of the signal, and are beneficial. Above all, the operation of collecting acoustic signals in actual working conditions is simpler.
Therefore, how to classify the bearing fault by using the intelligent inspection robot according to the sound and further accurately obtain the fault part of the bearing is a technical problem to be solved urgently at present.
Disclosure of Invention
The invention provides a bearing fault classification method and system based on an intelligent inspection robot, and aims to solve the problem of how to classify bearing faults according to sound by using the intelligent inspection robot so as to accurately acquire the fault part of a bearing.
The technical task of the invention is realized in the following way, and the bearing fault classification method based on the intelligent inspection robot specifically comprises the following steps:
s1, acquiring acoustic signals: the method comprises the steps that fault sounds of a bearing inner ring raceway, a bearing outer ring raceway, a bearing rolling body and a bearing assembly are collected based on a collection card and a microphone of an intelligent inspection robot, and sound information of normal bearing operation is collected;
s2, constructing a time-frequency signal: constructing a time-frequency signal to perform data dimension-increasing signal feature expression;
s3, feature extraction: extracting features through a self-coding network;
s4, feature classification: performing feature classification through a convolutional neural network;
s5, optimizing and classifying: and training and testing by using an Adam optimization algorithm, and continuously adjusting parameter setting in the network so as to obtain a more accurate classification result.
Preferably, the step S1 of acquiring the acoustic signals specifically includes:
s101, connecting or discovering acquisition equipment, wherein an acquisition signal source is a bearing and comprises a bearing inner ring raceway fault signal, a bearing outer ring raceway fault signal, a bearing rolling body fault signal and a bearing assembly fault signal;
s102, acquiring equipment state information: acquiring equipment information in a get request mode, distinguishing WebDAQ 504 distinguishing equipment according to an IP (Internet protocol) address and an MAC (media access control) address, and selecting a host address as manual input; for convenience of data analysis, converting the data into a json format;
s103, preliminarily judging whether the equipment is available:
(1) if yes, executing step S104;
(2) if not, jumping to the step S101;
s104, acquiring detailed information of the complete system for recording data sources;
s105, setting the sound acquisition task through an acquisition card, wherein the sound acquisition task comprises conditions of sampling rate, channel number and sound pressure for setting microphone working, and issuing the acquisition task;
s106, acquiring the state of an equipment working list;
s107, judging whether a task is in progress or is suspended:
(1) if yes, stopping the existing task or waiting for queuing, and executing the step S108;
(2) if not, executing step S108;
s108, acquiring the running state of the equipment and executing tasks;
and S109, converting and outputting the related sound data of the corresponding acquisition equipment.
Preferably, when the state of the equipment working list is acquired, whether the current task list directory is legal is judged:
and if the list task is legal, printing the information of the list task, wherein the information of the list task comprises the task name, the channel number and the progress condition of the task or automatically selecting the existing task for viewing.
Preferably, the acquisition card adopts an internet of things sound and vibration data recorder WebDAQ 504;
the microphone adopts two acoustic sensors of SKC AD7199 and SKC AD7018 to carry out double-channel acquisition.
Preferably, the time-frequency signal constructed in step S2 is specifically as follows:
the method adopts space-time frequency domain transformation and utilizes fast Fourier transformation to construct multi-scale frequency domain information;
two-dimensionally unfolding the time sequence signals on the separated frequency domain scale in a windowing mode to form two-dimensional signal representation under multiple scales; the separated time sequence signals are subjected to resampling on a time sequence to obtain a time sequence, the time sequence is arranged in a space dimension to form preprocessed signals, and the formula is as follows:
wherein L (i) represents the length of the original signal; p (j, k) represents the representation of a single pixel point in the image; i =1, \8230;, M2; j =1, \ 8230;, M; k =1, \ 8230 \ 8230;, M;
the purpose of performing two-dimensional rearrangement on the data by adopting rounding-down is that the data is between 0 and 255, the one-dimensional vibration data becomes an image through normalization, actually, the data is converted into the current 8bit from the original 64bit, and the data storage space of a single point is greatly reduced; the method for windowing and resampling the data actually increases effective analysis data, so that the data is effectively analyzed on the basis of not occupying too large data storage space as much as possible, and the correlation among the data is extracted.
Preferably, the feature extraction through the self-coding network in step S3 is specifically as follows:
carrying out data denoising and dimensionality reduction in a self-coding mode after data dimensionality increasing: carrying out data denoising and dimensionality reduction by adopting a Tensorflow end-to-end open source machine learning platform and a Keras-based deep learning framework;
in self-coding, a time-frequency signal with characteristic dimensionality defined by 2400 multiplied by 1 and channel number of 1 is input, 10 filters are constructed by two-dimensional convolution, a 128 multiplied by 1 sliding window is adopted, in order to keep data information as much as possible, a filtering mode is set as "sea", and a tangent function tanh is selected as an activation function;
the down-sampling selection uses more maximum pooling, the size of the sampling factor is 8 multiplied by 1, and the step length is 2 to obtain a 1200 multiplied by 1 multiplied by 10 characteristic diagram;
the method comprises the steps of up-sampling a 1200X 1X 10 feature diagram, although the dimension of externally viewed data returns to the original 2400X 1X 10, performing interpolation operation without introducing other parameters, amplifying the feature diagram and purifying feature information; wherein, after adding the encode and decode operation, inserting a standard self-encoding (Autoencode) operation; the self-coding (auto encoder) operation is characterized in that the input and output feature quantity is the same, the input and output feature quantity is defined as the input data compression by a neural network and the reconstruction after the essence extraction, the feature maps are continuously reduced by the convolution layer and the sub-sampling layer through the feature extraction of multiple stacking, and the quantity of the feature maps is continuously increased;
at the end of the feature extractor, all feature maps are expanded and arranged into a vector, called a feature vector, which is used as the input of a post-classifier;
and (5) the signals are sent to a neural network classifier for training, and classification of the bearing fault sound signals is realized.
Preferably, the feature classification performed by the convolutional neural network in step S4 is specifically as follows:
the learning process of the convolutional neural network aims to obtain a minimum error: label y of actual output ik And predicted label d ik The difference between should be reduced as much as possible, relying on the constant update of the weights ω and the deviations b during the training process; the loss function adopts a mean square error, the mean square error reflects the degree of closeness between probability distribution of various classes of softmax and actual classes, and if the difference between a model prediction result and the actual classes is small, the calculated mean square error is small;
carrying out classification test on the formed column vectors by using a trained network, and designing the column vectors step by step into 400 and 300 classes & ltcng & gtin sequence;
the method is finally divided into five types according to the category of the collected data, and the five types correspond to five different states of the collected signals;
wherein, the evaluation indexes of the classification comprise Accuracy (Accuracy), precision (Precision) and Recall (Recall); meanwhile, a parameter F1 is introduced for evaluating the quality of the model, and the formula is as follows:
F1＝2*(Precision*Recall)/(Precision+Recall)。
a bearing fault classification system based on an intelligent inspection robot comprises,
the signal acquisition module is used for acquiring fault sounds of the bearing inner ring raceway, the bearing outer ring raceway, the bearing rolling body and the bearing assembly based on an acquisition card and a microphone of the intelligent inspection robot and acquiring sound information of normal bearing operation;
the time-frequency signal construction module is used for constructing a time-frequency signal to perform data dimension-increasing signal feature expression;
the characteristic extraction module is used for extracting characteristics through a self-coding network;
the characteristic classification module is used for carrying out characteristic classification through a convolutional neural network;
and the optimization classification module is used for training and testing by using an Adam optimization algorithm, continuously adjusting parameter setting in the network and further acquiring a more accurate classification result.
An electronic device, comprising: a memory and at least one processor;
wherein the memory has stored thereon a computer program;
the at least one processor executes the computer program stored in the memory, so that the at least one processor performs the intelligent inspection robot-based bearing fault classification method according to any one of claims 1 to 7.
A computer-readable storage medium having stored thereon a computer program executable by a processor to implement a smart inspection robot-based bearing fault classification method as described above.
The bearing fault classification method and system based on the intelligent inspection robot have the following advantages:
the invention (I) carries out fault classification aiming at rolling bearings in the same rotating speed state, and aims to classify 5 faults under different conditions, from the construction of time-frequency signals to the data dimension-increasing and signal feature-enhancing expression to the design of a self-coding network to realize feature extraction, finally, the convolution spirit realizes feature classification of the network, the Adam optimization algorithm is utilized to train and test, the parameter setting in the network is continuously adjusted in the whole process, and more accurate classification effect is obtained;
based on the application of the existing intelligent patrol robot, the invention designs a sound acquisition program by using an acquisition card to finish the acquisition of sound signals, designs a self-coding network to extract characteristics, and adopts a signal classification method under a convolutional neural network to finish the state diagnosis of equipment according to the acquired sound data.
Drawings
The invention is further described below with reference to the accompanying drawings.
FIG. 1 is a flow chart diagram of a bearing fault classification method based on an intelligent inspection robot;
FIG. 2 is a block flow diagram of the acquisition of acoustic signals;
FIG. 3 is a waveform diagram of a rolling element fault signal;
FIG. 4 is a waveform diagram of a fault signal for a combination;
FIG. 5 is a waveform diagram of a fault signal of an inner ring raceway;
FIG. 6 is a waveform of a bearing in a normal state;
FIG. 7 is a waveform diagram of outer race raceway fault signals.
Detailed Description
The bearing fault classification method and system based on the intelligent inspection robot of the invention are described in detail below with reference to the drawings and specific embodiments of the specification.
Example 1:
as shown in fig. 1, the present embodiment provides a bearing fault classification method based on an intelligent inspection robot, which includes the following steps:
s1, acquiring acoustic signals: the method comprises the steps that fault sounds of a bearing inner ring raceway, a bearing outer ring raceway, a bearing rolling body and a bearing assembly are collected by a collection card and a microphone based on an intelligent inspection robot, sound information of normal bearing working is collected, matlab is used for selecting the length of 12000 points to represent signal amplitude, obviously, the frequency and the amplitude of signals are obviously distinguished under 5 conditions, and the method is used for proving the feasibility of acoustic signals as bearing fault classification, and a waveform diagram is shown in attached figures 3-7; the data acquisition mainly acquires equipment information and a system running state, so that data recording and tracing are facilitated, and specific acquisition is realized by writing according to documents provided by acquisition equipment in order to realize an independent acquisition task. Since the operation of the acquisition card is essentially complete.
S2, constructing a time-frequency signal: constructing a time-frequency signal to perform data dimension-increasing signal feature expression;
s3, feature extraction: extracting features through a self-coding network;
s4, feature classification: carrying out feature classification through a convolutional neural network;
s5, optimizing and classifying: and training and testing by using an Adam optimization algorithm, and continuously adjusting parameter settings in the network so as to obtain a more accurate classification result.
As shown in fig. 2, the acoustic signal acquisition in step S1 of this embodiment is specifically as follows:
s101, connecting or finding a collecting device, wherein a signal source is collected to be a bearing and comprises a bearing inner ring raceway fault signal, a bearing outer ring raceway fault signal, a bearing rolling element fault signal and a bearing assembly fault signal;
s102, acquiring equipment state information: acquiring equipment information by adopting a get request mode, distinguishing equipment according to IP and MAC addresses and WebDAQ 504, and selecting a host address as manual input; for convenience of data analysis, converting the data into a json format;
s103, preliminarily judging whether the equipment is available:
(1) if yes, executing step S104;
(2) if not, jumping to the step S101;
s104, acquiring detailed information of the complete system for recording data sources;
s105, setting the sound collection task through a collection card, wherein the setting comprises conditions of sampling rate, channel number and sound pressure for setting microphone work, and issuing the collection task;
s106, acquiring the state of an equipment working list;
s107, judging whether a task is in progress or is suspended:
(1) if yes, stopping the existing tasks or waiting for queuing, and executing the step S108;
(2) if not, executing step S108;
s108, acquiring the running state of the equipment and executing tasks;
and S109, converting and outputting the related sound data of the corresponding acquisition equipment.
In this embodiment, while the status of the device work list is obtained in step S106, it is determined whether the current task list directory is legal:
and if the list task is legal, printing the information of the list task, wherein the information of the list task comprises the task name, the channel number and the progress condition of the task or automatically selecting the existing task for viewing.
The acquisition card in this embodiment adopts an internet of things sound and vibration data recorder WebDAQ 504 developed by MCC corporation.
In order to realize the complexity of the sound signals as much as possible, the microphone in the embodiment adopts two acoustic sensors, namely SKC AD7199 and SKC AD7018, to perform two-channel acquisition.
In this embodiment, the time-frequency signal is constructed in step S2 as follows:
effective information of data in a raw signal measured by a common sensor cannot be deeply explored, and the data after preprocessing of the data can highlight the characteristics expressed by the raw data.
Data is amplified in a one-dimensional time domain signal in a resampling preprocessing mode, and implicit characteristics of the signal are mined. The preprocessed signal tends to highlight the features represented by the raw data. The method expands the traditional time domain changed signals to a two-dimensional space domain in a windowing resampling mode, not only can enrich the signal characteristic representation, but also can utilize the existing network weight for analysis, and essentially, the correlation and the change index between the sampled data in the signals are key indexes in network classification.
In order to take account of the computational complexity of the edge device and the sufficiency of data preprocessing, the time-space frequency domain transformation is adopted, and the multi-scale frequency domain information is constructed by utilizing the fast Fourier transformation;
two-dimensionally unfolding the time sequence signals on the separated frequency domain scale in a windowing mode to form two-dimensional signal representation under multiple scales; the separated time sequence signals are subjected to resampling on a time sequence to obtain a time sequence, the time sequence is arranged in a space dimension to form preprocessed signals, and the formula is as follows:
wherein L (i) represents the length of the original signal; p (j, k) represents the representation of a single pixel point in the image; i =1, \ 8230;, M2; j =1, \ 8230;, M; k =1, \8230;, M;
the purpose of performing two-dimensional rearrangement on the data by adopting downward rounding is that the data is between 0 and 255, the one-dimensional vibration data is normalized into an image, actually, the data is converted into the current 8bit from the original 64bit, and the data storage space of a single point is greatly reduced; the method for windowing and resampling the data actually increases effective analysis data, so that the data is effectively analyzed on the basis of not occupying too large data storage space as much as possible, and the correlation among the data is extracted.
The step is to use the algorithm idea of image extraction on the whole, and directly use the acquired original signal without independent noise reduction, thereby realizing certain feature enhancement on the original signal dimension enhancement.
In this embodiment, the feature extraction performed through the self-coding network in step S3 is specifically as follows:
carrying out data denoising and dimensionality reduction in a self-coding mode after data dimensionality increasing: carrying out data denoising and dimensionality reduction by adopting a Tensorflow end-to-end open source machine learning platform and a Keras-based deep learning framework;
in self-coding, a time-frequency signal with characteristic dimensionality defined by 2400 multiplied by 1 and channel number of 1 is input, 10 filters are constructed by two-dimensional convolution, a 128 multiplied by 1 sliding window is adopted, in order to keep data information as much as possible, a filtering mode is set as "sea", and a tangent function tanh is selected as an activation function;
the down-sampling selection uses more maximum pooling, the size of the sampling factor is 8 multiplied by 1, and the step length is 2 to obtain a 1200 multiplied by 1 multiplied by 10 characteristic diagram;
the method comprises the steps of up-sampling a 1200X 1X 10 feature diagram, although the dimension of externally viewed data returns to the original 2400X 1X 10, performing interpolation operation without introducing other parameters, amplifying the feature diagram and purifying feature information; wherein, after adding the encode and decode operation, inserting a standard self-encoding (Autoencode) operation; the self-coding (auto encoder) operation is characterized in that the input and output feature quantity is the same, the input and output feature quantity is defined as the input data compression by a neural network and the reconstruction after the essence extraction, the feature maps are continuously reduced by the convolution layer and the sub-sampling layer through the feature extraction of multiple stacking, and the quantity of the feature maps is continuously increased;
at the end of the feature extractor, all feature maps are expanded and arranged into a vector, called a feature vector, which is used as the input of a post-classifier;
and (5) the signals are sent to a neural network classifier for training, so that the classification of the sound signals of the bearing fault is realized.
The feature classification performed by the convolutional neural network in step S4 in this embodiment is specifically as follows:
adam optimization algorithms are used in classification, simply Adam uses momentum and adaptive learning rates to speed up convergence. The method is suitable for the problem that the non-fixed target is suitable for very noisy or sparse gradient, and can replace a first-order optimization algorithm of the traditional random gradient descent process.
The learning process of the convolutional neural network aims to obtain a minimum error: label y of actual output ik And predicted label d ik The difference between should be reduced as much as possible, relying on the constant update of the weights ω and the deviations b during the training process; the loss function adopts a mean square error, the mean square error reflects the degree of closeness between probability distribution of various classes of softmax and actual classes, and if the difference between a model prediction result and the actual classes is small, the calculated mean square error is small;
carrying out classification test on the formed column vectors by using a trained network, and designing the column vectors step by step into 400 and 300 classes & ltcng & gtin sequence;
the method is finally divided into five types according to the category of the collected data, and the five types correspond to five different states of the collected signals;
wherein, the evaluation indexes of the classification comprise Accuracy (Accuracy), precision (Precision) and Recall rate (Recall); meanwhile, a parameter F1 is introduced for evaluating the quality of the model, and the formula is as follows:
F1＝2*(Precision*Recall)/(Precision+Recall)。
example 2:
the embodiment provides a bearing fault classification system based on an intelligent inspection robot, which comprises,
the signal acquisition module is used for acquiring fault sounds of the bearing inner ring raceway, the bearing outer ring raceway, the bearing rolling body and the bearing assembly based on an acquisition card and a microphone of the intelligent inspection robot and acquiring sound information of normal bearing operation;
the time-frequency signal construction module is used for constructing a time-frequency signal to perform data dimension-increasing signal feature expression;
the characteristic extraction module is used for extracting characteristics through a self-coding network;
the characteristic classification module is used for carrying out characteristic classification through a convolutional neural network;
and the optimization classification module is used for training and testing by using an Adam optimization algorithm, continuously adjusting parameter setting in the network and further acquiring a more accurate classification result.
The working process of the system is as follows:
(1) The method comprises the following steps of arranging an acquisition card, a microphone device and the like on an intelligent inspection robot, and calling the device to acquire bearing fault acoustic signals at the working condition position which needs to be actually acquired and analyzed;
(2) Continuously adjusting the parameter optimization training effect, finally packaging the classification model of the convolutional neural network into the intelligent inspection robot, and calling the model to perform feature extraction and classification on the acquired original signals;
the system can be set to be collected at regular time to realize full-automatic bearing fault monitoring.
The feature classification through the convolutional neural network is specifically as follows:
(1) decomposing the signal into signals under different frequency bands by utilizing fast Fourier change, expanding a data set in a time domain resampling mode, and expanding the data set in a two-dimensional mode to form two-dimensional signal representation under multiple scales;
(2) dividing a training set and a testing set, and setting a label for each sample;
(3) inputting the training sample obtained in the step (2) into a self-coding network for training to obtain the feature representation of the training and testing data of the last full-link layer, and finally calculating a loss function by using the features obtained from the training sample and the original label;
(4) and sending the test sample into the trained model, and verifying the effectiveness of the diagnosis model.
Example 3:
the present embodiment further provides an electronic device, including: a memory and a processor;
wherein the memory stores computer execution instructions;
the processor executes the computer-executable instructions stored in the memory, so that the processor executes the bearing fault classification method based on the intelligent inspection robot in any embodiment of the invention.
The processor may be a Central Processing Unit (CPU), but may also be other general purpose processors, digital Signal Processors (DSPs), application Specific Integrated Circuits (ASICs), off-the-shelf programmable gate arrays (FPGAs) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, and the like. The processor may be a microprocessor or the processor may be any conventional processor or the like.
The memory may be used to store computer programs and/or modules, and the processor may implement various functions of the electronic device by executing or executing the computer programs and/or modules stored in the memory and invoking data stored in the memory. The memory may mainly include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application program required for at least one function, and the like; the storage data area may store data created according to the use of the terminal, and the like. The memory may also include high speed random access memory, and may also include non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a memory only card (SMC), a Secure Digital (SD) card, a flash memory card, at least one disk storage period, a flash memory device, or other volatile solid state memory device.
Example 4:
the embodiment of the invention also provides a computer-readable storage medium, wherein a plurality of instructions are stored, and the instructions are loaded by the processor, so that the processor executes the bearing fault classification method based on the intelligent inspection robot in any embodiment of the invention. Specifically, a system or an apparatus equipped with a storage medium on which software program codes that realize the functions of any of the embodiments described above are stored may be provided, and a computer (or a CPU or MPU) of the system or the apparatus is caused to read out and execute the program codes stored in the storage medium.
In this case, the program code itself read from the storage medium can realize the functions of any of the above-described embodiments, and thus the program code and the storage medium storing the program code constitute a part of the present invention.
Examples of the storage medium for supplying the program code include a floppy disk, a hard disk, a magneto-optical disk, an optical disk (e.g., CD-ROM, CD-R, CD-RW, DVD-ROM, DVD-RYM, DVD-RW, DVD + RW), a magnetic tape, a nonvolatile memory card, and a ROM. Alternatively, the program code may be downloaded from a server computer by a communications network.
Further, it should be clear that the functions of any one of the above-described embodiments may be implemented not only by executing the program code read out by the computer, but also by causing an operating system or the like operating on the computer to perform a part or all of the actual operations based on instructions of the program code.
Further, it is to be understood that the program code read out from the storage medium is written to a memory provided in an expansion board inserted into the computer or to a memory provided in an expansion unit connected to the computer, and then causes a CPU or the like mounted on the expansion board or the expansion unit to perform part or all of the actual operations based on instructions of the program code, thereby realizing the functions of any of the above-described embodiments.
Finally, it should be noted that: the above embodiments are only used to illustrate the technical solution of the present invention, and not to limit the same; while the invention has been described in detail and with reference to the foregoing embodiments, it will be understood by those skilled in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some or all of the technical features may be equivalently replaced; and the modifications or the substitutions do not make the essence of the corresponding technical solutions depart from the scope of the technical solutions of the embodiments of the present invention.
Claims (10)
1. A bearing fault classification method based on an intelligent inspection robot is characterized by comprising the following steps:
s1, acquiring acoustic signals: the method comprises the steps that fault sounds of a bearing inner ring raceway, a bearing outer ring raceway, a bearing rolling body and a bearing assembly are collected based on a collection card and a microphone of an intelligent inspection robot, and sound information of normal bearing operation is collected;
s2, constructing a time-frequency signal: constructing a time-frequency signal to perform data dimension-increasing signal feature expression;
s3, feature extraction: extracting features through a self-coding network;
s4, feature classification: performing feature classification through a convolutional neural network;
s5, optimizing and classifying: and training and testing by using an Adam optimization algorithm, and continuously adjusting parameter setting in the network so as to obtain a more accurate classification result.
2. The intelligent inspection robot-based bearing fault classification method according to claim 1, wherein the acoustic signals collected in the step S1 are specifically as follows:
s101, connecting or finding a collecting device, wherein a signal source is collected to be a bearing and comprises a bearing inner ring raceway fault signal, a bearing outer ring raceway fault signal, a bearing rolling element fault signal and a bearing assembly fault signal;
s102, acquiring equipment state information: acquiring equipment information by adopting a get request mode, distinguishing WebDAQ 504 distinguishing equipment according to IP and MAC addresses, selecting a host address as manual input, and converting data into a json format;
s103, preliminarily judging whether the equipment is available:
(1) if yes, executing step S104;
(2) if not, jumping to the step S101;
s104, acquiring detailed information of the complete system;
s105, setting the sound collection task through a collection card, wherein the setting comprises conditions of sampling rate, channel number and sound pressure for setting microphone work, and issuing the collection task;
s106, acquiring the state of an equipment working list;
s107, judging whether a task is in progress or is suspended:
(1) if yes, stopping the existing task or waiting for queuing, and executing the step S108;
(2) if not, executing step S108;
s108, acquiring the running state of the equipment and executing tasks;
and S109, converting and outputting the related sound data of the corresponding acquisition equipment.
3. The intelligent inspection robot-based bearing fault classification method according to claim 2, wherein the status of the equipment work list is obtained while judging whether the current task list directory is legal:
and if the list task is legal, printing the information of the list task, wherein the information of the list task comprises a task name, a channel number and the progress condition of the task or automatically selecting the existing task for viewing.
4. The intelligent inspection robot-based bearing fault classification method according to claim 1, wherein the acquisition card employs an internet of things sound and vibration data recorder WebDAQ 504;
the microphone adopts two acoustic sensors of SKC AD7199 and SKC AD7018 to carry out double-channel acquisition.
5. The intelligent inspection robot-based bearing fault classification method according to claim 1, wherein the time-frequency signals constructed in the step S2 are specifically as follows:
the method adopts space-time frequency domain transformation and utilizes fast Fourier transformation to construct multi-scale frequency domain information;
two-dimensionally unfolding the time sequence signals on the separated frequency domain scale in a windowing mode to form two-dimensional signal representation under multiple scales; the separated time sequence signals are subjected to resampling on a time sequence to obtain a time sequence, the time sequence is arranged in a space dimension to form preprocessed signals, and the formula is as follows:
wherein L (i) represents the length of the original signal; p (j, k) represents the representation of a single pixel point in the image; i =1, \8230;, M2; j =1, \ 8230;, M; k =1, \ 8230;, M;
the purpose of rounding down for the two-dimensional rearrangement of the data is that the data is between 0 and 255, the one-dimensional vibration data becomes an image through normalization, and actually the data is converted into the current 8 bits from the original 64 bits.
6. The intelligent inspection robot-based bearing fault classification method according to claim 1, wherein the feature extraction through the self-coding network in the step S3 is specifically as follows:
carrying out data denoising and dimensionality reduction in a self-coding mode after data dimensionality increasing: carrying out data denoising and dimensionality reduction by adopting a Tensorflow end-to-end open source machine learning platform and a Keras-based deep learning framework;
in self-encoding, a time-frequency signal with characteristic dimension definition of 2400 multiplied by 1 and channel number of 1 is input, 10 filters are constructed by two-dimensional convolution, a 128 multiplied by 1 sliding window is adopted, a filter mode is set to 'team', and an activation function selects a double tangent function tanh;
the maximum pooling is selected for the down-sampling, the size of the sampling factor is 8 multiplied by 1, and the step length is 2 to obtain a 1200 multiplied by 1 multiplied by 10 characteristic diagram;
up-sampling the 1200 × 1 × 10 feature map, amplifying the feature map, and purifying feature information; after adding the encode and decode operation, inserting a standard self-encoding operation; the self-coding operation is characterized in that the input and output feature quantity is the same, the self-coding operation is defined as the reconstruction of input data after compression and essence extraction by using a neural network, feature graphs are continuously reduced by a convolution layer and a sub-sampling layer through feature extraction stacked for multiple times, and the quantity of the feature graphs is continuously increased;
at the end of the feature extractor, all feature maps are expanded and arranged into a vector, called a feature vector, and the feature vector is used as the input of a rear-layer classifier;
and (5) the signals are sent to a neural network classifier for training, so that the classification of the sound signals of the bearing fault is realized.
7. The intelligent inspection robot-based bearing fault classification method according to any one of claims 1 to 6, wherein the feature classification through the convolutional neural network in the step S4 is specifically as follows:
the learning process of the convolutional neural network aims to obtain a minimum error: label y of actual output ik And predicted label d ik The difference between should be reduced as much as possible, relying on the constant update of the weights ω and the deviations b during the training process; the loss function adopts a mean square error, the mean square error reflects the degree of closeness between probability distribution of various classes of softmax and actual classes, and if the difference between a model prediction result and the actual classes is small, the calculated mean square error is small;
carrying out classification test on the formed column vectors by using a trained network step by step, and designing the classification test into 400 and 300 classes-;
the method is finally divided into five types according to the category of the collected data, and the five types correspond to five different states of the collected signals;
wherein, the evaluation indexes of the classification comprise accuracy, precision and recall rate; meanwhile, a parameter F1 is introduced for evaluating the quality of the model, and the formula is as follows:
F1＝2*(Precision*Recall)/(Precision+Recall)。
8. a bearing fault classification system based on an intelligent inspection robot is characterized by comprising,
the signal acquisition module is used for acquiring fault sounds of the bearing inner ring raceway, the bearing outer ring raceway, the bearing rolling body and the bearing assembly based on an acquisition card and a microphone of the intelligent inspection robot and acquiring sound information of normal bearing operation;
the time-frequency signal construction module is used for constructing a time-frequency signal to perform data dimension-increasing signal feature expression;
the characteristic extraction module is used for extracting characteristics through a self-coding network;
the characteristic classification module is used for carrying out characteristic classification through a convolutional neural network;
and the optimization classification module is used for training and testing by using an Adam optimization algorithm, continuously adjusting parameter setting in the network and further acquiring a more accurate classification result.
9. An electronic device, comprising: a memory and at least one processor;
wherein the memory has stored thereon a computer program;
the at least one processor executes the computer program stored in the memory to cause the at least one processor to perform the intelligent inspection robot-based bearing fault classification method according to any one of claims 1 to 7.
10. A computer-readable storage medium, in which a computer program is stored, the computer program being executable by a processor to implement the intelligent inspection robot-based bearing fault classification method according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211367420.4A CN115758237A (en) | 2022-11-03 | 2022-11-03 | Bearing fault classification method and system based on intelligent inspection robot |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211367420.4A CN115758237A (en) | 2022-11-03 | 2022-11-03 | Bearing fault classification method and system based on intelligent inspection robot |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115758237A true CN115758237A (en) | 2023-03-07 |
Family
ID=85356209
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211367420.4A Pending CN115758237A (en) | 2022-11-03 | 2022-11-03 | Bearing fault classification method and system based on intelligent inspection robot |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115758237A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117668528A (en) * | 2024-02-01 | 2024-03-08 | 成都华泰数智科技有限公司 | Natural gas voltage regulator fault detection method and system based on Internet of things |
-
2022
- 2022-11-03 CN CN202211367420.4A patent/CN115758237A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117668528A (en) * | 2024-02-01 | 2024-03-08 | 成都华泰数智科技有限公司 | Natural gas voltage regulator fault detection method and system based on Internet of things |
| CN117668528B (en) * | 2024-02-01 | 2024-04-12 | 成都华泰数智科技有限公司 | Natural gas voltage regulator fault detection method and system based on Internet of things |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Priyadarshani et al. | Automated birdsong recognition in complex acoustic environments: a review | |
| CN110047512B (en) | Environmental sound classification method, system and related device | |
| CN105971901A (en) | Centrifugal pump fault diagnosis method based on complete ensemble empirical mode decomposition and random forest | |
| EP3767551A1 (en) | Inspection system, image recognition system, recognition system, discriminator generation system, and learning data generation device | |
| CN111508480B (en) | Training method of audio recognition model, audio recognition method, device and equipment | |
| CN113314144A (en) | Voice recognition and power equipment fault early warning method, system, terminal and medium | |
| CN112131907A (en) | Method and device for training classification model | |
| Islam et al. | Motor bearing fault diagnosis using deep convolutional neural networks with 2d analysis of vibration signal | |
| CN115798516B (en) | Migratable end-to-end acoustic signal diagnosis method and system | |
| JP2020042519A (en) | Abnormality detection device, abnormality detection method, and abnormality detection program | |
| CN115758237A (en) | Bearing fault classification method and system based on intelligent inspection robot | |
| CN114638633A (en) | Abnormal flow detection method and device, electronic equipment and storage medium | |
| CN116012681A (en) | Method and system for diagnosing motor faults of pipeline robot based on sound vibration signal fusion | |
| JP4760614B2 (en) | Method for selecting learning data of signal identification device | |
| JP2007327885A (en) | Abnormality determination method and abnormality determination device | |
| Verma et al. | Android app for intelligent CBM | |
| JP4886460B2 (en) | Abnormality monitoring device | |
| Coelho et al. | Deep dense and convolutional autoencoders for machine acoustic anomaly detection | |
| JP2008040684A (en) | Learning method of signal identification device | |
| CN116659856A (en) | Fault diagnosis method of motor bearing based on WPT-1DCNN | |
| Hao et al. | New fusion features convolutional neural network with high generalization ability on rolling bearing fault diagnosis | |
| CN112700016B (en) | Predictive maintenance method and apparatus | |
| JP4513796B2 (en) | Abnormality monitoring device | |
| CN115062685A (en) | Fault diagnosis method and device, electronic equipment and storage medium | |
| CN114186593A (en) | Bearing fault identification method, device, equipment and storage medium under noise condition |
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 |