CN111680665B - Motor mechanical fault diagnosis method adopting current signals based on data driving - Google Patents
Motor mechanical fault diagnosis method adopting current signals based on data driving Download PDFInfo
- Publication number
- CN111680665B CN111680665B CN202010597800.1A CN202010597800A CN111680665B CN 111680665 B CN111680665 B CN 111680665B CN 202010597800 A CN202010597800 A CN 202010597800A CN 111680665 B CN111680665 B CN 111680665B
- Authority
- CN
- China
- Prior art keywords
- signal
- current
- layer
- reconstruction
- current signal
- 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.)
- Active
Links
Images
Classifications
-
- 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
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/02—Preprocessing
- G06F2218/04—Denoising
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
- Y04S10/52—Outage or fault management, e.g. fault detection or location
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Life Sciences & Earth Sciences (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Mathematical Physics (AREA)
- Medical Informatics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Control Of Electric Motors In General (AREA)
Abstract
The invention discloses a motor mechanical fault diagnosis method adopting a current signal based on data driving, which comprises the steps of obtaining a stator current signal, wherein the length of the stator current signal is D and N samples are included; carrying out noise signal reconstruction on the stator current signal to obtain a reconstructed signal; subtracting the reconstruction signal from the original stator current signal to obtain residual current; converting the residual current signal into a frequency domain to obtain a frequency domain signal; and inputting the frequency domain signal of the residual current into a pre-trained machine learning feature extraction and fault classification model to obtain the motor state. The invention adopts the current signal to diagnose the motor fault, can be used for diagnosing mechanical faults, does not need to additionally install a vibration sensor and a current sensor, can collect data through a current transformer which is already installed in a protection system or a control system, is a non-invasive low-cost mode, and has the advantages of high diagnosis efficiency, high diagnosis accuracy and low false alarm rate.
Description
Technical Field
The invention relates to a motor mechanical fault diagnosis technology, in particular to a motor mechanical fault diagnosis method adopting current signals based on data driving.
Background
Motor faults can be divided into mechanical faults and electrical faults, and vibration signals are used for diagnosis aiming at the mechanical faults in the prior art. However, the price of the vibration sensor is high, and the application of the vibration sensor in small and medium-sized motors is limited. And the vibration sensor needs to be in direct contact with the motor, so that the problem of installation and placement exists, and some devices which are put into production can not be additionally provided with the vibration sensor. Therefore, the conventional vibration-based sensor has a problem of limited application range.
Disclosure of Invention
The invention aims to solve the technical problems: aiming at the problems in the prior art, the invention provides a motor mechanical fault diagnosis method based on data driving and adopting current signals, and the motor mechanical fault diagnosis method adopts the current signals to diagnose the motor faults, and can be used for diagnosing the mechanical faults.
In order to solve the technical problems, the invention adopts the following technical scheme:
a data-driven electromechanical fault diagnosis method employing a current signal, comprising:
1) Acquiring a stator current signal, wherein the length of the stator current signal is D and N samples are included;
2) Carrying out noise signal reconstruction on the stator current signal to obtain a reconstructed signal;
3) Subtracting the reconstruction signal from the original stator current signal to obtain residual current;
4) Converting the time domain signal of the residue current into a frequency domain to obtain a frequency domain signal;
5) And inputting the frequency domain signal of the residual current into a pre-trained machine learning feature extraction and fault classification model to obtain the motor state.
Optionally, the step of converting the time domain signal of the residue current to the frequency domain in step 4) includes: first, a Hilbert envelope spectrum of the residue current is calculated, and then an FFT spectrum of the Hilbert envelope spectrum is calculated.
Optionally, in step 2), the noise signal reconstruction of the stator current signal to obtain the reconstructed signal adopts a signal noise reconstruction model built by an automatic encoder, and the training signal noise reconstruction model comprises an input layer and an hidden layer which are sequentially connectedAn output layer, wherein the input layer comprises D+1 nodes, the last node is a bias node, and a hidden layer D h +1 nodes, and the last node is a bias node, and the output layer comprises D nodes; the input layer and the hidden layer form an encoder, the hidden layer and the output layer form a decoder, and the weight matrix of the encoder and the decoder are respectively W 1 (D h X D) and W 2 (D×D h ),D h Is the dimension of the hidden layer; the input of the input layer is as follows: x= { X (1) ,x (2) ,…,x (N) X is the input stator current signal, X comprises X (1) ~x (N) N samples, any ith sample x (i) Is expressed as x (i) =[x 1 ,x 2 ,…,x D ]Wherein x is 1 ~x D Respectively sampling values of stator current signals; the functional expression of the hidden layer is: h is a (i) =f(W 1 x (i) +b 1 ) Wherein h is (i) For the ith hidden layer, f is the activation function, W 1 Weight vector, x, for the ith hidden layer (i) For the ith sample, b 1 Is a bias vector; the functional expression of the output layer is:wherein->For the ith output layer,/th>To activate the function, W 2 For the weight vector of the ith output layer, h (i) For the ith hidden layer, b 2 Is a bias vector.
Optionally, step 2) is preceded by the step of training a signal noise reconstruction model: collecting stator current signals of a motor in a health state and N motor mechanical faults and forming a sample library, wherein the length of each sample is D, and each motor state comprises N samples, so that the total amount of data is D, N (n+1); training signal noise through motor samples in health state in sample libraryThe reconstruction model is trained, and the training target is the input signal X and the reconstruction signal thereofIs the least mean square error.
Optionally, the machine learning feature extraction and fault classification model in step 5) includes a feature extraction model and a fault classification model, where the feature extraction model is formed by stacking M automatic encoders to implement automatic feature extraction, where an input of a bottom automatic encoder is a frequency domain signal of residual current, and an output end of a top automatic encoder is connected to a Softmax classifier to form the fault classification model, and the Softmax classifier is used for outputting a motor state obtained by classification.
Optionally, step 5) further includes a step of training machine learning feature extraction and fault classification model composed of M automatic encoder stacks: inputting the frequency domain signals of residual current obtained by sample extraction in a sample library into a machine learning feature extraction model and a feature extraction model in a fault classification model to extract features, then finely adjusting parameters of each layer of automatic encoder in the feature extraction model and a Softmax classifier in the fault classification model through a back propagation algorithm according to the motor state obtained by classifying the features by the Softmax classifier, and repeating the process until the iteration times are equal to a preset threshold or the error is lower than the preset threshold.
In addition, the invention also provides a motor mechanical fault diagnosis system adopting current signals based on data driving, which comprises the following components:
a signal acquisition program unit for acquiring a stator current signal;
the noise reconstruction program unit is used for carrying out noise signal reconstruction on the stator current signal to obtain a reconstructed signal;
the noise elimination program unit is used for subtracting the reconstruction signal from the original stator current signal to obtain residual current;
converting the time domain signal of the residue current into a frequency domain to obtain a frequency domain signal; and the feature extraction and classification diagnosis program unit is used for inputting the frequency domain signal of the residual current into a pre-trained machine learning feature extraction and fault classification model to obtain the motor state.
In addition, the invention also provides a data-driven electromechanical fault diagnosis system adopting the current signal, which comprises computer equipment, wherein the computer equipment is programmed or configured to execute the steps of the data-driven electromechanical fault diagnosis method adopting the current signal.
In addition, the invention also provides a data-driven electromechanical fault diagnosis system adopting the current signal, which comprises computer equipment, wherein a memory of the computer equipment is stored with a computer program which is programmed or configured to execute the data-driven electromechanical fault diagnosis method adopting the current signal.
Furthermore, the present invention provides a computer-readable storage medium having stored therein a computer program programmed or configured to perform the data-driven electromechanical fault diagnosis method employing a current signal.
Compared with the prior art, the invention has the following advantages:
1. for mechanical faults, vibration signals are mostly used for diagnosis in the prior art. However, the price of the vibration sensor is high, and the application of the vibration sensor in small and medium-sized motors is limited. And the vibration sensor needs to be in direct contact with the motor, so that the problem of installation and placement exists, and some devices which are put into production can not be additionally provided with the vibration sensor. The method for diagnosing the motor faults by adopting the current signals can be used for diagnosing mechanical faults. The method does not need to additionally install a vibration sensor and a current sensor, can collect data through a current transformer which is already installed in a protection system or a control system, and is a non-invasive low-cost mode.
2. In the prior art, a model-based method is mostly adopted, and characteristics are manually extracted through fault mechanism analysis and signal processing means. Because the motor is a complex electromechanical-magnetic coupling system, the modeling difficulty is high, and the requirement on professional knowledge is high. Meanwhile, the fault is likely to generate a linking effect, and the difficulty of building an accurate model is increased. And the high-order model has the defects of sensitive parameters, low robustness and the like. The motor fault diagnosis method based on data driving can utilize motor operation data to mine information contained in the data, automatically extract fault characteristics which are difficult to model, has high robustness, reduces requirements on professional knowledge and improves diagnosis efficiency.
3. When diagnosis is performed based on a data driving mode, if original time domain data or data converted into a frequency domain are directly used, the accuracy of diagnosis is low, and the false alarm rate is high. Because the mechanical fault is reflected in the fault information in the current, the fault information can be modulated by the current fundamental wave, the signal to noise ratio is low, and the noise interference is large. The signal has not only fundamental wave and odd harmonics thereof, but also unknown components such as slot harmonics, saturated harmonics, environmental noise and the like. The fault characteristics are very unobvious, and the fault diagnosis difficulty is high. The invention provides a data-driven noise elimination method. The "noise signal" is defined as a component that is independent of the fault signature. And the operation data of the healthy motor is utilized to build a reconstruction model of the current noise signal, the noise component is automatically extracted, and the signal-to-noise ratio of the signal is greatly improved. The accuracy of the subsequent fault feature extraction and classification algorithm based on data driving is improved, and the false alarm rate is reduced.
Drawings
FIG. 1 is a schematic diagram of a basic flow of a method according to an embodiment of the present invention.
Fig. 2 is a schematic diagram of a method according to an embodiment of the present invention.
Fig. 3 is a schematic diagram of an automatic encoder according to an embodiment of the present invention.
Fig. 4 is a schematic structural diagram of a machine learning feature extraction and fault classification model in an embodiment of the present invention.
Detailed Description
The principle of the motor mechanical fault diagnosis method based on data driving and adopting current signals is as follows:
when a motor is subjected to typical mechanical faults (rotor breakage, eccentric faults, bearing faults), the motor generates periodic torque pulsation in operation, thereby influencing current. Each fault will appear at a specific frequency in the current spectrum, namely the fault signature frequency, as follows:
characteristic frequency f of rotor broken bar fault brb1 And f brb2 The method comprises the following steps:
f brb1 =(1±2k 1 s)·f s ,k 1 =1,2,3,… (1)
f brb2 =[(k 2 /p)(1-s)±s]·f s ,k 2 /p=1,2,3,… (2)
frequency of eccentric failure ecc1 The method comprises the following steps:
f ecc1 =[(k·R±n d )·(1-s)/P p ±v]·f s ,k=1,2,3,…v=1,3,5,… (3)
n d when=0, the static eccentric fault is formed, n d When=1, 2,3, …, the dynamic eccentric failure occurs.
Bearing failure characteristic frequency:
characteristic frequency f at inner ring failure I The method comprises the following steps: f (f) I =(N b /2)·f s ·[1+b d ·cos(β)/d p ]/P p (4)
Characteristic frequency f of outer ring failure O The method comprises the following steps: f (f) O =(N b /2)·f s ·[1-b d ·cos(β)/d p ]/P p (5)
Characteristic frequency f of rolling element failure B The method comprises the following steps: f (f) B =d p ·f s {1-[b d ·cos(β)/d p ] 2 }/2/b d /P p (6)
In the formulae (1) to (6), f s Is the power frequency, s is the slip, R is the number of rotor grooves, P p Is the polar logarithm, N b Is the number of rolling bodies, b d Is the diameter of the rolling element d p Is the rolling element pitch diameter and β is the contact angle.
According to the above formulas (1) to (6), the characteristic frequency of each fault can be reduced to formula (7), where f c Is a general representation of the above-mentioned characteristic frequencies of the faults.
f=f s ±k·f c ,k=1,2,3,… (7)
From equation (7), the diagnosis of the mechanical failure of the motor can be accomplished by detecting a specific frequency. However, the above formula is derived based on the theory of an ideal motor, and under actual conditions, the current has odd harmonics besides the fundamental wave, and other unknown source components caused by slot harmonics, saturated harmonics, environmental noise and the like. These components are independent of fault characteristics and their presence may interfere with the extraction and discrimination of fault characteristics, so that the current fundamental wave and other unknown components are collectively defined as "noise signals" in the data-driven electromechanical fault diagnosis method employing current signals according to the present invention.
As shown in fig. 1, the electromechanical fault diagnosis method using a current signal based on data driving of the present embodiment includes:
1) Acquiring a stator current signal, wherein the length of the stator current signal is D and N samples are included;
2) Carrying out noise signal reconstruction on the stator current signal to obtain a reconstructed signal;
3) Subtracting the reconstruction signal from the original stator current signal to obtain residual current;
4) Converting the time domain signal of the residue current into a frequency domain to obtain a frequency domain signal;
5) And inputting the frequency domain signal of the residual current into a pre-trained machine learning feature extraction and fault classification model to obtain the motor state.
Referring to fig. 1, the method of the present embodiment adopts a data-driven mode, and utilizes motor stator current data to complete automatic noise elimination and fault feature extraction and identification. And during the period, the motor mechanical fault diagnosis can be realized by simply processing the denoised residual current. The method mainly comprises five steps of motor stator current signal acquisition, noise reconstruction, noise elimination, frequency domain conversion, feature extraction and classification diagnosis.
In this embodiment, a signal which is independent of the fault characteristics is defined as a noise signal. Referring to fig. 2 and 3, in step 2) of the present embodiment, the noise signal reconstruction of the stator current signal to obtain the reconstructed signal uses a signal noise reconstruction model built by an automatic encoder, and the training signal noise reconstruction model includes an input layer and an hidden layer which are sequentially connectedAn output layer, wherein the input layer comprises D+1 nodes, the last node is a bias node, and a hidden layer D h +1 nodes, and the last node is a bias node, and the output layer comprises D nodes; the input layer and the hidden layer form an encoder, the hidden layer and the output layer form a decoder, and the weight matrix of the encoder and the decoder are respectively W 1 (D h X D) and W 2 (D×D h ),D h Is the dimension of the hidden layer; the input of the input layer is as follows: x= { X (1) ,x (2) ,…,x (N) X is the input stator current signal, X comprises X (1) ~x (N) N samples, any ith sample x (i) Is expressed as x (i) =[x 1 ,x 2 ,…,x D ](represented by x by taking sample 1 as an example (1) =[x 1 ,x 2 ,…,x D ]) Wherein x is 1 ~x D Respectively sampling values of stator current signals; the functional expression of the hidden layer is: h is a (i) =f(W 1 x (i) +b 1 ) Wherein h is (i) For the ith hidden layer, f is the activation function, W 1 Weight vector, x, for the ith hidden layer (i) For the ith sample, b 1 Is a bias vector; the functional expression of the output layer is:wherein->For the ith output layer,/th>To activate the function, W 2 For the weight vector of the ith output layer, h (i) For the ith hidden layer, b 2 Is a bias vector. Taking sample 1 as an example, the output is the reconstructed signal for this sample +.>
Referring to fig. 2 and 3, step 2) in this embodiment is also followed byTraining a signal noise reconstruction model: collecting stator current signals of a motor in a health state and N motor mechanical faults and forming a sample library, wherein the length of each sample is D, and each motor state comprises N samples, so that the total amount of data is D, N (n+1); training a training signal noise reconstruction model through samples of a motor in a healthy state in a sample library, wherein the training target is an input signal X and a reconstruction signal thereofThe least mean square error (reconstruction error), i.e.:
in the above formula, J is the mean square error, X is the input signal,is the reconstructed signal of X. Upon reconstruction of healthy motor current data, only the fundamental and other unknown source components are included as no fault signature is included in the data. That is, the reconstructed signal is a noise signal. The method can automatically and effectively extract the noise signals through the operation data without establishing a complex mathematical model. And inputting the motor current signal into the trained reconstruction model to obtain a corresponding reconstruction signal. Current signal X of faulty motor F Inputting the fault motor current into a trained model to obtain a reconstructed signal of the fault motor current under the modelSince the model is a reconstruction of the noise signal, the current signal of the faulty motor passes through the model, and its unique fault signal is not reconstructed.
Step 3) subtracting the reconstruction signal from the original stator current signal to obtain residual current, subtracting the reconstruction signal from the original motor (health and fault) signal to eliminate noise and obtain corresponding residual current signals R and R F Wherein R represents healthResidual current obtained by motor, R F Representing the residual current drawn by the faulty motor.
Referring to fig. 2, the step of converting the time domain signal of the residue current into the frequency domain in step 4) to obtain a frequency domain signal includes: first a Hilbert envelope spectrum of the residue current is calculated, and then an FFT (fourier) spectrum of the Hilbert envelope spectrum is calculated.
Referring to fig. 4, the machine learning feature extraction and fault classification model in step 5) includes a feature extraction model and a fault classification model, wherein the feature extraction model is formed by stacking M automatic encoders to implement automatic feature extraction, wherein the input of the lowest automatic encoder is a frequency domain signal of residual current, and the output end of the top automatic encoder is connected with a Softmax classifier to form the fault classification model, and the Softmax classifier is used for outputting the motor state obtained by classification.
Referring to fig. 4, step 5) further includes a machine learning feature extraction and fault classification model step of training M automatic encoder stacks: inputting the frequency domain signals of residual current obtained by sample extraction in a sample library into a machine learning feature extraction model and a feature extraction model in a fault classification model to extract features, then finely adjusting parameters of each layer of automatic encoder in the feature extraction model and a Softmax classifier in the fault classification model through a back propagation algorithm according to the motor state obtained by classifying the features by the Softmax classifier, and repeating the process until the iteration times are equal to a preset threshold or the error is lower than the preset threshold.
As shown in fig. 4, the feature extraction and classification model of the residual current is built by stacking an automatic encoder algorithm in this embodiment. Stacked auto encoders are a type of deep learning algorithm that is stacked from multiple auto encoders. The first layer automatic encoder inputs are residue currents R and R F Envelope spectrum X of (2) P Outputs a reconstructed signal thereforObtaining hidden layer H (1) . The hidden layer H (1) As input to the next layer auto encoder and performs layer-by-layer training. Such layer-by-layer pretrainingBy means of the method, the depth of an algorithm can be increased, and the condition that gradients disappear due to a multi-layer structure is avoided. The hidden layers obtained are H (1) ,H (2) ,…,H (M) M is the number of layers of the stacked auto encoder. And adding a Softmax classifier on the top layer of the trained model, and fine-tuning parameters of each layer and parameters of the classifier through a back propagation algorithm to complete fault classification training. The diagnosis of the motor state can be realized.
In summary, the method for diagnosing the mechanical fault of the motor using the current signal based on the data driving according to the embodiment can detect the mechanical fault by using the current signal, and does not need an additional sensor, which is a non-invasive low-cost way. Based on a data driving mode, the noise characteristic irrelevant to faults is automatically extracted to eliminate power fundamental waves and odd harmonics thereof and other harmonics which are not easy to model in the original signals. The automatic noise elimination processing based on data driving can greatly improve the signal to noise ratio, avoid a complex fault modeling process and reduce the requirement on professional knowledge. After the denoised signal is subjected to simple signal processing, the fault characteristics are automatically extracted and the classification identification of the faults is completed by combining a deep learning algorithm. The invention fully utilizes the advantages of the data driving method, improves the diagnosis accuracy, reduces the false alarm rate and the requirements on professional knowledge, and improves the diagnosis efficiency.
In addition, the embodiment also provides a motor mechanical fault diagnosis system adopting current signals based on data driving, which comprises the following components:
a signal acquisition program unit for acquiring a stator current signal;
the noise reconstruction program unit is used for carrying out noise signal reconstruction on the stator current signal to obtain a reconstructed signal;
the noise elimination program unit is used for subtracting the reconstruction signal from the original stator current signal to obtain residual current;
converting the time domain signal of the residue current into a frequency domain to obtain a frequency domain signal;
and the feature extraction and classification diagnosis program unit is used for inputting the frequency domain signal of the residual current into a pre-trained machine learning feature extraction and fault classification model to obtain the motor state.
The present embodiment also provides a data-driven electromechanical fault diagnosis system employing a current signal, comprising a computer device programmed or configured to perform the steps of the foregoing data-driven electromechanical fault diagnosis method employing a current signal.
The present embodiment also provides a data-driven electromechanical fault diagnosis system using a current signal, including a computer device having a memory stored with a computer program programmed or configured to perform the foregoing data-driven electromechanical fault diagnosis method using a current signal.
The present embodiment further provides a computer-readable storage medium having stored therein a computer program programmed or configured to perform the foregoing data-driven electromechanical fault diagnosis method employing a current signal.
The above description is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above examples, and all technical solutions belonging to the concept of the present invention belong to the protection scope of the present invention. It should be noted that modifications and adaptations to the present invention may occur to one skilled in the art without departing from the principles of the present invention and are intended to be within the scope of the present invention.
Claims (9)
1. A method for diagnosing a motor mechanical fault using a current signal based on data driving, comprising:
1) Acquiring a stator current signal, wherein the length of the stator current signal is D and N samples are included;
2) Carrying out noise signal reconstruction on the stator current signal to obtain a reconstructed signal; the method comprises the steps that noise signal reconstruction is carried out on a stator current signal to obtain a reconstructed signal, a signal noise reconstruction model built by an automatic encoder is adopted, the training signal noise reconstruction model comprises an input layer, an hidden layer and an output layer which are sequentially connected, the input layer comprises D+1 nodes, and the last one is the last oneThe node is a bias node, implying layer D h +1 nodes, and the last node is a bias node, and the output layer comprises D nodes; the input layer and the hidden layer form an encoder, the hidden layer and the output layer form a decoder, and the weight matrix of the encoder and the decoder are respectively W 1 (D h X D) and W 2 (D×D h ),D h Is the dimension of the hidden layer; the input of the input layer is as follows: x= { X (1) ,x (2) ,…,x (N) X is the input stator current signal, X comprises X (1) ~x (N) N samples, any ith sample x (i) Is expressed as x (i) =[x 1 ,x 2 ,…,x D ]Wherein x is 1 ~x D Respectively sampling values of stator current signals; the functional expression of the hidden layer is: h is a (i) =f(W 1 x (i) +b 1 ) Wherein h is (i) For the ith hidden layer, f is the activation function, W 1 Weight vector, x, for the ith hidden layer (i) For the ith sample, b 1 Is a bias vector; the functional expression of the output layer is:wherein->For the ith output layer,/th>To activate the function, W 2 For the weight vector of the ith output layer, h (i) For the ith hidden layer, b 2 Is a bias vector;
3) Subtracting the reconstruction signal from the original stator current signal to obtain residual current;
4) Converting the time domain signal of the residue current into a frequency domain to obtain a frequency domain signal;
5) And inputting the frequency domain signal of the residual current into a pre-trained machine learning feature extraction and fault classification model to obtain the motor state.
2. The method for diagnosing a motor mechanical failure using a current signal based on data driving as recited in claim 1, wherein the step of converting the time domain signal of the residual current to the frequency domain in the step 4) includes: first, a Hilbert envelope spectrum of the residue current is calculated, and then an FFT spectrum of the Hilbert envelope spectrum is calculated.
3. The method for diagnosing a motor mechanical failure using a current signal based on data driving as recited in claim 1, wherein the step 2) is preceded by the step of training a signal noise reconstruction model: collecting stator current signals of a motor in a health state and N motor mechanical faults and forming a sample library, wherein the length of each sample is D, and each motor state comprises N samples, so that the total amount of data is D, N (n+1); training a training signal noise reconstruction model through samples of a motor in a healthy state in a sample library, wherein the training target is an input signal X and a reconstruction signal thereofIs the least mean square error.
4. The method for diagnosing motor mechanical faults using current signals based on data driving according to claim 1, wherein the machine learning feature extraction and fault classification model in the step 5) comprises a feature extraction model and a fault classification model, wherein the feature extraction model is formed by stacking M automatic encoders to achieve automatic feature extraction, wherein the input of the automatic encoder at the bottom layer is the frequency domain signal of residual current, and the output end of the automatic encoder at the top layer is connected with a Softmax classifier to form the fault classification model, and the Softmax classifier is used for outputting the motor state obtained by classification.
5. The method for data-driven electromechanical fault diagnosis using current signals according to claim 4, further comprising the step of training machine learning feature extraction and fault classification model composed of M automatic encoder stacks before step 5): inputting the frequency domain signals of residual current obtained by sample extraction in a sample library into a machine learning feature extraction model and a feature extraction model in a fault classification model to extract features, then finely adjusting parameters of each layer of automatic encoder in the feature extraction model and a Softmax classifier in the fault classification model through a back propagation algorithm according to the motor state obtained by classifying the features by the Softmax classifier, and repeating the process until the iteration times are equal to a preset threshold or the error is lower than the preset threshold.
6. A data-driven electromechanical fault diagnosis system employing a current signal, comprising:
a signal acquisition program unit for acquiring a stator current signal;
the noise reconstruction program unit is used for carrying out noise signal reconstruction on the stator current signal to obtain a reconstructed signal; the method comprises the steps that noise signal reconstruction is carried out on a stator current signal to obtain a reconstructed signal, a signal noise reconstruction model built by an automatic encoder is adopted, the training signal noise reconstruction model comprises an input layer, an hidden layer and an output layer which are sequentially connected, the input layer comprises D+1 nodes, the last node is a bias node, and the hidden layer D h +1 nodes, and the last node is a bias node, and the output layer comprises D nodes; the input layer and the hidden layer form an encoder, the hidden layer and the output layer form a decoder, and the weight matrix of the encoder and the decoder are respectively W 1 (D h X D) and W 2 (D×D h ),D h Is the dimension of the hidden layer; the input of the input layer is as follows: x= { X (1) ,x (2) ,…,x (N) X is the input stator current signal, X comprises X (1) ~x (N) N samples, any ith sample x (i) Is expressed as x (i) =[x 1 ,x 2 ,…,x D ]Wherein x is 1 ~x D Respectively sampling values of stator current signals; the functional expression of the hidden layer is: h is a (i) =f(W 1 x (i) +b 1 ) Wherein h is (i) For the ith hidden layer, f is the activation function, W 1 The weighting direction for the ith hidden layerQuantity, x (i) For the ith sample, b 1 Is a bias vector; the functional expression of the output layer is:wherein the method comprises the steps ofFor the ith output layer,/th>To activate the function, W 2 For the weight vector of the ith output layer, h (i) For the ith hidden layer, b 2 Is a bias vector;
the noise elimination program unit is used for subtracting the reconstruction signal from the original stator current signal to obtain residual current;
the frequency domain conversion program unit is used for converting the residual current signal in the time domain into a frequency domain to obtain a frequency domain signal;
and the feature extraction and classification diagnosis program unit is used for inputting the frequency domain signal of the residual current into a pre-trained machine learning feature extraction and fault classification model to obtain the motor state.
7. A data-driven electromechanical fault diagnosis system employing current signals, comprising a computer device, characterized in that the computer device is programmed or configured to perform the steps of the data-driven electromechanical fault diagnosis method employing current signals according to any one of claims 1 to 5.
8. A data-driven current-signal-based electromechanical fault diagnosis system comprising a computer device, characterized in that a memory of the computer device has stored therein a computer program programmed or configured to perform the data-driven current-signal-based electromechanical fault diagnosis method according to any one of claims 1 to 5.
9. A computer-readable storage medium, characterized in that the computer-readable storage medium has stored therein a computer program programmed or configured to perform the data-driven electromechanical fault diagnosis method using a current signal according to any one of claims 1 to 5.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010597800.1A CN111680665B (en) | 2020-06-28 | 2020-06-28 | Motor mechanical fault diagnosis method adopting current signals based on data driving |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010597800.1A CN111680665B (en) | 2020-06-28 | 2020-06-28 | Motor mechanical fault diagnosis method adopting current signals based on data driving |
Publications (2)
Publication Number | Publication Date |
---|---|
CN111680665A CN111680665A (en) | 2020-09-18 |
CN111680665B true CN111680665B (en) | 2023-06-20 |
Family
ID=72456808
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010597800.1A Active CN111680665B (en) | 2020-06-28 | 2020-06-28 | Motor mechanical fault diagnosis method adopting current signals based on data driving |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111680665B (en) |
Families Citing this family (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112347944B (en) * | 2020-11-10 | 2023-10-13 | 南京大学 | Mechanical transmission fault detection method based on zero value proportion frequency spectrum characteristics |
CN112504678B (en) * | 2020-11-12 | 2022-12-23 | 重庆科技学院 | Motor bearing fault diagnosis method based on knowledge distillation |
CN112541524B (en) * | 2020-11-18 | 2024-04-02 | 湖南大学 | BP-Adaboost multisource information motor fault diagnosis method based on attention mechanism improvement |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108414923A (en) * | 2018-02-05 | 2018-08-17 | 武汉大学 | A kind of analog-circuit fault diagnosis method based on the extraction of depth confidence network characterization |
Family Cites Families (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH0894499A (en) * | 1994-09-20 | 1996-04-12 | Ishikawajima Harima Heavy Ind Co Ltd | Failure diagnostic system for rotating machine |
CN106295023A (en) * | 2016-08-15 | 2017-01-04 | 南京航空航天大学 | A kind of diagnostic method of asynchronous machine rotor combined failure |
WO2019061006A1 (en) * | 2017-09-26 | 2019-04-04 | Schaeffler Technologies AG & Co. KG | Bearing failure diagnosis method and device, readable storage medium, and electronic device |
CN107957551A (en) * | 2017-12-12 | 2018-04-24 | 南京信息工程大学 | Stacking noise reduction own coding Method of Motor Fault Diagnosis based on vibration and current signal |
CN108761332A (en) * | 2018-05-08 | 2018-11-06 | 郑州轻工业学院 | A kind of set empirical mode decomposition current diagnostic method of motor broken bar fault |
CN108919116B (en) * | 2018-05-16 | 2020-08-28 | 上海海事大学 | Ocean current generator unbalanced stator current fault diagnosis method based on MCCKAF-FFT-Softmax |
CN108731945B (en) * | 2018-08-02 | 2019-12-13 | 南昌航空大学 | Method for extracting fault signal characteristic information of aircraft engine rotor system |
CN109211546B (en) * | 2018-08-28 | 2020-05-26 | 电子科技大学 | Rotary machine fault diagnosis method based on noise reduction automatic encoder and increment learning |
CN109145886A (en) * | 2018-10-12 | 2019-01-04 | 西安交通大学 | A kind of asynchronous machine method for diagnosing faults of Multi-source Information Fusion |
CN109061474A (en) * | 2018-10-15 | 2018-12-21 | 株洲中车时代电气股份有限公司 | A kind of motor bearings trouble-shooter |
CN109774740A (en) * | 2019-02-03 | 2019-05-21 | 湖南工业大学 | A kind of wheel tread damage fault diagnostic method based on deep learning |
CN110110768B (en) * | 2019-04-24 | 2021-02-26 | 西安电子科技大学 | Rolling bearing fault diagnosis method based on parallel feature learning and multiple classifiers |
-
2020
- 2020-06-28 CN CN202010597800.1A patent/CN111680665B/en active Active
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108414923A (en) * | 2018-02-05 | 2018-08-17 | 武汉大学 | A kind of analog-circuit fault diagnosis method based on the extraction of depth confidence network characterization |
Also Published As
Publication number | Publication date |
---|---|
CN111680665A (en) | 2020-09-18 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111680665B (en) | Motor mechanical fault diagnosis method adopting current signals based on data driving | |
Isham et al. | Variational mode decomposition: mode determination method for rotating machinery diagnosis | |
CN106443316B (en) | Multi-information detection method and device for deformation state of power transformer winding | |
Chen et al. | Multi-fault condition monitoring of slurry pump with principle component analysis and sequential hypothesis test | |
US20170356936A1 (en) | Enhanced system and method for conducting pca analysis on data signals | |
CN112052796A (en) | Permanent magnet synchronous motor fault diagnosis method based on deep learning | |
CN106096562B (en) | Gearbox of wind turbine method for diagnosing faults based on vibration signal blind sources separation and sparse component analysis | |
CN108388692B (en) | Rolling bearing fault feature extraction method based on layered sparse coding | |
Zhang et al. | A GOA-MSVM based strategy to achieve high fault identification accuracy for rotating machinery under different load conditions | |
CN113486868B (en) | Motor fault diagnosis method and system | |
CN109655266B (en) | Wind turbine generator bearing fault diagnosis method based on AVMD and spectrum correlation analysis | |
CN112182490B (en) | Reactor state diagnosis method and system | |
CN113391207A (en) | Motor fault detection method, medium and system | |
Song et al. | Data and decision level fusion-based crack detection for compressor blade using acoustic and vibration signal | |
CN114462446A (en) | Rolling bearing fault diagnosis method based on vibration signal and electronic equipment | |
CN114705432A (en) | Method and system for evaluating health state of explosion-proof motor bearing | |
CN113758709A (en) | Rolling bearing fault diagnosis method and system combining edge calculation and deep learning | |
Gao et al. | Machine‐Learning‐Based Intelligent Mechanical Fault Detection and Diagnosis of Wind Turbines | |
CN112729825A (en) | Method for constructing bearing fault diagnosis model based on convolution cyclic neural network | |
CN110222390B (en) | Gear crack identification method based on wavelet neural network | |
CN117093938A (en) | Fan bearing fault detection method and system based on deep learning | |
Li et al. | Rotating machinery anomaly detection using data reconstruction generative adversarial networks with vibration energy analysis | |
CN116524273A (en) | Method, device, equipment and storage medium for detecting draft tube of power station | |
Zheng et al. | Wavelet packet decomposition and neural network based fault diagnosis for elevator excessive vibration | |
CN115326396A (en) | Bearing fault diagnosis method and device |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |