CN113869339A - Deep learning classification model for fault diagnosis and fault diagnosis method - Google Patents
Deep learning classification model for fault diagnosis and fault diagnosis method Download PDFInfo
- Publication number
- CN113869339A CN113869339A CN202110542679.7A CN202110542679A CN113869339A CN 113869339 A CN113869339 A CN 113869339A CN 202110542679 A CN202110542679 A CN 202110542679A CN 113869339 A CN113869339 A CN 113869339A
- Authority
- CN
- China
- Prior art keywords
- fault
- sample data
- data set
- dimensional feature
- information
- 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
Classifications
-
- 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
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- 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
Abstract
The invention provides a fault diagnosis classification model for deep learning, and the training process of the fault diagnosis classification model for deep learning comprises the following steps: data collection, feature extraction and fault diagnosis classification; the data collection is a sample data set for collecting multiple fault types of equipment; the feature extraction comprises the steps of obtaining one-dimensional features, two-dimensional features and three-dimensional features; and acquiring the reason of the equipment failure. And obtaining the fault classification result of the equipment through a deep learning classifier. And (3) the equipment fault diagnosis classification result is associated with the operation parameter information of the equipment, and an algorithm model is trained according to the sampling point number and the sampling frequency which are acquired and set by the vibration acceleration sensor and the data acquired by the data. Meanwhile, the invention also provides a fault diagnosis method. Through multi-dimensional vibration signal feature extraction, the feature value of the vibration signal can be deeply excavated, the relation between the feature and the fault label can be fully extracted before fault diagnosis data are trained, and the accuracy of fault classification and identification is improved.
Description
Technical Field
The invention belongs to the field of equipment detection, and is applied to an equipment detection process. The invention particularly relates to a deep learning classification model for fault diagnosis and a fault diagnosis method.
Background
With the development of computers, from the performance processing of hardware to the gradual maturity of software algorithms, the deep learning algorithm is widely applied to the fields of images and voices, and has remarkable effect and is successfully applied to the practical applications of voice recognition, face recognition, license plate recognition and the like. With the improvement of industrial intelligence, the intelligent monitoring system is particularly important for the state monitoring and fault diagnosis display of equipment. At present, in scientific research articles, bearing faults (rolling element faults, retainer faults, inner ring faults and outer ring faults) are researched most, the average accuracy can reach 92.24%, and the bearing fault detection method can be used in industry. However, only the bearing fault classification research is abundant, but the fault diagnosis classification and fault identification of field industrial equipment are insufficient, and due to different mechanical structures of the equipment, corresponding parts are different, and fault causes generated in the operation process are different. It is necessary to provide intelligent diagnosis decisions for the operation of the equipment, to determine the current operation state of the equipment, or to predict the upcoming degradation trend of the equipment. The intelligent device diagnosis can judge the fault position and reason of the device, so as to guide the field device to maintain.
In the equipment fault diagnosis classification and identification, the method mainly comprises the steps of feature extraction and classifier model construction, generally, a direct fault class data set based on a deep learning algorithm model and a machine learning algorithm model is directly input into a neural network model for feature extraction and classification, the model migration has certain limitation, the model training cannot be well converged, and fitting problems exist in the training process, therefore, the performance of the model can be ensured to be gradually improved only by fully extracting the characteristics of the fault data set, therefore, the method for extracting the always multi-dimensional characteristic fault diagnosis characteristic is extracted, and by knowing the characteristics of the data set, the data set is preprocessed, the data features are processed by machine learning, deep learning and traditional signal processing modes, the extracted multi-dimensional features are trained, and the classification performance of the model can be greatly improved.
Disclosure of Invention
The invention aims to provide a deep learning classification model for fault diagnosis, which is used for detecting operation faults of various shaft equipment by establishing the deep learning classification model, improving the detection precision and reducing the maintenance cost.
On the other hand, the invention provides a fault diagnosis method, which is realized by a deep learning classification model for fault diagnosis, can detect the operation faults of various shafts, improves the detection precision and reduces the maintenance and monitoring cost.
In one aspect of the present invention, a deep learning classification model for fault diagnosis is provided, where the deep learning classification model is obtained by a training method including:
step S101, collecting a sample data set of multiple fault types of equipment and corresponding fault information; the fault information comprises operation parameter information of the equipment; the data in the sample data set are data which continuously change along with time information and have amplitude change;
step S102, acquiring one-dimensional characteristics of the sample data set according to the amplitude information of the sample data set; acquiring two-dimensional characteristics of the sample data set according to the time information and the amplitude information in the sample data set; acquiring three-dimensional characteristics of the sample data set according to the time information and the amplitude information in the sample data set;
step S103, training one-dimensional features, two-dimensional features and three-dimensional features through a deep learning classifier, and obtaining a fault diagnosis classification result; the fault diagnosis classification result is associated with the operation parameter information of the equipment; and obtaining an output result of the deep learning classification model according to the fault diagnosis classification result.
In an embodiment of the deep learning classification model for fault diagnosis according to the present invention, the step of acquiring sample data sets of multiple fault types of a device in step S101 includes:
step S1011, placing an acceleration sensor at the joint of the shaft member and the rigid support member in the fault equipment; the shaft part is connected with the rigid support part through a bearing type connecting piece;
step S1012, operating fault equipment with fault types, collecting an output map of the acceleration sensor, setting the number of sampling points to 262144 and the sampling frequency to 102.4 KHz; corresponding the fault type to the spectrogram information, and recording the power, the rotating speed and the bearing model of the fault equipment;
step S102 further includes:
and normalizing the sample data set to obtain the sample data set with set characteristic dimensions.
In another embodiment of the deep learning classification model for fault diagnosis according to the present invention, the step of obtaining the one-dimensional features of the sample data set according to the amplitude information of the sample data set in step S102 includes:
calculating amplitude information of the sample data set by a wavelet change method to obtain a fault amplitude characteristic; taking the fault amplitude characteristic as a one-dimensional characteristic of the sample data set;
in step S102, the step of obtaining the two-dimensional features of the sample data set according to the time information and the amplitude information in the sample data set includes:
corresponding time information in the sample data set to a time domain signal; converting the time domain signal into a frequency domain signal by fourier transform;
training time domain signals and frequency domain signals through an LSTM long and short term memory network, and extracting time characteristics and fault amplitude characteristics or extracting fault frequency characteristics and speed value characteristics; taking the time characteristic and the fault amplitude characteristic or the fault frequency characteristic and the fault speed value characteristic as two-dimensional characteristics;
step S102, according to time information, amplitude information and energy in the sample data set; the step of obtaining three-dimensional features of the sample data set comprises:
extracting time characteristic, fault frequency characteristic and fault energy characteristic information through time information and amplitude information processed by an MFCC Mel cepstrum coefficient method; and taking the time characteristic, the fault frequency characteristic and the fault energy characteristic information as three-dimensional characteristics.
In another embodiment of the deep learning classification model for fault diagnosis of the present invention, the deep learning classification model for fault diagnosis includes:
the equipment fault data collection module is configured to collect sample data sets of multiple fault types of equipment and fault information corresponding to the sample data sets; the fault information comprises operation parameter information of the equipment; the data in the sample data set are data which continuously change along with time information and have amplitude change;
a one-dimensional feature extraction module configured to obtain one-dimensional features of the sample data set according to the amplitude information of the sample data set;
a two-dimensional feature extraction module configured to obtain two-dimensional features of the sample data set according to the time information and the amplitude information in the sample data set;
a three-dimensional feature extraction module configured to obtain three-dimensional features of the sample data set according to the time information and the amplitude information in the sample data set;
the data input ends of the one-dimensional feature extraction module, the two-dimensional feature extraction module and the three-dimensional feature extraction module simultaneously receive a sample data set of multiple fault types of equipment from the equipment fault data set acquisition module;
the data input end of the classifier is connected with the data output ends of the one-dimensional feature extraction module, the two-dimensional feature extraction module and the three-dimensional feature extraction module, and the classifier is configured to train the one-dimensional feature, the two-dimensional feature and the three-dimensional feature through a deep learning classifier to obtain a fault diagnosis classification result; the fault diagnosis classification result is associated with the operation parameter information of the equipment;
a fault classification result module configured to receive fault diagnosis classification results from a data output of the classifier; and the fault classification result module is configured to obtain an output result of the deep learning classification model according to the fault diagnosis classification result.
In another embodiment of the deep learning classification model for fault diagnosis of the present invention, the deep learning classification model for fault diagnosis includes:
the equipment fault data collection module is configured to collect sample data sets of multiple fault types of equipment and fault information corresponding to the sample data sets; the fault information comprises operation parameter information of the equipment; the data in the sample data set are data which continuously change along with time information and have amplitude change;
a one-dimensional feature extraction module configured to obtain one-dimensional features of the sample data set according to the amplitude information of the sample data set;
a two-dimensional feature extraction module configured to obtain two-dimensional features of the sample data set according to the time information and the amplitude information in the sample data set;
a three-dimensional feature extraction module configured to obtain three-dimensional features of the sample data set according to the time information and the amplitude information in the sample data set;
the system comprises a one-dimensional feature extraction module, a two-dimensional feature extraction module and a three-dimensional feature extraction module, wherein the one-dimensional feature extraction module, the two-dimensional feature extraction module and the three-dimensional feature extraction module are sequentially connected in series; a data input end of the one-dimensional characteristic extraction module receives a sample data set of multiple fault types of equipment from the equipment fault data set acquisition module at the same time; the data output end of the one-dimensional feature extraction module is connected with the data input end of the two-dimensional feature extraction module; the data input end of the two-dimensional feature extraction module is connected with the data output end of the one-dimensional feature extraction module; the data output end of the two-dimensional feature extraction module is connected with the data input end of the three-dimensional feature extraction module; the data input end of the three-dimensional feature extraction module is connected with the data output end of the two-dimensional feature extraction module;
the data input end of the classifier is connected with the data output end of the three-dimensional feature extraction module, and the classifier is configured to train the one-dimensional feature, the two-dimensional feature and the three-dimensional feature through a deep learning classifier to obtain a fault diagnosis classification result; the fault diagnosis classification result is associated with the operation parameter information of the equipment;
a fault classification result module configured to receive fault diagnosis classification results from a data output of the classifier; and the fault classification result module is configured to obtain an output result of the deep learning classification model according to the fault diagnosis classification result.
In a second aspect of the present invention, there is provided a fault diagnosis method including:
step S201, collecting a vibration signal of the current equipment to be detected;
step S202, obtaining a deep learning classification model for fault diagnosis; the deep learning classification model for fault diagnosis is obtained by the following training method, wherein the training method comprises the following steps:
step S101, collecting a sample data set of multiple fault types of equipment and corresponding fault information; the fault information comprises operation parameter information of the equipment; the data in the sample data set are data which continuously change along with time information and have amplitude change;
step S102, acquiring one-dimensional characteristics of the sample data set according to the amplitude information of the sample data set; acquiring two-dimensional characteristics of the sample data set according to the time information and the amplitude information in the sample data set; acquiring three-dimensional characteristics of the sample data set according to the time information and the amplitude information in the sample data set;
step S103, training one-dimensional features, two-dimensional features and three-dimensional features through a deep learning classifier, and obtaining a fault diagnosis classification result; the fault diagnosis classification result is associated with the operation parameter information of the equipment; acquiring an output result of the deep learning classification model according to the fault diagnosis classification result;
step S203, inputting the current operation vibration signal of the equipment to be detected into a deep learning classification model for fault diagnosis to obtain an output result;
and S204, acquiring the fault information of the current equipment to be detected according to the output result.
In an embodiment of the fault diagnosis method of the present invention, the step of acquiring sample data sets of multiple fault types of a device in step S101 includes:
step S1011, placing an acceleration sensor at the joint of the shaft member and the rigid support member in the fault equipment; the shaft part is connected with the rigid support part through a bearing type connecting piece;
step S1012, operating fault equipment with fault types, collecting an output map of the acceleration sensor, setting the number of sampling points to 262144 and the sampling frequency to 102.4 KHz; corresponding the fault type to the spectrogram information, and recording the power, the rotating speed and the bearing model of the fault equipment;
step S102 further includes:
and normalizing the sample data set to obtain the sample data set with set characteristic dimensions.
In another embodiment of the fault diagnosis method of the present invention, the step of obtaining the one-dimensional features of the sample data set according to the amplitude information of the sample data set in step S102 includes:
calculating amplitude information of the sample data set by a wavelet change method to obtain a fault amplitude characteristic; taking the fault amplitude characteristic as a one-dimensional characteristic of the sample data set;
in step S102, the step of obtaining the two-dimensional features of the sample data set according to the time information and the amplitude information in the sample data set includes:
corresponding time information in the sample data set to a time domain signal; converting the time domain signal into a frequency domain signal by fourier transform;
training time domain signals and frequency domain signals through an LSTM long and short term memory network, and extracting time characteristics and fault amplitude characteristics or extracting fault frequency characteristics and speed value characteristics; taking the time characteristic and the fault amplitude characteristic or the fault frequency characteristic and the fault speed value characteristic as two-dimensional characteristics;
in step S102, the step of obtaining the three-dimensional features of the sample data set according to the time information and the amplitude information in the sample data set includes:
extracting time characteristic, fault frequency characteristic and fault energy characteristic information through time information and amplitude information processed by an MFCC Mel cepstrum coefficient method; and taking the time characteristic, the fault frequency characteristic and the fault energy characteristic information as three-dimensional characteristics.
In another embodiment of the fault diagnosis method of the present invention, the deep learning classification model for fault diagnosis includes:
the equipment fault data collection module is configured to collect sample data sets of multiple fault types of equipment and fault information corresponding to the sample data sets; the fault information comprises operation parameter information of the equipment; the data in the sample data set are data which continuously change along with time information and have amplitude change;
a one-dimensional feature extraction module configured to obtain one-dimensional features of the sample data set according to the amplitude information of the sample data set;
a two-dimensional feature extraction module configured to obtain two-dimensional features of the sample data set according to the time information and the amplitude information in the sample data set;
a three-dimensional feature extraction module configured to obtain three-dimensional features of the sample data set according to the time information and the amplitude information in the sample data set;
the data input ends of the one-dimensional feature extraction module, the two-dimensional feature extraction module and the three-dimensional feature extraction module simultaneously receive a sample data set of multiple fault types of equipment from the equipment fault data set acquisition module;
the data input end of the classifier is connected with the data output ends of the one-dimensional feature extraction module, the two-dimensional feature extraction module and the three-dimensional feature extraction module, and the classifier is configured to train the one-dimensional feature, the two-dimensional feature and the three-dimensional feature through a deep learning classifier to obtain a fault diagnosis classification result; the fault diagnosis classification result is associated with the operation parameter information of the equipment;
a fault classification result module configured to receive fault diagnosis classification results from a data output of the classifier; and the fault classification result module is configured to obtain an output result of the deep learning classification model according to the fault diagnosis classification result.
In another embodiment of the fault diagnosis method of the present invention, the deep learning classification model for fault diagnosis includes:
the equipment fault data collection module is configured to collect sample data sets of multiple fault types of equipment and fault information corresponding to the sample data sets; the fault information comprises operation parameter information of the equipment; the data in the sample data set are data which continuously change along with time information and have amplitude change;
a one-dimensional feature extraction module configured to obtain one-dimensional features of the sample data set according to the amplitude information of the sample data set;
a two-dimensional feature extraction module configured to obtain two-dimensional features of the sample data set according to the time information and the amplitude information in the sample data set;
a three-dimensional feature extraction module configured to obtain three-dimensional features of the sample data set according to the time information and the amplitude information in the sample data set;
the system comprises a one-dimensional feature extraction module, a two-dimensional feature extraction module and a three-dimensional feature extraction module, wherein the one-dimensional feature extraction module, the two-dimensional feature extraction module and the three-dimensional feature extraction module are sequentially connected in series; a data input end of the one-dimensional characteristic extraction module receives a sample data set of multiple fault types of equipment from the equipment fault data set acquisition module at the same time; the data output end of the one-dimensional feature extraction module is connected with the data input end of the two-dimensional feature extraction module; the data input end of the two-dimensional feature extraction module is connected with the data output end of the one-dimensional feature extraction module; the data output end of the two-dimensional feature extraction module is connected with the data input end of the three-dimensional feature extraction module; the data input end of the three-dimensional feature extraction module is connected with the data output end of the two-dimensional feature extraction module;
the data input end of the classifier is connected with the data output end of the three-dimensional feature extraction module, and the classifier is configured to train the one-dimensional feature, the two-dimensional feature and the three-dimensional feature through a deep learning classifier to obtain a fault diagnosis classification result; the fault diagnosis classification result is associated with the operation parameter information of the equipment;
a fault classification result module configured to receive fault diagnosis classification results from a data output of the classifier; and the fault classification result module is configured to obtain an output result of the deep learning classification model according to the fault diagnosis classification result.
The characteristics, technical features, advantages and implementation manners of the deep learning classification model for fault diagnosis and the fault diagnosis method are further described in a clear and easy manner with reference to the accompanying drawings.
Drawings
FIG. 1 is a schematic diagram for explaining the steps of a deep learning classification model training method for fault diagnosis according to an embodiment of the present invention.
Fig. 2 is a schematic diagram for explaining a method of collecting a sample data set of multiple fault types of a device in an embodiment of the present invention.
Fig. 3 is a schematic diagram illustrating the composition of a deep learning classification model for fault diagnosis according to an embodiment of the present invention.
Fig. 4 is a schematic diagram illustrating the composition of a deep learning classification model for fault diagnosis according to another embodiment of the present invention.
Fig. 5 is a schematic flowchart for explaining a fault diagnosis method in one embodiment of the present invention.
Detailed Description
In order to more clearly understand the technical features, objects and effects of the present invention, embodiments of the present invention will now be described with reference to the accompanying drawings, in which the same reference numerals indicate the same or structurally similar but functionally identical elements.
"exemplary" means "serving as an example, instance, or illustration" herein, and any illustration, embodiment, or steps described as "exemplary" herein should not be construed as a preferred or advantageous alternative. For the sake of simplicity, the drawings only schematically show the parts relevant to the present exemplary embodiment, and they do not represent the actual structure and the true scale of the product.
The main technical scheme of the invention is as follows: and analyzing the acquired diagnosis data set, wherein in one embodiment, the data set for equipment fault diagnosis is acquired by a vibration sensor, the signal is a time domain signal, and the signal is changed along with the vibration amplitude of a time axis. According to the characteristics of a data set, firstly, directly adopting a wavelet change or other machine learning algorithm model to directly carry out feature acquisition on a one-dimensional waveform; and secondly, carrying out Fourier transformation on the time domain signals to convert the time domain signals into frequency domain signals, and then carrying out feature acquisition on one-dimensional feature frequency values of the frequency domain signals.
Considering that the vibration signal is similar to the vibration signal, the time domain two-dimensional feature information can be processed by introducing a long-time memory network and a short-time memory network through feature extraction on two dimensions of a time dimension and an amplitude dimension, the time dimension feature extraction can be performed on the vibration signal through the two-dimensional information, fault features are fully collected, or the frequency domain signal is subjected to feature extraction on a two-dimensional signal of fault frequency and speed values; except for a two-dimensional feature extraction mode, MFCC feature processing can be adopted for time domain signals to obtain three-dimensional feature information including time, frequency and energy, vibration data features are fully extracted, data and features can be in one-to-one correspondence, deep learning training is carried out subsequently, and classification performance is greatly improved.
The vibration data features can be extracted deeply by fusing one-dimensional, two-dimensional and three-dimensional feature extraction modes, and large-scale data set training can be performed by inputting the vibration data features into a deep learning network model, so that the convergence speed of the model can be improved, and the occurrence of overfitting problems in the training process, the network model parameters and the model training time can be reduced.
In one aspect of the present invention, a deep learning classification model for fault diagnosis is provided, where the deep learning classification model is obtained by a training method, as shown in fig. 1, where the training method includes:
and S101, collecting a sample data set of multiple fault types of equipment.
In this step, a sample data set of multiple fault types of the equipment and corresponding fault information are acquired. The fault information includes operation parameter information of the device. The data in the sample data set is data having amplitude variation that continuously varies with time information.
Step S102, obtaining multi-dimensional characteristics.
In this step, one-dimensional features of the sample data set are obtained according to the amplitude information of the sample data set. And acquiring the two-dimensional characteristics of the sample data set according to the time information and the amplitude information in the sample data set. And acquiring the three-dimensional characteristics of the sample data set according to the time information and the amplitude information in the sample data set.
And step S103, acquiring an output result.
In this step, the one-dimensional, two-dimensional, and three-dimensional features are trained by the deep learning classifier 501, and a fault diagnosis classification result is obtained. And the fault diagnosis classification result is associated with the operation parameter information of the equipment. And obtaining an output result of the deep learning classification model according to the fault diagnosis classification result.
In an embodiment of the deep learning classification model for fault diagnosis according to the present invention, as shown in fig. 2, the step of acquiring sample data sets of multiple fault types of a device in step S101 includes:
in step S1011, an acceleration sensor is provided.
In the step, the acceleration sensor is placed at the joint of the shaft element and the rigid support element in the fault equipment. The shaft member is connected with the rigid support member through a bearing-like connecting member. Acceleration sensor sets sampling point number and sampling frequency
In step S1012, a vibration signal is acquired by the acceleration sensor.
In this step, a fault device with a fault type is operated, an output map of the acceleration sensor is collected, the number of sampling points is set to 262144, and the sampling frequency is set to 102.4 KHz. And (4) corresponding the fault type with the spectrogram information, and recording the power, the rotating speed and the bearing model of the fault equipment.
Step S102 further includes:
and normalizing the sample data set to obtain the sample data set with set characteristic dimensions.
In another embodiment of the deep learning classification model for fault diagnosis according to the present invention, the step of obtaining the one-dimensional features of the sample data set according to the amplitude information of the sample data set in step S102 includes:
and calculating the amplitude information of the sample data set by a wavelet change method to obtain the fault amplitude characteristic. And taking the fault amplitude characteristic as a one-dimensional characteristic of the sample data set.
In step S102, the step of obtaining the two-dimensional features of the sample data set according to the time information and the amplitude information in the sample data set includes:
and corresponding the time information in the sample data set to a time domain signal. The time domain signal is converted into a frequency domain signal by fourier transform.
And training time domain signals and frequency domain signals through an LSTM long and short term memory network, and extracting time characteristics and fault amplitude characteristics or extracting fault frequency characteristics and speed value characteristics. And taking the time characteristic and the fault amplitude characteristic or the fault frequency characteristic and the fault speed value characteristic as two-dimensional characteristics.
In step S102, the step of obtaining the three-dimensional features of the sample data set according to the time information and the amplitude information in the sample data set includes:
and extracting time characteristic, fault frequency characteristic and fault energy characteristic information through the time information and amplitude information processed by the MFCC Mel cepstrum coefficient method. And taking the time characteristic, the fault frequency characteristic and the fault energy characteristic information as three-dimensional characteristics.
In another embodiment of the deep learning classification model for fault diagnosis according to the present invention, as shown in fig. 3, the deep learning classification model for fault diagnosis includes:
an equipment fault data set collection module 101 is configured to collect sample data sets of multiple fault types of equipment and fault information corresponding to the sample data sets. The fault information includes operation parameter information of the device. The data in the sample data set is data having amplitude variation that continuously varies with time information.
A one-dimensional feature extraction module 201 configured to obtain one-dimensional features of the sample data set according to the amplitude information of the sample data set.
A two-dimensional feature extraction module 301 configured to obtain two-dimensional features of the sample data set according to the time information and the amplitude information in the sample data set.
A three-dimensional feature extraction module 401 configured to obtain three-dimensional features of the sample data set according to the time information and the amplitude information in the sample data set.
The data input ends of the one-dimensional feature extraction module 201, the two-dimensional feature extraction module 301 and the three-dimensional feature extraction module 401 receive sample data sets of multiple fault types of equipment from the equipment fault data set module at the same time.
And a data input end of the classifier 501 is connected with data output ends of the one-dimensional feature extraction module 201, the two-dimensional feature extraction module 301 and the three-dimensional feature extraction module 401, and the classifier 501 is configured to train the one-dimensional feature, the two-dimensional feature and the three-dimensional feature through the deep learning classifier 501 to obtain a fault diagnosis classification result. And the fault diagnosis classification result is associated with the operation parameter information of the equipment.
A fault classification result module 601 configured to receive fault diagnosis classification results from the data output of the classifier 501. The fault classification result module 601 is configured to obtain an output result of the deep learning classification model according to the fault diagnosis classification result.
In another embodiment of the deep learning classification model for fault diagnosis according to the present invention, as shown in fig. 4, the deep learning classification model for fault diagnosis includes:
an equipment failure data set collection module 101 is configured to collect sample data sets of multiple types of failures of equipment and failure information corresponding to the sample data sets. The fault information includes operation parameter information of the device. The data in the sample data set is data having amplitude variation that continuously varies with time information.
A one-dimensional feature extraction module 201 configured to obtain one-dimensional features of the sample data set according to the amplitude information of the sample data set.
A two-dimensional feature extraction module 301 configured to obtain two-dimensional features of the sample data set according to the time information and the amplitude information in the sample data set.
A three-dimensional feature extraction module 401 configured to obtain three-dimensional features of the sample data set according to the time information and the amplitude information in the sample data set.
The one-dimensional feature extraction module 201, the two-dimensional feature extraction module 301 and the three-dimensional feature extraction module 401 are connected in series in sequence. The data input of the one-dimensional feature extraction module 201 receives a sample data set of multiple types of equipment faults from the equipment fault data set module at the same time. The data output end of the one-dimensional feature extraction module 201 is connected to the data input end of the two-dimensional feature extraction module 301. The data input end of the two-dimensional feature extraction module 301 is connected to the data output end of the one-dimensional feature extraction module 201. The data output end of the two-dimensional feature extraction module 301 is connected to the data input end of the three-dimensional feature extraction module 401. The data input end of the three-dimensional feature extraction module 401 is connected to the data output end of the two-dimensional feature extraction module 301.
And a data input end of the classifier 501 is connected with a data output end of the three-dimensional feature extraction module 401, and the classifier 501 is configured to train the one-dimensional features, the two-dimensional features and the three-dimensional features through the deep learning classifier 501 to obtain a fault diagnosis classification result. And the fault diagnosis classification result is associated with the operation parameter information of the equipment.
A fault classification result module 601 configured to receive fault diagnosis classification results from the data output of the classifier 501. The fault classification result module 601 is configured to obtain an output result of the deep learning classification model according to the fault diagnosis classification result.
In a second aspect of the present invention, as shown in fig. 5, there is provided a fault diagnosis method including:
step S201, collecting a current operation vibration signal of the equipment to be detected.
In step S202, a deep learning classification model for fault diagnosis is acquired.
In this step, a deep learning classification model for fault diagnosis is obtained. The deep learning classification model for fault diagnosis is obtained by the following training method, wherein the training method comprises the following steps:
step S101, collecting a sample data set of multiple fault types of equipment and corresponding fault information. The fault information includes operation parameter information of the device. The data in the sample data set is data having amplitude variation that continuously varies with time information.
And S102, acquiring one-dimensional characteristics of the sample data set according to the amplitude information of the sample data set. And acquiring the two-dimensional characteristics of the sample data set according to the time information and the amplitude information in the sample data set. And acquiring the three-dimensional characteristics of the sample data set according to the time information and the amplitude information in the sample data set.
Step S103, training one-dimensional features, two-dimensional features and three-dimensional features through the deep learning classifier 501, and obtaining a fault diagnosis classification result. And the fault diagnosis classification result is associated with the operation parameter information of the equipment. And obtaining an output result of the deep learning classification model according to the fault diagnosis classification result.
In step S203, an output result of the deep learning classification model is obtained.
In the step, the current operation vibration signal of the equipment to be detected is input into a deep learning classification model for fault diagnosis to obtain an output result.
And step S204, acquiring the fault information of the current equipment to be detected.
In the step, the current fault information of the equipment to be detected is obtained according to the output result.
In an embodiment of the fault diagnosis method of the present invention, the step of acquiring sample data sets of multiple fault types of a device in step S101 includes:
step S1011, the acceleration sensor is placed at the joint of the shaft member and the rigid support member in the fault equipment. The shaft member is connected with the rigid support member through a bearing-like connecting member.
Step S1012, operating the fault equipment having the fault type, collecting an output map of the acceleration sensor, setting the number of sampling points to 262144 and the sampling frequency to 102.4 KHz. And (4) corresponding the fault type with the spectrogram information, and recording the power, the rotating speed and the bearing model of the fault equipment.
Step S102 further includes:
and normalizing the sample data set to obtain the sample data set with set characteristic dimensions.
In another embodiment of the fault diagnosis method of the present invention, the step of obtaining the one-dimensional features of the sample data set according to the amplitude information of the sample data set in step S102 includes:
and calculating the amplitude information of the sample data set by a wavelet change method to obtain the fault amplitude characteristic. And taking the fault amplitude characteristic as a one-dimensional characteristic of the sample data set.
In step S102, the step of obtaining the two-dimensional features of the sample data set according to the time information and the amplitude information in the sample data set includes:
and corresponding the time information in the sample data set to a time domain signal. The time domain signal is converted into a frequency domain signal by fourier transform.
And training time domain signals and frequency domain signals through an LSTM long and short term memory network, and extracting time characteristics and fault amplitude characteristics or extracting fault frequency characteristics and speed value characteristics. And taking the time characteristic and the fault amplitude characteristic or the fault frequency characteristic and the fault speed value characteristic as two-dimensional characteristics.
In step S102, the step of obtaining the three-dimensional features of the sample data set according to the time information and the amplitude information in the sample data set includes:
and extracting time characteristic, fault frequency characteristic and fault energy characteristic information through the time information and amplitude information processed by the MFCC Mel cepstrum coefficient method. And taking the time characteristic, the fault frequency characteristic and the fault energy characteristic information as three-dimensional characteristics.
In another embodiment of the fault diagnosis method of the present invention, the deep learning classification model for fault diagnosis includes:
an equipment failure data set collection module 101 is configured to collect sample data sets of multiple types of failures of equipment and failure information corresponding to the sample data sets. The fault information includes operation parameter information of the device. The data in the sample data set is data having amplitude variation that continuously varies with time information.
A one-dimensional feature extraction module 201 configured to obtain one-dimensional features of the sample data set according to the amplitude information of the sample data set.
A two-dimensional feature extraction module 301 configured to obtain two-dimensional features of the sample data set according to the time information and the amplitude information in the sample data set.
A three-dimensional feature extraction module 401 configured to obtain three-dimensional features of the sample data set according to the time information and the amplitude information in the sample data set.
The data input ends of the one-dimensional feature extraction module 201, the two-dimensional feature extraction module 301 and the three-dimensional feature extraction module 401 receive sample data sets of multiple fault types of equipment from the equipment fault data set module at the same time.
And a data input end of the classifier 501 is connected with data output ends of the one-dimensional feature extraction module 201, the two-dimensional feature extraction module 301 and the three-dimensional feature extraction module 401, and the classifier 501 is configured to train the one-dimensional feature, the two-dimensional feature and the three-dimensional feature through the deep learning classifier 501 to obtain a fault diagnosis classification result. And the fault diagnosis classification result is associated with the operation parameter information of the equipment.
A fault classification result module 601 configured to receive fault diagnosis classification results from the data output of the classifier 501. The fault classification result module 601 is configured to obtain an output result of the deep learning classification model according to the fault diagnosis classification result.
In another embodiment of the fault diagnosis method of the present invention, the deep learning classification model for fault diagnosis includes:
an equipment failure data set collection module 101 is configured to collect sample data sets of multiple types of failures of equipment and failure information corresponding to the sample data sets. The fault information includes operation parameter information of the device. The data in the sample data set is data having amplitude variation that continuously varies with time information.
A one-dimensional feature extraction module 201 configured to obtain one-dimensional features of the sample data set according to the amplitude information of the sample data set.
A two-dimensional feature extraction module 301 configured to obtain two-dimensional features of the sample data set according to the time information and the amplitude information in the sample data set.
A three-dimensional feature extraction module 401 configured to obtain three-dimensional features of the sample data set according to the time information and the amplitude information in the sample data set.
The one-dimensional feature extraction module 201, the two-dimensional feature extraction module 301 and the three-dimensional feature extraction module 401 are connected in series in sequence. The data input of the one-dimensional feature extraction module 201 receives a sample data set of multiple types of equipment faults from the equipment fault data set module at the same time. The data output end of the one-dimensional feature extraction module 201 is connected to the data input end of the two-dimensional feature extraction module 301. The data input end of the two-dimensional feature extraction module 301 is connected to the data output end of the one-dimensional feature extraction module 201. The data output end of the two-dimensional feature extraction module 301 is connected to the data input end of the three-dimensional feature extraction module 401. The data input end of the three-dimensional feature extraction module 401 is connected to the data output end of the two-dimensional feature extraction module 301.
And a data input end of the classifier 501 is connected with a data output end of the three-dimensional feature extraction module 401, and the classifier 501 is configured to train the one-dimensional features, the two-dimensional features and the three-dimensional features through the deep learning classifier 501 to obtain a fault diagnosis classification result. And the fault diagnosis classification result is associated with the operation parameter information of the equipment.
A fault classification result module 601 configured to receive fault diagnosis classification results from the data output of the classifier 501. The fault classification result module 601 is configured to obtain an output result of the deep learning classification model according to the fault diagnosis classification result.
The invention has the advantages that: firstly, a vibration data set is preprocessed, and through multi-dimensional vibration signal feature extraction, feature extraction can be performed on similar vibration signals for many times, so that interference caused by similar features in a training process is reduced, and the difference of model training is reduced.
Meanwhile, the vibration signal features cannot be fully extracted by a feature extraction mode of a single model, the feature values of the vibration signals can be deeply mined by multi-dimensional feature extraction, the relation between the features and the fault labels can be fully extracted before fault diagnosis data are trained, the input to a classifier model is more suitable for a deep learning algorithm model, and the accuracy of fault classification and identification is gradually improved.
And through multi-dimensional feature extraction, effective analysis can be carried out on the data set, feature extraction is carried out by combining the characteristics of the data set, deep fault feature information can be collected, and multi-dimensional feature extraction modes are fused, so that the multi-dimensional feature information can correspond to a fault data set label.
According to the general flow of the fault diagnosis model under deep learning, the data set preprocessing, the feature extraction, the classification model and the classification result are adopted, the problem that the data set cannot be fully trained or the gradient problem and the fitting problem occur in the training process is solved, and the sampling points and the sampling frequency collected by the vibration data set are different.
Firstly, the data set is normalized to ensure the consistent dimensionality of the data set, secondly, the feature extraction is optimized, the fault diagnosis is carried out according to data collected by field equipment operation, influence factors of the data include field working conditions, equipment aging, hardware products and other related factor influences, so the feature extraction of corresponding fault data is particularly important, for feature extraction including one-dimensional, two-dimensional and three-dimensional feature extraction modes, for the three extraction modes, different weight ranges can be selected for optimization training, or the traditional feature extraction mode is fused for feature extraction.
The main key points comprise weights of connections among multi-dimensional feature extraction and feature dimensions input into the deep learning classifier model; how effectively the connection mode between the multi-dimensional feature extractions avoids the loss of the information of the vibration signal feature quantity.
Fig. 3 and 4 show model structures for optimization training by two different weight and connected feature extraction methods.
In one embodiment of the invention.
1) Firstly, collecting data sets of different equipment fault types, measuring speed values by adopting an acceleration sensor, setting the number of sampling points to be 262144 and setting the sampling frequency to be 102.4 KHz; acquiring fault information of equipment operation, corresponding the fault information with spectrogram information, and recording equipment parameters, such as power, rotating speed, bearing model and other related equipment parameters;
2) firstly, carrying out normalization processing on data acquisition and reducing feature dimension number;
3) performing multi-dimensional feature extraction and one-dimensional feature extraction, and mainly extracting feature information of the waveform; extracting two-dimensional characteristics, namely introducing characteristic information on a time axis, wherein the characteristic information mainly comprises two-dimensional characteristic information such as time, vibration amplitude, frequency and speed values; the three-dimensional characteristic extraction mode is that time, frequency and energy three-dimensional information characteristics are connected through a network model and connected according to the weight and the deviation value, and a vibration fault signal data set is fully extracted.
4) Inputting a unified deep learning classifier, belonging to the characteristic dimension in an input classification model, wherein the number of layers and the parameter quantity of deep learning are unified for training; and the mode of multi-dimensional feature extraction is gradually optimized through the output fault diagnosis classification result, and the generalization capability of the model is improved.
It should be understood that although the present description is described in terms of various embodiments, not every embodiment includes only a single embodiment, and such description is for clarity purposes only, and those skilled in the art will recognize that the embodiments described herein as a whole may be suitably combined to form other embodiments as will be appreciated by those skilled in the art.
The above-listed detailed description is only a specific description of a possible embodiment of the present invention, and they are not intended to limit the scope of the present invention, and equivalent embodiments or modifications made without departing from the technical spirit of the present invention should be included in the scope of the present invention.
Claims (10)
1. A deep learning classification model for fault diagnosis, wherein the deep learning classification model is obtained by a training method comprising:
step S101, collecting a sample data set of multiple fault types of equipment and corresponding fault information; the fault information comprises operation parameter information of the equipment; the data in the sample data set are data which continuously change along with time information and have amplitude change;
step S102, acquiring one-dimensional characteristics of the sample data set according to the amplitude information of the sample data set; acquiring two-dimensional characteristics of the sample data set according to the time information and the amplitude information in the sample data set; acquiring three-dimensional characteristics of the sample data set according to the time information and the amplitude information in the sample data set;
step S103, training the one-dimensional feature, the two-dimensional feature and the three-dimensional feature through a deep learning classifier, and obtaining a fault diagnosis classification result; the fault diagnosis classification result is associated with the operation parameter information of the equipment; and obtaining an output result of the deep learning classification model according to the fault diagnosis classification result.
2. The deep learning classification model for fault diagnosis according to claim 1, wherein the step of collecting sample data sets of multiple fault types of the equipment in step S101 includes:
step S1011, placing an acceleration sensor at the joint of the shaft member and the rigid support member in the fault equipment; the shaft part is connected with the rigid support part through a bearing type connecting piece;
step S1012, operating the fault equipment with fault type, collecting an output map of the acceleration sensor, setting the number of sampling points to 262144 and the sampling frequency to 102.4 KHz; corresponding the fault type to the spectrogram information, and recording the power, the rotating speed and the bearing model of the fault equipment;
the step S102 further includes:
and normalizing the sample data set to obtain the sample data set with set characteristic dimensions.
3. The deep learning classification model for fault diagnosis according to claim 1 or 2, wherein the step of obtaining one-dimensional features of the sample data set according to the amplitude information of the sample data set in step S102 comprises:
calculating amplitude information of the sample data set by a wavelet change method to obtain fault amplitude characteristics; taking the fault amplitude characteristic as a one-dimensional characteristic of a sample data set;
the step of obtaining the two-dimensional features of the sample data set according to the time information and the amplitude information in the sample data set in the step S102 includes:
corresponding the time information in the sample data set to a time domain signal; converting the time domain signal into a frequency domain signal by a fourier transform;
training the time domain signal and the frequency domain signal through an LSTM long-short term memory network, and extracting time characteristics and fault amplitude characteristics or extracting fault frequency characteristics and speed value characteristics; taking the time characteristic and the fault amplitude characteristic or the fault frequency characteristic and the fault speed value characteristic as two-dimensional characteristics;
the step of obtaining the three-dimensional features of the sample data set according to the time information and the amplitude information in the sample data set in the step S102 includes:
processing the time information and the amplitude information by an MFCC Mel cepstrum coefficient method, and extracting time characteristic, fault frequency characteristic and fault energy characteristic information; and taking the time characteristic, the fault frequency characteristic and the fault energy characteristic information as three-dimensional characteristics.
4. The deep-learning classification model for fault diagnosis according to claim 1, characterized in that the deep-learning classification model for fault diagnosis comprises:
the equipment fault data collection module is configured to collect sample data sets of multiple fault types of equipment and fault information corresponding to the sample data sets; the fault information comprises operation parameter information of the equipment; the data in the sample data set are data which continuously change along with time information and have amplitude change;
a one-dimensional feature extraction module configured to obtain one-dimensional features of the sample data set according to the amplitude information of the sample data set;
a two-dimensional feature extraction module configured to obtain two-dimensional features of the sample data set according to the time information and the amplitude information in the sample data set;
a three-dimensional feature extraction module configured to extract a three-dimensional feature from the time information, the magnitude information, and the energy information in the sample data set; acquiring three-dimensional characteristics of the sample data set;
the data input ends of the one-dimensional feature extraction module, the two-dimensional feature extraction module and the three-dimensional feature extraction module simultaneously receive sample data sets of multiple fault types of the equipment from the equipment fault data set acquisition module;
the data input end of the classifier is connected with the data output ends of the one-dimensional feature extraction module, the two-dimensional feature extraction module and the three-dimensional feature extraction module, and the classifier is configured to train the one-dimensional feature, the two-dimensional feature and the three-dimensional feature through a deep learning classifier to obtain a fault diagnosis classification result; the fault diagnosis classification result is associated with the operation parameter information of the equipment;
a fault classification result module configured to receive said fault diagnosis classification result from a data output of said classifier; and the fault classification result module is configured to obtain an output result of the deep learning classification model according to the fault diagnosis classification result.
5. The deep-learning classification model for fault diagnosis according to claim 1, characterized in that the deep-learning classification model for fault diagnosis comprises:
the equipment fault data collection module is configured to collect sample data sets of multiple fault types of equipment and fault information corresponding to the sample data sets; the fault information comprises operation parameter information of the equipment; the data in the sample data set are data which continuously change along with time information and have amplitude change;
a one-dimensional feature extraction module configured to obtain one-dimensional features of the sample data set according to the amplitude information of the sample data set;
a two-dimensional feature extraction module configured to obtain two-dimensional features of the sample data set according to the time information and the amplitude information in the sample data set;
a three-dimensional feature extraction module configured to obtain three-dimensional features of the sample data set according to the time information and the amplitude information in the sample data set;
the one-dimensional feature extraction module, the two-dimensional feature extraction module and the three-dimensional feature extraction module are sequentially connected in series; a data input end of the one-dimensional feature extraction module receives a sample data set of multiple fault types of the equipment from the equipment fault data set acquisition module at the same time; the data output end of the one-dimensional feature extraction module is connected with the data input end of the two-dimensional feature extraction module; the data input end of the two-dimensional feature extraction module is connected with the data output end of the one-dimensional feature extraction module; the data output end of the two-dimensional feature extraction module is connected with the data input end of the three-dimensional feature extraction module; the data input end of the three-dimensional feature extraction module is connected with the data output end of the two-dimensional feature extraction module;
the data input end of the classifier is connected with the data output end of the three-dimensional feature extraction module, and the classifier is configured to train the one-dimensional feature, the two-dimensional feature and the three-dimensional feature through a deep learning classifier to obtain a fault diagnosis classification result; the fault diagnosis classification result is associated with the operation parameter information of the equipment;
a fault classification result module configured to receive said fault diagnosis classification result from a data output of said classifier; and the fault classification result module is configured to obtain an output result of the deep learning classification model according to the fault diagnosis classification result.
6. A fault diagnosis method characterized by comprising:
step S201, collecting a current operation vibration signal of equipment to be detected;
step S202, obtaining a deep learning classification model for fault diagnosis; the deep learning classification model for fault diagnosis is obtained by the following training method, wherein the training method comprises the following steps:
step S101, collecting a sample data set of multiple fault types of equipment and corresponding fault information; the fault information comprises operation parameter information of the equipment; the data in the sample data set are data which continuously change along with time information and have amplitude change;
step S102, acquiring one-dimensional characteristics of the sample data set according to the amplitude information of the sample data set; acquiring two-dimensional characteristics of the sample data set according to the time information and the amplitude information in the sample data set; acquiring three-dimensional characteristics of the sample data set according to the time information and the amplitude information in the sample data set;
step S103, training the one-dimensional feature, the two-dimensional feature and the three-dimensional feature through a deep learning classifier, and obtaining a fault diagnosis classification result; the fault diagnosis classification result is associated with the operation parameter information of the equipment; obtaining an output result of the deep learning classification model according to the fault diagnosis classification result;
step S203, inputting the current operation vibration signal of the equipment to be detected into the deep learning classification model for fault diagnosis to obtain an output result;
and step S204, acquiring the fault information of the current equipment to be detected according to the output result.
7. The method according to claim 6, wherein the step of collecting sample data sets of multiple fault types of the device in step S101 includes:
step S1011, placing an acceleration sensor at the joint of the shaft member and the rigid support member in the fault equipment; the shaft part is connected with the rigid support part through a bearing type connecting piece;
step S1012, operating the fault equipment with fault type, collecting an output map of the acceleration sensor, setting the number of sampling points to 262144 and the sampling frequency to 102.4 KHz; corresponding the fault type to the spectrogram information, and recording the power, the rotating speed and the bearing model of the fault equipment;
the step S102 further includes:
and normalizing the sample data set to obtain the sample data set with set characteristic dimensions.
8. The method according to claim 6 or 7, wherein the step of obtaining the one-dimensional features of the sample data set according to the amplitude information of the sample data set in the step S102 includes:
calculating amplitude information of the sample data set by a wavelet change method to obtain fault amplitude characteristics; taking the fault amplitude characteristic as a one-dimensional characteristic of a sample data set;
the step of obtaining the two-dimensional features of the sample data set according to the time information and the amplitude information in the sample data set in the step S102 includes:
corresponding the time information in the sample data set to a time domain signal; converting the time domain signal into a frequency domain signal by a fourier transform;
training the time domain signal and the frequency domain signal through an LSTM long-short term memory network, and extracting time characteristics and fault amplitude characteristics or extracting fault frequency characteristics and speed value characteristics; taking the time characteristic and the fault amplitude characteristic or the fault frequency characteristic and the fault speed value characteristic as two-dimensional characteristics;
the step of obtaining the three-dimensional features of the sample data set according to the time information and the amplitude information in the sample data set in the step S102 includes:
processing the time information and the amplitude information by an MFCC Mel cepstrum coefficient method, and extracting time characteristic, fault frequency characteristic and fault energy characteristic information; and taking the time characteristic, the fault frequency characteristic and the fault energy characteristic information as three-dimensional characteristics.
9. The fault diagnosis method according to claim 6, wherein the deep learning classification model for fault diagnosis includes:
the equipment fault data collection module is configured to collect sample data sets of multiple fault types of equipment and fault information corresponding to the sample data sets; the fault information comprises operation parameter information of the equipment; the data in the sample data set are data which continuously change along with time information and have amplitude change;
a one-dimensional feature extraction module configured to obtain one-dimensional features of the sample data set according to the amplitude information of the sample data set;
a two-dimensional feature extraction module configured to obtain two-dimensional features of the sample data set according to the time information and the amplitude information in the sample data set;
a three-dimensional feature extraction module configured to obtain three-dimensional features of the sample data set according to the time information and the amplitude information in the sample data set;
the data input ends of the one-dimensional feature extraction module, the two-dimensional feature extraction module and the three-dimensional feature extraction module simultaneously receive sample data sets of multiple fault types of the equipment from the equipment fault data set acquisition module;
the data input end of the classifier is connected with the data output ends of the one-dimensional feature extraction module, the two-dimensional feature extraction module and the three-dimensional feature extraction module, and the classifier is configured to train the one-dimensional feature, the two-dimensional feature and the three-dimensional feature through a deep learning classifier to obtain a fault diagnosis classification result; the fault diagnosis classification result is associated with the operation parameter information of the equipment;
a fault classification result module configured to receive said fault diagnosis classification result from a data output of said classifier; and the fault classification result module is configured to obtain an output result of the deep learning classification model according to the fault diagnosis classification result.
10. The fault diagnosis method according to claim 6, wherein the deep learning classification model for fault diagnosis includes:
the equipment fault data collection module is configured to collect sample data sets of multiple fault types of equipment and fault information corresponding to the sample data sets; the fault information comprises operation parameter information of the equipment; the data in the sample data set are data which continuously change along with time information and have amplitude change;
a one-dimensional feature extraction module configured to obtain one-dimensional features of the sample data set according to the amplitude information of the sample data set;
a two-dimensional feature extraction module configured to obtain two-dimensional features of the sample data set according to the time information and the amplitude information in the sample data set;
a three-dimensional feature extraction module configured to obtain three-dimensional features of the sample data set according to the time information and the amplitude information in the sample data set;
the one-dimensional feature extraction module, the two-dimensional feature extraction module and the three-dimensional feature extraction module are sequentially connected in series; a data input end of the one-dimensional feature extraction module receives a sample data set of multiple fault types of the equipment from the equipment fault data set acquisition module at the same time; the data output end of the one-dimensional feature extraction module is connected with the data input end of the two-dimensional feature extraction module; the data input end of the two-dimensional feature extraction module is connected with the data output end of the one-dimensional feature extraction module; the data output end of the two-dimensional feature extraction module is connected with the data input end of the three-dimensional feature extraction module; the data input end of the three-dimensional feature extraction module is connected with the data output end of the two-dimensional feature extraction module;
the data input end of the classifier is connected with the data output end of the three-dimensional feature extraction module, and the classifier is configured to train the one-dimensional feature, the two-dimensional feature and the three-dimensional feature through a deep learning classifier to obtain a fault diagnosis classification result; the fault diagnosis classification result is associated with the operation parameter information of the equipment;
a fault classification result module configured to receive said fault diagnosis classification result from a data output of said classifier; and the fault classification result module is configured to obtain an output result of the deep learning classification model according to the fault diagnosis classification result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110542679.7A CN113869339A (en) | 2021-05-18 | 2021-05-18 | Deep learning classification model for fault diagnosis and fault diagnosis method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110542679.7A CN113869339A (en) | 2021-05-18 | 2021-05-18 | Deep learning classification model for fault diagnosis and fault diagnosis method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113869339A true CN113869339A (en) | 2021-12-31 |
Family
ID=78989876
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110542679.7A Pending CN113869339A (en) | 2021-05-18 | 2021-05-18 | Deep learning classification model for fault diagnosis and fault diagnosis method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113869339A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114941890A (en) * | 2022-05-24 | 2022-08-26 | 日照安泰科技发展有限公司 | Central air conditioner fault diagnosis method and system based on image and depth blurring |
| CN115865630A (en) * | 2023-02-28 | 2023-03-28 | 广东名阳信息科技有限公司 | Network equipment fault diagnosis method and system based on deep learning |
-
2021
- 2021-05-18 CN CN202110542679.7A patent/CN113869339A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114941890A (en) * | 2022-05-24 | 2022-08-26 | 日照安泰科技发展有限公司 | Central air conditioner fault diagnosis method and system based on image and depth blurring |
| CN114941890B (en) * | 2022-05-24 | 2024-04-30 | 日照安泰科技发展有限公司 | Central air conditioner fault diagnosis method and system based on image and depth blur |
| CN115865630A (en) * | 2023-02-28 | 2023-03-28 | 广东名阳信息科技有限公司 | Network equipment fault diagnosis method and system based on deep learning |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107228766B (en) | Based on the Fault Diagnosis of Roller Bearings for improving multiple dimensioned fuzzy entropy | |
| CN111914883B (en) | Spindle bearing state evaluation method and device based on deep fusion network | |
| CN111947928B (en) | Multi-source information fusion bearing fault prediction system and method | |
| CN111191740B (en) | Fault diagnosis method for rolling bearing | |
| CN113869339A (en) | Deep learning classification model for fault diagnosis and fault diagnosis method | |
| Pan et al. | A deep learning network via shunt-wound restricted Boltzmann machines using raw data for fault detection | |
| CN106441896A (en) | Characteristic vector extraction method for rolling bearing fault mode identification and state monitoring | |
| CN111898644B (en) | Intelligent identification method for health state of aerospace liquid engine under fault-free sample | |
| CN113008559B (en) | Bearing fault diagnosis method and system based on sparse self-encoder and Softmax | |
| CN109375010A (en) | Winding in Power Transformer method for diagnosing faults based on vibration distribution characteristic model | |
| CN112329329B (en) | Simulation data driven rotary machine depth semi-supervised migration diagnosis method | |
| CN108760300A (en) | A method of intelligent fault diagnosis being carried out to it according to bearing vibration signal | |
| CN112665852A (en) | Variable working condition planetary gearbox fault diagnosis method and device based on deep learning | |
| CN112305388B (en) | On-line monitoring and diagnosing method for insulation partial discharge faults of generator stator winding | |
| CN116593157A (en) | Complex working condition gear fault diagnosis method based on matching element learning under small sample | |
| CN114429152A (en) | Rolling bearing fault diagnosis method based on dynamic index antagonism self-adaption | |
| Qi et al. | Feature classification method of frequency cepstrum coefficient based on weighted extreme gradient boosting | |
| Li et al. | Transformer-based meta learning method for bearing fault identification under multiple small sample conditions | |
| CN112541524A (en) | BP-Adaboost multi-source information motor fault diagnosis method based on attention mechanism improvement | |
| CN112861275A (en) | Rotary machine fault diagnosis method based on minimum information entropy feature learning model | |
| CN102778632A (en) | Double normalization recognition method for directly forecasting and recognizing transformer winding fault type | |
| CN115795341A (en) | Two-dimensional piston pump health state assessment method based on variable rotating speed | |
| CN113790890A (en) | Wavelet packet decomposition weight fuzzy entropy and ELM-based bearing fault classification method and device | |
| Wang et al. | Adversarial Based Unsupervised Domain Adaptation for Bearing Fault Diagnosis | |
| CN112528849B (en) | Structure health monitoring method based on inner product matrix and deep learning |
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 |