CN113569486A - Fault detection method of industrial equipment, computing equipment and readable storage medium - Google Patents
Fault detection method of industrial equipment, computing equipment and readable storage medium Download PDFInfo
- Publication number
- CN113569486A CN113569486A CN202110876571.1A CN202110876571A CN113569486A CN 113569486 A CN113569486 A CN 113569486A CN 202110876571 A CN202110876571 A CN 202110876571A CN 113569486 A CN113569486 A CN 113569486A
- Authority
- CN
- China
- Prior art keywords
- fault
- fault detection
- data
- sample data
- frequency spectrum
- 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
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- 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
- G06F18/23—Clustering techniques
-
- 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
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2111/00—Details relating to CAD techniques
- G06F2111/08—Probabilistic or stochastic CAD
Abstract
The invention discloses a fault detection method of industrial equipment, which is executed in computing equipment and comprises the following steps: acquiring a frequency spectrum of vibration data within a preset time length when a device to be detected operates; normalizing the frequency spectrum of the vibration data to obtain a normalized frequency spectrum of the vibration data; and inputting the normalized frequency spectrum of the vibration data into a trained fault detection model for processing to obtain the fault category of the equipment to be detected. The fault detection model is constructed based on a metric element learning model. Therefore, the fault detection method can obtain good fault detection effect under the condition of small samples, so that the fault detection accuracy under the small samples can be improved.
Description
Technical Field
The present invention relates to the field of computers, and in particular, to a fault detection method for an industrial device, a computing device, and a readable storage medium.
Background
In recent years, deep learning has made significant progress in the field of fault detection with its unique advantages in feature extraction and pattern recognition. At present, deep learning models such as a deep confidence network, a self-coding network, a convolutional neural network and a recurrent neural network are widely applied to the field of fault detection.
However, when applying these deep learning models to the field of fault detection, it is necessary to train them with a large amount of sample data and sufficient computational resources. Otherwise, the effect of fault classification will be poor. That is, fault detection based on deep learning requires relying on a large amount of sample data to ensure the accuracy of its detection. However, it is very difficult for industrial equipment to obtain a large number of fault instance samples, or it can even be said that a large number of fault instance samples cannot be obtained at all.
Therefore, a new fault detection method is needed to solve the above problems.
Disclosure of Invention
To this end, the present invention provides a fault detection method, a computing device and a readable storage medium for an industrial device in an attempt to solve or at least alleviate the above-presented problems.
According to an aspect of the present invention, there is provided a fault detection method of an industrial device, executed in a computing device, the method comprising: acquiring a frequency spectrum of vibration data within a preset time length when a device to be detected operates; normalizing the frequency spectrum of the vibration data to obtain a normalized frequency spectrum of the vibration data; and inputting the normalized frequency spectrum of the vibration data into a trained fault detection model for processing to obtain the fault category of the equipment to be detected, and constructing the fault detection model based on a metric learning model.
Optionally, in the method for detecting a fault of an industrial device according to the present invention, the step of obtaining a frequency spectrum of vibration data within a preset time period when the device to be detected operates includes: obtaining vibration data within a preset time length when a device to be detected operates; and acquiring the frequency spectrum of the vibration data by using short-time Fourier transform.
Alternatively, in the fault detection method of an industrial device according to the present invention, the vibration data is acceleration data.
Optionally, in the fault detection method of the industrial equipment according to the present invention, the metric learning model is a prototype network.
Optionally, in the fault detection method of the industrial equipment according to the present invention, the fault detection model includes a feature extraction module and a metric module.
Optionally, in the fault detection method of the industrial device according to the present invention, the feature extraction module includes four layers of convolutional networks, each layer of convolutional network including one convolutional layer, one active layer, and one pooling layer.
Alternatively, in the fault detection method of the industrial equipment according to the present invention, the fault detection model is trained based on the following method: randomly extracting a first preset number of fault types from a total training sample set, wherein each fault type in the total training sample set comprises a plurality of sample data, and the sample data is a normalized frequency spectrum of vibration data within a preset time length when the industrial equipment operates; for each extracted fault category, randomly extracting a second preset number of sample data from the total training sample set; taking a third preset number of sample data in the second preset number of sample data as a meta-training support set, and taking the residual sample data as a meta-training query set; training the fault detection model by using the support set and the query set to obtain an initial fault detection model; and repeating the steps until the preset times are reached, and obtaining the trained fault detection model.
Optionally, in the method for detecting a fault of an industrial device according to the present invention, the step of training the fault detection model using the support set and the query set includes: inputting sample data supporting each fault category in the set into a feature extraction module to obtain a feature vector of each sample data; acquiring prototype representation of each fault category in the support set based on the obtained feature vector of each sample data; inputting sample data under each fault category in the query set into a feature extraction module to obtain a feature vector of each sample data in the query set; acquiring a predicted fault category corresponding to each sample data in the query set based on the feature vector of each sample data in the query set and the acquired prototype representation of each fault category; and updating the fault detection model based on the loss value between the predicted fault category and the real fault category corresponding to each sample data in the query set to obtain an initial fault detection model.
According to yet another aspect of the invention, there is provided a computing device comprising: at least one processor; and a memory storing program instructions, wherein the program instructions are configured to be executed by the at least one processor, the program instructions comprising instructions for performing the fault detection method of the industrial device according to the present invention.
According to still another aspect of the present invention, there is provided a readable storage medium storing program instructions which, when read and executed by a computing device, cause the computing device to execute a fault detection method of an industrial device according to the present invention.
According to the fault detection method of the industrial equipment, firstly, the frequency spectrum of the vibration data within the preset time length when the equipment to be detected operates is obtained. Then, the frequency spectrum of the vibration data is normalized to obtain the normalized frequency spectrum of the vibration data. And finally, acquiring the fault category of the equipment to be detected by using a fault detection model constructed based on the metric element learning model according to the normalized frequency spectrum of the vibration data. Therefore, the fault detection method provided by the invention adopts the fault detection model constructed based on the metric learning model, so that a good fault detection effect can be obtained under the condition of a small sample, and the fault detection accuracy under the condition of the small sample is improved.
Moreover, the invention carries out normalization processing on the frequency spectrum of the vibration data of the equipment to be detected, thus eliminating the influence on fault detection caused by different amplitude ranges of the frequency spectrum, and further improving the accuracy of fault detection.
Drawings
To the accomplishment of the foregoing and related ends, certain illustrative aspects are described herein in connection with the following description and the annexed drawings, which are indicative of various ways in which the principles disclosed herein may be practiced, and all aspects and equivalents thereof are intended to be within the scope of the claimed subject matter. The above and other objects, features and advantages of the present disclosure will become more apparent from the following detailed description read in conjunction with the accompanying drawings. Throughout this disclosure, like reference numerals generally refer to like parts or elements.
FIG. 1 shows a block diagram of a computing device 100, according to one embodiment of the invention;
FIG. 2 illustrates a flow diagram of a method 200 of fault detection of an industrial device according to one embodiment of the present invention;
FIG. 3 shows a schematic diagram of a structure diagram of a fault detection model according to one embodiment of the invention.
Detailed Description
Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be embodied in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.
Fault diagnosis (and fault detection) plays a critical role in the safe operation of equipment and the orderly progress of industrial production. The failure diagnosis methods can be roughly classified into two types, a mechanism-based failure diagnosis method and a data-driven failure diagnosis method. The mechanism-based fault diagnosis method can clearly depict the mechanism of each fault and the relation between input and output, and is simple and accurate. However, it has certain limitations-it is not suitable for complex fault diagnosis because it cannot acquire all information of internal mechanisms for a too complex device.
The data-driven fault diagnosis method can be used for complex fault diagnosis, such as fault diagnosis of rotating equipment with complex internal structure and diverse external operating environments. The data-driven fault diagnosis method is to directly use the state monitoring data to infer mechanical faults without making any assumptions about potential fault mechanisms.
The fault diagnosis method based on data driving goes through two stages, namely an intelligent fault diagnosis stage based on artificial features and an intelligent fault diagnosis stage based on deep learning. The fault diagnosis based on artificial features is that features are extracted from original data manually, and then sensitive features are input into machine learning models such as a Support Vector Machine (SVM), k-nearest neighbor (KNN), Artificial Neural Network (ANN) and the like. The method requires that workers have abundant professional knowledge and diagnosis experience, and redesigns a feature extraction algorithm according to the change of a diagnosis task, which obviously brings certain difficulty to intelligent fault diagnosis based on artificial features. Based on the method, the data-driven fault diagnosis method enters a second stage, namely an intelligent fault diagnosis stage based on deep learning.
The intelligent fault diagnosis based on deep learning needs to train a model by means of a large amount of data and sufficient computing resources so as to ensure the accuracy of fault diagnosis. However, the large number of fault instances and sufficient computing resources required for deep learning are not sufficient for industrial equipment.
Reasons for not obtaining enough samples to make the classifier robust to each fault type mainly include: (1) equipment failure, particularly of critical systems, can have serious consequences and therefore equipment is not normally allowed to fail. (2) Most faults follow a degradation path with slow degradation, so the fault degradation of the device can be months or even years long, which makes it very difficult to collect relevant data. (3) The mechanical systems are very complex in their operating conditions and often vary according to production requirements, which makes it impractical to collect and label sufficient training samples. In particular, in practical applications, the fault classes and operating conditions are often unbalanced, so it is difficult to collect enough samples for each fault type under different operating conditions. (4) Collecting and labeling sufficient training samples costs a significant amount of manpower and material. The reasons why sufficient computing resources cannot be satisfied mainly include: (1) the sufficient computational resources and training time required for deep learning are also contrary to the practical requirements in the field of device failure diagnosis. Excessive computing resources represent a significant cost expenditure that makes enterprises unwilling to afford. (2) The equipment fault diagnosis needs rapid response capability and diagnosis capability, and the actual requirements of the equipment fault diagnosis cannot be met due to the overlong training time.
Therefore, when the fault prediction method based on deep learning is applied to fault detection of industrial equipment, the fault prediction method based on deep learning is restricted by operation resources and the number of samples. However, the number of samples is too small, and the deep learning model may be at risk of overfitting. Overfitting refers to a problem in the model parameter fitting process. Specifically, the training data contains sampling errors, and the complex model takes the sampling errors into account during training (i.e., the sampling errors are well fitted), so that the model has a good effect on the training set, but has a poor effect on the test set, i.e., the generalization capability of the model is weak. That is, in the case of a small sample, the fault classification effect of the deep learning model is poor.
Based on the above, the invention provides a fault detection method based on metric learning, which is characterized in that the fault identification (namely fault detection and fault diagnosis) learned in the training process is rapidly applied to the fault identification under the new working condition through the metric learning such as a prototype network, so that the dependence of the fault diagnosis based on deep learning on a large amount of sample data can be overcome, and the fault detection accuracy can be ensured under different complex working conditions. In addition, compared with a deep learning network, the metric learning network has a simpler structure, so that the required computing resources and training time can be effectively reduced.
FIG. 1 shows a block diagram of a computing device 100, according to one embodiment of the invention. It should be noted that the computing device 100 shown in fig. 1 is only an example, and in practice, the computing device for implementing the fault detection method of the industrial device of the present invention may be any type of device, and the hardware configuration thereof may be the same as the computing device 100 shown in fig. 1 or different from the computing device 100 shown in fig. 1. In practice, the computing device for implementing the fault detection method of the industrial device of the present invention may add or delete hardware components of the computing device 100 shown in fig. 1, and the present invention does not limit the specific hardware configuration of the computing device.
As shown in FIG. 1, in a basic configuration 102, a computing device 100 typically includes a system memory 106 and one or more processors 104. A memory bus 108 may be used for communication between the processor 104 and the system memory 106.
Depending on the desired configuration, the processor 104 may be any type of processing, including but not limited to: a microprocessor (μ P), a microcontroller (μ C), a Digital Signal Processor (DSP), or any combination thereof. The processor 104 may include one or more levels of cache, such as a level one cache 110 and a level two cache 112, a processor core 114, and registers 116. The example processor core 114 may include an Arithmetic Logic Unit (ALU), a Floating Point Unit (FPU), a digital signal processing core (DSP core), or any combination thereof. The example memory controller 118 may be used with the processor 104, or in some implementations the memory controller 118 may be an internal part of the processor 104.
Depending on the desired configuration, system memory 106 may be any type of memory, including but not limited to: volatile memory (such as RAM), non-volatile memory (such as ROM, flash memory, etc.), or any combination thereof. The physical memory in the computing device is usually referred to as a volatile memory RAM, and data in the disk needs to be loaded into the physical memory to be read by the processor 104. System memory 106 may include an operating system 120, one or more applications 122, and program data 124. In some implementations, the application 122 can be arranged to execute instructions on an operating system with program data 124 by one or more processors 104. Operating system 120 may be, for example, Linux, Windows, etc., which includes program instructions for handling basic system services and performing hardware dependent tasks. The application 122 includes program instructions for implementing various user-desired functions, and the application 122 may be, for example, but not limited to, a browser, instant messenger, a software development tool (e.g., an integrated development environment IDE, a compiler, etc.), and the like. When the application 122 is installed into the computing device 100, a driver module may be added to the operating system 120.
When the computing device 100 is started, the processor 104 reads program instructions of the operating system 120 from the system memory 106 and executes them. The application 122 runs on top of the operating system 120, utilizing the operating system 120 and interfaces provided by the underlying hardware to implement various user-desired functions. When the user launches the application 122, the application 122 is loaded into the system memory 106, and the processor 104 reads and executes the program instructions of the application 122 from the system memory 106.
The computing device 100 also includes a storage device 132, the storage device 132 including removable storage 136 and non-removable storage 138, the removable storage 136 and the non-removable storage 138 each connected to the storage interface bus 134.
Computing device 100 may also include an interface bus 140 that facilitates communication from various interface devices (e.g., output devices 142, peripheral interfaces 144, and communication devices 146) to the basic configuration 102 via the bus/interface controller 130. The example output device 142 includes a graphics processing unit 148 and an audio processing unit 150. They may be configured to facilitate communication with various external devices, such as a display or speakers, via one or more a/V ports 152. Example peripheral interfaces 144 may include a serial interface controller 154 and a parallel interface controller 156, which may be configured to facilitate communication with external devices such as input devices (e.g., keyboard, mouse, pen, voice input device, touch input device) or other peripherals (e.g., printer, scanner, etc.) via one or more I/O ports 158. An example communication device 146 may include a network controller 160, which may be arranged to facilitate communications with one or more other computing devices 162 over a network communication link via one or more communication ports 164.
A network communication link may be one example of a communication medium. Communication media may typically be embodied by computer readable instructions, data structures, program modules, and may include any information delivery media, such as carrier waves or other transport mechanisms, in a modulated data signal. A "modulated data signal" may be a signal that has one or more of its data set or its changes made in such a manner as to encode information in the signal. By way of non-limiting example, communication media may include wired media such as a wired network or private-wired network, and various wireless media such as acoustic, Radio Frequency (RF), microwave, Infrared (IR), or other wireless media. The term computer readable media as used herein may include both storage media and communication media.
In the computing device 100 according to the present invention, the application 122 includes instructions for performing the fault detection method 200 of the industrial device of the present invention, which may instruct the processor 104 to perform the fault detection method of the industrial device of the present invention. It will be appreciated by those skilled in the art that the application 122 may include other applications 126 for implementing other functions in addition to instructions for performing the fault detection method 200 of an industrial device.
FIG. 2 illustrates a flow diagram of a method 200 for fault detection of industrial equipment, the method 200 being suitable for execution in a computing device (e.g., the computing device 100 shown in FIG. 1), according to one embodiment of the invention. As shown in fig. 2, the method begins at step S210.
In step S210, a frequency spectrum of vibration data within a preset time duration when the device to be detected operates is obtained. Specifically, the method comprises the following steps: firstly, obtaining vibration data within a preset time length when the equipment to be detected operates. Then, a frequency spectrum of the vibration data is acquired by using short-time fourier transform. In short, the obtained vibration signal data is converted into frequency spectrum data by short-time fourier transform, thereby obtaining a frequency spectrum of the vibration data. The device to be detected can be a rotating device, and the vibration data can be acceleration data.
When the vibration data is acceleration data, it can be acquired by an acceleration sensor. According to one embodiment of the invention, the step of obtaining the frequency spectrum of the acceleration data within the preset time length when the device to be detected operates is specifically as follows: the method comprises the steps of firstly, obtaining acceleration data within a preset time length when the equipment to be detected runs through an acceleration sensor, then converting the acceleration data into frequency spectrum information by utilizing short-time Fourier transform, and obtaining a frequency spectrum of the acceleration data within the preset time length when the equipment to be detected runs. The preset time period may be two minutes, which is not limited in the present invention. In the specific embodiment, a person skilled in the art can set the setting according to actual needs.
In addition, considering that the amplitude ranges of the frequency spectrums of the vibration data obtained under different working conditions (for example, different loads) are different, and the difference in the amplitude ranges affects the accuracy of fault diagnosis, in order to improve the accuracy of fault detection, normalization processing may be performed on the frequency spectrums of the obtained vibration data.
Then, step S220 is performed to normalize the frequency spectrum of the vibration data, so as to obtain a normalized frequency spectrum of the vibration data. According to an embodiment of the present invention, the acquired frequency spectrum of the vibration data within the preset time period may be normalized by the following formula:
wherein x' (t) is the normalized amplitude of the t-th sampling point in the frequency spectrum of the vibration data in the preset time length, x (t) is the amplitude of the t-th sampling point in the frequency spectrum of the vibration data in the preset time length, mean (x) is the average amplitude of the frequency spectrum of the vibration data in the preset time length, max (x) is the maximum amplitude in the frequency spectrum of the vibration data in the preset time length, and min (x) is the minimum amplitude in the frequency spectrum of the vibration data in the preset time length.
And then, step S230 is carried out, the normalized frequency spectrum of the vibration data is input into a trained fault detection model for processing, the fault category of the equipment to be detected is obtained, and the fault detection model is constructed on the basis of a metric element learning model.
The metric meta-learning model is a meta-learning model based on metrics, and mainly comprises a prototype Network (Prototypical Networks), a Matching Network (Matching Networks) and a relationship Network (relationship Networks). The prototype network projects a Support set (Support set) to a measurement space by using a clustering idea, obtains a vector mean value on the basis of Euclidean distance measurement, calculates the distance from a test sample to each prototype and realizes classification. After feature extraction is carried out on the support set, the matching network measures the distance of Cosine (Cosine) in an embedding (embedding) space, and classification is realized by calculating the matching degree of the test samples. The cosine and Euclidean distance measurement in the matching network and the prototype network is replaced by the relationship module structure provided by the relationship network, so that the relationship module structure becomes a learning nonlinear classifier for judging the relationship and realizing classification.
According to one embodiment of the invention, the fault detection model is constructed based on a prototype network. FIG. 3 shows a schematic diagram of a structure diagram of a fault detection model according to one embodiment of the invention. As shown in fig. 3, the fault detection model includes a feature extraction module and a metric module. The feature extraction module is formed by connecting four layers of convolution networks in series. Each convolutional network comprises a convolutional layer, a ReLu function for de-linearization and a max-pooling layer. The convolution kernel size may be set to 3 × 1, and the pooling kernel size may be set to 2 × 1. Of course, the invention is not limited with respect to the size of the convolution kernel and the size of the pooling kernel. In the specific embodiment, a person skilled in the art can set the setting according to actual needs. Therefore, the model discards the full connection layer, so that the calculation of parameters can be reduced. The constructed fault detection model can be trained through the following steps.
In the first step, a total training sample set is constructed. The total training sample set comprises a plurality of fault categories, and each fault category comprises a plurality of sample data. The sample data is a normalized frequency spectrum of vibration data within a preset time length when the industrial equipment operates. For the same fault category, the frequency spectrum of the vibration data within the preset time length during the operation of the industrial equipment can be obtained under different working conditions.
And secondly, constructing a meta-training set. After the total training sample set is constructed, a first preset number of fault categories are randomly extracted from the total training sample set, and for each extracted fault category, a second preset number of sample data are randomly extracted from the total training sample set to serve as a meta-training set.
And thirdly, dividing the meta-training set. Specifically, a third preset number of sample data in the second preset number of sample data is used as a support set of the meta-training, and the remaining sample data is used as a query set of the meta-training.
And fourthly, training a fault detection model. After the meta-training set is divided into a support set and a query set, the constructed fault detection model is trained by using the support set and the query set to obtain an initial fault model. Specifically, the method comprises the following steps:
firstly, inputting sample data under each fault category in a support set into a feature extraction module to obtain a feature vector of each sample data. Wherein the feature extraction module extracts an encoded representation (i.e., a feature vector) of each fault sample data by a convolution operation. The process can be expressed as an embedding function:
RD→RM. D is the dimension of original fault sample data, and M is the dimension of the code representation of the fault sample data learned by the fault detection model.
And then, acquiring prototype representation of each fault category in the support set based on the obtained feature vector of each sample data. Specifically, the feature vectors of all sample data in each fault category in the support set are averaged, and the obtained average value is used as a prototype representation of the fault category, as follows:
wherein, CkFor prototype representation of failure class k, m is the number of samples under failure class k in the support set, xiTo support the ith sample data in the failure category k in the set,
is sample data xiCharacteristic vector of (y)iIs sample data xiThe fault class to which it belongs, SkIs a collection of sample data under the supporting centralized failure category k.
After prototype representation of each fault category in the support set is obtained, sample data under each fault category in the query set are sequentially input into the feature extraction module, and feature vectors of each sample data in the query set are obtained.
And then, acquiring a predicted fault category corresponding to each sample data in the query set based on the feature vector of each sample data in the query set and the acquired prototype representation of each fault category. Specifically, after the feature vector of any sample data in the query set is obtained, the euclidean distance between the feature vector and the prototype representation of each fault category is calculated as follows:
wherein the content of the first and second substances,
a feature vector representing sample data z
Prototype representation C with failure class kkThe euclidean distance between them.
Then, all the obtained Euclidean distances are compared, and the fault class corresponding to the prototype representation with the minimum distance to the feature vector of the sample data is used as the predicted fault class of the sample data. For the feature vector of any sample data in the query set, after the feature vector obtains euclidean distances between the feature vector and prototype representations of all fault classes, each distance value can be converted into a probability value by using a classifier softmax, and then the maximum probability value is used as a classification result of the sample data, as shown in fig. 3, the fault class with the maximum probability value in fault 1, fault 2, and fault 3 … is used as the classification result of the sample data.
And after the predicted fault category corresponding to each sample data in the query set is obtained, updating the fault detection model based on the cross entropy loss between the predicted fault category corresponding to each sample data in the query set and the real fault category to obtain an initial fault detection model.
And fifthly, repeating the second step to the fourth step until reaching the preset times, and obtaining the trained fault detection model. Wherein the invention is not limited with respect to the size of the predetermined number of times. In the specific embodiment, a person skilled in the art can set the setting according to actual needs.
After a trained fault detection model is obtained, the normalized frequency spectrum of the vibration data within a preset time length when any equipment works is input into the model, and then the fault type of the equipment can be obtained. Among the common faults in the equipment are bearing faults, gear faults and power frequency faults. Common fault categories among bearing faults include inner race faults, outer race faults, cage faults, and rolling element faults. Common failure categories among gear failures include flank wear, flank pitting, root cracking, tooth breakage, flank spalling, and plastic deformation. Common types of faults in line frequency faults include imbalance (uneven distribution of the rotating body mass along the centerline of rotation is called imbalance, and other problems resulting from machine vibration or operation are called imbalance faults), misalignment (misalignment of the centerlines of the shafts coupled to each other), and looseness.
In summary, the invention builds a model based on a prototype network, trains the model by using a meta-learning strategy, optimizes the model by learning a plurality of meta-tasks, and obtains prototype representations of different faults by accumulating feature representations of the faults. Therefore, when the data to be measured is input into the model, the prototype representation with the minimum distance from the characteristic representation of the data to be measured can be obtained, and the fault type corresponding to the prototype representation is the fault type of the data to be measured.
Therefore, the invention trains the prototype network by using the meta-training method and extracts the coding representation of the fault data, thereby solving the problems of poor robustness and low fault diagnosis accuracy of the deep neural network under the condition of small samples. In addition, the fault detection model adopts a simple four-layer convolution network structure, and simultaneously abandons a full connection layer, thereby avoiding network parameter explosion and effectively reducing the requirement of computing resources and the training time.
In order to better understand the training method of the fault detection model of the present invention, it is described below with a specific example. There are K types of faults in the training set T.
Firstly, a mini-batcn method is adopted to randomly extract M types from K types of faults as the types of primary mini-batch training. For the M fault classes, N + P samples are randomly extracted under each fault class, wherein the N samples serve as a support set, and the P samples serve as a query set.
The fault detection model then computes a prototype representation of each fault using the N support samples for each fault.
Then, each sample in the query set is input into the fault detection model, and the coded representation of each query sample is obtained. Distances between the coded representation of each query sample and the prototype representation of each fault are calculated and the smallest of these distances is found. And taking the fault class corresponding to the prototype representation with the minimum distance to the coded representation of each query sample as a classification result of the fault detection model for the query sample.
And finally, acquiring the cross entropy loss between the classification result of each query sample and the real class of the query sample, and updating the fault detection model through the cross entropy losses of the M × P query samples.
After the fault detection model is updated, the steps are repeated by using the updated fault detection model, namely a new meta-training set is extracted, new fault prototype representation, distance and the like are calculated, and a final fault detection model is trained.
According to the fault detection method of the industrial equipment, firstly, the frequency spectrum of the vibration data within the preset time length when the equipment to be detected operates is obtained. Then, the frequency spectrum of the vibration data is normalized to obtain the normalized frequency spectrum of the vibration data. And finally, acquiring the fault category of the equipment to be detected by using a fault detection model constructed based on the metric element learning model according to the normalized frequency spectrum of the vibration data. Therefore, the fault detection method provided by the invention adopts the fault detection model constructed based on the metric learning model, and can be used for rapidly applying the fault detection learned in the training process to the fault detection under the new working condition, so that a good fault detection effect can be obtained under the condition of a small sample, and the fault detection accuracy under the condition of the small sample is improved.
Moreover, the invention carries out normalization processing on the frequency spectrum of the vibration data of the equipment to be detected, thus eliminating the influence on fault detection caused by different amplitude ranges of the frequency spectrum, and further improving the accuracy of fault detection.
The various techniques described herein may be implemented in connection with hardware or software or, alternatively, with a combination of both. Thus, the methods and apparatus of the present invention, or certain aspects or portions thereof, may take the form of program code (i.e., instructions) embodied in tangible media, such as removable hard drives, U.S. disks, floppy disks, CD-ROMs, or any other machine-readable storage medium, wherein, when the program is loaded into and executed by a machine, such as a computer, the machine becomes an apparatus for practicing the invention.
In the case of program code execution on programmable computers, the computing device will generally include a processor, a storage medium readable by the processor (including volatile and non-volatile memory and/or storage elements), at least one input device, and at least one output device. Wherein the memory is configured to store program code; the processor is configured to execute the document loading method of the present invention according to instructions in the program code stored in the memory.
By way of example, and not limitation, readable media may comprise readable storage media and communication media. Readable storage media store information such as computer readable instructions, data structures, program modules or other data. Communication media typically embodies computer readable instructions, data structures, program modules or other data in a modulated data signal such as a carrier wave or other transport mechanism and includes any information delivery media. Combinations of any of the above are also included within the scope of readable media.
In the description provided herein, algorithms and displays are not inherently related to any particular computer, virtual system, or other apparatus. Various general purpose systems may also be used with examples of this invention. The required structure for constructing such a system will be apparent from the description above. Moreover, the present invention is not directed to any particular programming language. It is appreciated that a variety of programming languages may be used to implement the teachings of the present invention as described herein, and any descriptions of specific languages are provided above to disclose the best mode of the invention.
In the description provided herein, numerous specific details are set forth. It is understood, however, that embodiments of the invention may be practiced without these specific details. In some instances, well-known methods, structures and techniques have not been shown in detail in order not to obscure an understanding of this description.
It should be appreciated that in the foregoing description of exemplary embodiments of the invention, various features of the invention are sometimes grouped together in a single embodiment, figure, or description thereof for the purpose of streamlining the disclosure and aiding in the understanding of one or more of the various inventive aspects. However, the disclosed method should not be interpreted as reflecting an intention that: that the invention as claimed requires more features than are expressly recited in each claim. Rather, as the following claims reflect, inventive aspects lie in less than all features of a single foregoing disclosed embodiment. Thus, the claims following the detailed description are hereby expressly incorporated into this detailed description, with each claim standing on its own as a separate embodiment of this invention.
Those skilled in the art will appreciate that the modules or units or components of the devices in the examples disclosed herein may be arranged in a device as described in this embodiment or alternatively may be located in one or more devices different from the devices in this example. The modules in the foregoing examples may be combined into one module or may be further divided into multiple sub-modules.
Those skilled in the art will appreciate that the modules in the device in an embodiment may be adaptively changed and disposed in one or more devices different from the embodiment. The modules or units or components of the embodiments may be combined into one module or unit or component, and furthermore they may be divided into a plurality of sub-modules or sub-units or sub-components. All of the features disclosed in this specification (including any accompanying claims, abstract and drawings), and all of the processes or elements of any method or apparatus so disclosed, may be combined in any combination, except combinations where at least some of such features and/or processes or elements are mutually exclusive. Each feature disclosed in this specification (including any accompanying claims, abstract and drawings) may be replaced by alternative features serving the same, equivalent or similar purpose, unless expressly stated otherwise.
Furthermore, those skilled in the art will appreciate that while some embodiments described herein include some features included in other embodiments, rather than other features, combinations of features of different embodiments are meant to be within the scope of the invention and form different embodiments. For example, in the following claims, any of the claimed embodiments may be used in any combination.
Furthermore, some of the described embodiments are described herein as a method or combination of method elements that can be performed by a processor of a computer system or by other means of performing the described functions. A processor having the necessary instructions for carrying out the method or method elements thus forms a means for carrying out the method or method elements. Further, the elements of the apparatus embodiments described herein are examples of the following apparatus: the apparatus is used to implement the functions performed by the elements for the purpose of carrying out the invention.
As used herein, unless otherwise specified the use of the ordinal adjectives "first", "second", "third", etc., to describe a common object, merely indicate that different instances of like objects are being referred to, and are not intended to imply that the objects so described must be in a given sequence, either temporally, spatially, in ranking, or in any other manner.
While the invention has been described with respect to a limited number of embodiments, those skilled in the art, having benefit of this description, will appreciate that other embodiments can be devised which do not depart from the scope of the invention as described herein. Furthermore, it should be noted that the language used in the specification has been principally selected for readability and instructional purposes, and may not have been selected to delineate or circumscribe the inventive subject matter. Accordingly, many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the appended claims. The present invention has been disclosed in an illustrative rather than a restrictive sense, and the scope of the present invention is defined by the appended claims.
Claims (10)
1. A fault detection method of an industrial device, adapted to be executed in a computing device, the method comprising:
acquiring a frequency spectrum of vibration data within a preset time length when a device to be detected operates;
normalizing the frequency spectrum of the vibration data to obtain a normalized frequency spectrum of the vibration data;
and inputting the normalized frequency spectrum of the vibration data into a trained fault detection model for processing to obtain the fault category of the equipment to be detected, wherein the fault detection model is constructed based on a metric learning model.
2. The method of claim 1, wherein the step of obtaining the frequency spectrum of the vibration data within a preset time period when the device to be detected operates comprises:
obtaining vibration data within a preset time length when a device to be detected operates;
and acquiring the frequency spectrum of the vibration data by using short-time Fourier transform.
3. The method of claim 1 or 2, wherein the vibration data is acceleration data.
4. The method of any of claims 1-3, wherein the metric learning model is a prototype network.
5. The method of claim 4, wherein the fault detection model comprises a feature extraction module and a metrics module.
6. The method of claim 5, wherein the feature extraction module comprises four convolutional networks, each convolutional network comprising one convolutional layer, one active layer, and one pooling layer.
7. The method according to any of claims 4-6, wherein the fault detection model is trained based on the following method:
randomly extracting a first preset number of fault types from a total training sample set, wherein each fault type in the total training sample set comprises a plurality of sample data, and the sample data is a normalized frequency spectrum of vibration data within a preset time length when the industrial equipment operates;
for each extracted fault category, randomly extracting a second preset number of sample data from the total training sample set;
taking a third preset number of sample data in the second preset number of sample data as a meta-training support set, and taking the remaining sample data as a meta-training query set;
training the fault detection model by using the support set and the query set to obtain an initial fault detection model;
and repeating the steps until the preset times are reached, and obtaining the trained fault detection model.
8. The method of claim 7, wherein the step of training the fault detection model using the support set and the query set comprises:
inputting the sample data under each fault category in the support set into the feature extraction module to obtain a feature vector of each sample data;
acquiring prototype representation of each fault category in the support set based on the obtained feature vector of each sample data;
inputting sample data under each fault category in the query set into the feature extraction module to obtain a feature vector of each sample data in the query set;
acquiring a predicted fault category corresponding to each sample data in the query set based on the feature vector of each sample data in the query set and the acquired prototype representation of each fault category;
and updating the fault detection model based on the loss value between the predicted fault category and the real fault category corresponding to each sample data in the query set to obtain an initial fault detection model.
9. A computing device, comprising:
at least one processor; and
a memory storing program instructions, wherein the program instructions are configured to be executed by the at least one processor, the program instructions comprising instructions for performing the method of any of claims 1-8.
10. A readable storage medium storing program instructions that, when read and executed by a computing device, cause the computing device to perform the method of any of claims 1-8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110876571.1A CN113569486A (en) | 2021-07-30 | 2021-07-30 | Fault detection method of industrial equipment, computing equipment and readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110876571.1A CN113569486A (en) | 2021-07-30 | 2021-07-30 | Fault detection method of industrial equipment, computing equipment and readable storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113569486A true CN113569486A (en) | 2021-10-29 |
Family
ID=78169783
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110876571.1A Pending CN113569486A (en) | 2021-07-30 | 2021-07-30 | Fault detection method of industrial equipment, computing equipment and readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113569486A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114417931A (en) * | 2022-01-21 | 2022-04-29 | 石家庄铁道大学 | Bearing fault diagnosis method based on prototype network |
| CN114898136A (en) * | 2022-03-14 | 2022-08-12 | 武汉理工大学 | Small sample image classification method based on feature self-adaption |
| CN115062678A (en) * | 2022-08-19 | 2022-09-16 | 山东能源数智云科技有限公司 | Training method of equipment fault detection model, fault detection method and device |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111709448A (en) * | 2020-05-20 | 2020-09-25 | 西安交通大学 | Mechanical fault diagnosis method based on migration relation network |
| CN112084974A (en) * | 2020-09-14 | 2020-12-15 | 北京工商大学 | Multi-label rolling bearing fault diagnosis method based on meta-learning |
| CN112270334A (en) * | 2020-08-07 | 2021-01-26 | 武汉理工大学 | Few-sample image classification method and system based on abnormal point exposure |
| CN112418013A (en) * | 2020-11-09 | 2021-02-26 | 贵州大学 | Complex working condition bearing fault diagnosis method based on meta-learning under small sample |
| CN112989792A (en) * | 2021-04-25 | 2021-06-18 | 中国人民解放军国防科技大学 | Case detection method and electronic equipment |
| CN113095179A (en) * | 2021-03-30 | 2021-07-09 | 西安交通大学 | Metametric learning driven helicopter planetary gearbox fault diagnosis method |
| CN113111820A (en) * | 2021-04-21 | 2021-07-13 | 北京科技大学 | Rotary component fault diagnosis method and device based on improved CNN and relation module |
| US20210234365A1 (en) * | 2020-01-29 | 2021-07-29 | Witricity Corporation | Systems and methods for auxiliary power dropout protection |
-
2021
- 2021-07-30 CN CN202110876571.1A patent/CN113569486A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20210234365A1 (en) * | 2020-01-29 | 2021-07-29 | Witricity Corporation | Systems and methods for auxiliary power dropout protection |
| CN111709448A (en) * | 2020-05-20 | 2020-09-25 | 西安交通大学 | Mechanical fault diagnosis method based on migration relation network |
| CN112270334A (en) * | 2020-08-07 | 2021-01-26 | 武汉理工大学 | Few-sample image classification method and system based on abnormal point exposure |
| CN112084974A (en) * | 2020-09-14 | 2020-12-15 | 北京工商大学 | Multi-label rolling bearing fault diagnosis method based on meta-learning |
| CN112418013A (en) * | 2020-11-09 | 2021-02-26 | 贵州大学 | Complex working condition bearing fault diagnosis method based on meta-learning under small sample |
| CN113095179A (en) * | 2021-03-30 | 2021-07-09 | 西安交通大学 | Metametric learning driven helicopter planetary gearbox fault diagnosis method |
| CN113111820A (en) * | 2021-04-21 | 2021-07-13 | 北京科技大学 | Rotary component fault diagnosis method and device based on improved CNN and relation module |
| CN112989792A (en) * | 2021-04-25 | 2021-06-18 | 中国人民解放军国防科技大学 | Case detection method and electronic equipment |
Non-Patent Citations (2)
| Title |
|---|
| DUO WANG等: "Metric-based meta-learning model for few-shot fault diagnosis under multiple limited data conditions", MECHANICAL SYSTEMS AND SIGNAL PROCESSING, pages 1 - 33 * |
| 余浩帅等: "小样本下混合自注意力原型网络的风电齿轮箱故障诊断方法", 中国机械工程, pages 1 - 9 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114417931A (en) * | 2022-01-21 | 2022-04-29 | 石家庄铁道大学 | Bearing fault diagnosis method based on prototype network |
| CN114898136A (en) * | 2022-03-14 | 2022-08-12 | 武汉理工大学 | Small sample image classification method based on feature self-adaption |
| CN114898136B (en) * | 2022-03-14 | 2024-04-19 | 武汉理工大学 | Small sample image classification method based on characteristic self-adaption |
| CN115062678A (en) * | 2022-08-19 | 2022-09-16 | 山东能源数智云科技有限公司 | Training method of equipment fault detection model, fault detection method and device |
| CN115062678B (en) * | 2022-08-19 | 2022-11-04 | 山东能源数智云科技有限公司 | Training method of equipment fault detection model, fault detection method and device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113569486A (en) | Fault detection method of industrial equipment, computing equipment and readable storage medium | |
| Rajabi et al. | Fault diagnosis in industrial rotating equipment based on permutation entropy, signal processing and multi-output neuro-fuzzy classifier | |
| CN111458142B (en) | Sliding bearing fault diagnosis method based on generation of countermeasure network and convolutional neural network | |
| CN113159147B (en) | Image recognition method and device based on neural network and electronic equipment | |
| CN110110804B (en) | Flight control system residual life prediction method based on CNN and LSTM | |
| CN110033018B (en) | Graph similarity judging method and device and computer readable storage medium | |
| CN111144548B (en) | Method and device for identifying working condition of oil pumping well | |
| CN111814956B (en) | Multi-task learning air quality prediction method based on multi-dimensional secondary feature extraction | |
| Wang et al. | Intelligent fault diagnosis for planetary gearbox using transferable deep q network under variable conditions with small training data | |
| CN112364974B (en) | YOLOv3 algorithm based on activation function improvement | |
| Jiménez et al. | Machine learning and neural network for maintenance management | |
| CN110826624A (en) | Time series classification method based on deep reinforcement learning | |
| CN116186578A (en) | Wind turbine generator system fault identification method and device and terminal equipment | |
| CN113127552A (en) | Food safety identification method and system based on big data | |
| CN114662386A (en) | Bearing fault diagnosis method and system | |
| CN112633362B (en) | Rotary machine self-adaptive fault diagnosis method, device, equipment and medium | |
| CN104504696A (en) | Embedded parallel optimization method for image salient region detection | |
| CN116541838A (en) | Malware detection method based on contrast learning | |
| Wu et al. | A fault diagnosis method based on Auxiliary Classifier Generative Adversarial Network for rolling bearing | |
| CN113554075A (en) | Method for generating fault classification model and method for determining fault type of rotating equipment | |
| Hao et al. | New fusion features convolutional neural network with high generalization ability on rolling bearing fault diagnosis | |
| CN115146641A (en) | NER noisy training method and device | |
| CN114595730A (en) | Rolling bearing fault diagnosis method based on GAF-DRSN | |
| CN113233270A (en) | Elevator internal and external judgment method based on robot running safety and related equipment | |
| CN111178630A (en) | Load prediction 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 |