Health detection method and device for industrial equipment, and storage medium

Abstract

The invention provides a health detection method and device for industrial equipment and a storage medium, and is applied to the technical field of industrial equipment, and the method comprises the steps: obtaining first feature data corresponding to to-be-detected industrial equipment; wherein the first feature data is determined based on vibration data corresponding to the to-be-detected industrial equipment; inputting the first feature data into a health detection model to obtain a health detection result output by the health detection model; wherein the health detection model is obtained by performing unsupervised learning and semi-supervised learning training on an initial model, and the initial model is a type of support vector machine. The AI-based data-driven development mode has relatively high development efficiency, so that the problem of relatively low development efficiency of a traditional expert system is solved. Besides, in the process of obtaining the health detection model through the catenary, an unsupervised learning mode and a semi-supervised learning mode are combined, so that the confidence coefficient of the label is improved, and the cost of performing health examination on the industrial equipment can be reduced.

Description

Technical Field

[0001] The present application relates to the technical field of industrial equipment, and in particular to a health detection method and device for industrial equipment, and a storage medium. Background Art

[0002] In recent years, with the introduction of the concept of Industry 4.0, more and more industrial equipment has been connected to the Prognostics and Health Management (PHM) system. The system regularly issues equipment physical examination reports and pushes them to the site so that operation and maintenance personnel can understand the health status of the equipment in a timely manner. Most of the current PHM systems are expert systems. The system simulates the diagnostic engineer's interpretation of the equipment signals stored in the central platform from a human perspective, extracts various indicator values ​​based on the vibration signals, and combines multiple rule groups to develop an expert system with a large number of parameters. The diagnostic engineer writes the physical examination report based on the physical examination results of the expert system.

[0003] Although the expert system can accurately give the health status of the equipment, with the continuous increase in indicators and policy rules, there are a large number of artificial hyperparameters in the expert system. The reasonable configuration of effective hyperparameter values ​​requires a high level of business awareness and parameter adjustment experience from the developers, which increases the development cost of the expert system. In addition, with the increase in the number of devices to be monitored, the algorithm iteration cycle continues to lengthen, and the original development method that relies on manpower is inefficient. Summary of the invention

[0004] The purpose of the embodiments of the present application is to provide a health detection method and device for industrial equipment, and a storage medium, so as to solve the technical problems in the prior art of high cost and low development efficiency of health checks on industrial equipment.

[0005] In a first aspect, an embodiment of the present application provides a health detection method for industrial equipment, comprising: obtaining first feature data corresponding to the industrial equipment to be detected; wherein the first feature data is determined based on vibration data corresponding to the industrial equipment to be detected; inputting the first feature data into a health detection model to obtain a health detection result output by the health detection model; wherein the health detection model is obtained by performing unsupervised learning and semi-supervised learning training on an initial model, and the initial model is a first-class support vector machine.

[0006] In the above scheme, the health detection model can be used directly based on the first feature data corresponding to the industrial equipment to be detected to achieve the purpose of detecting the health status of the industrial equipment to be detected, wherein the data-driven development method based on artificial intelligence (AI) has a high development efficiency, thereby solving the problem of low development efficiency of traditional expert systems. In addition, in the process of obtaining the health detection model of the catenary, the unsupervised learning method and the semi-supervised learning method are combined to improve the confidence of the label, thereby reducing the cost of health checks on industrial equipment.

[0007] In an optional embodiment, the obtaining of the first characteristic data corresponding to the industrial equipment to be detected includes: obtaining the vibration data corresponding to the industrial equipment to be detected; performing data cleaning on the vibration data to obtain the time domain data corresponding to the industrial equipment to be detected; performing Fourier transform on the time domain data to obtain the frequency domain data and time-frequency domain data corresponding to the industrial equipment to be detected; and determining the first characteristic data based on the time domain data, the frequency domain data, and the time-frequency domain data. In the above scheme, in the process of extracting the corresponding time domain data based on the vibration data corresponding to the industrial equipment to be detected, the high-order eigenvalues ​​of the waveform or spectrum can be used, thereby avoiding the problem of feature redundancy and large amount of calculation caused by using the original waveform or spectrum with too large dimensions, thereby improving the calculation efficiency and effect of the algorithm.

[0008] In an optional embodiment, the determining of the first feature data based on the time domain data, the frequency domain data and the time-frequency domain data includes: respectively extracting the time domain features corresponding to the time domain data, the frequency domain features corresponding to the frequency domain data and the time-frequency domain features corresponding to the time-frequency domain data; fusing the time domain features, the frequency domain features and the time-frequency domain features to obtain the first feature data. In the above scheme, in the process of extracting the corresponding time domain data based on the vibration data corresponding to the industrial equipment to be detected, the multi-dimensional statistical features of the time domain signal, the frequency domain signal and the time-frequency domain signal can be fused, which provides a more comprehensive data consideration compared to the scheme of using only time domain or frequency domain signal features, thereby improving the accuracy of the health detection model. In addition, by extracting high-dimensional features of data of three signal types, namely, time domain, frequency domain and time-frequency domain, the information contained in the feature dimension can be made more comprehensive.

[0009] In an optional embodiment, the health detection method for industrial equipment further includes: training the initial model using the following steps to obtain the health detection model: obtaining second feature data corresponding to healthy industrial equipment; training the initial model using the second feature data to obtain an intermediate model; obtaining third feature data and label data corresponding to sample industrial equipment; wherein the sample industrial equipment is industrial equipment with unknown health status; training the intermediate model using the third feature data and the label data to obtain the health detection model. In the above scheme, in the process of training the initial model, firstly, an unsupervised learning method can be used to train the initial model using the second feature data corresponding to the healthy industrial equipment, and then a semi-supervised learning method can be used to use the initial model to predict the results of unlabeled data, and label the unlabeled data according to the posterior probability value, thereby effectively reducing the labeling cost and providing more data support to help the model capture richer data distribution characteristics, thereby improving the generalization ability of the model.

[0010] In an optional embodiment, the obtaining of the third characteristic data and label data corresponding to the sample industrial equipment includes: obtaining the vibration sample data corresponding to the sample industrial equipment; determining the third characteristic data according to the vibration sample data; inputting the third characteristic data into the initial model to obtain the classification result corresponding to the third characteristic data; and determining the label data according to the classification result. In the above scheme, semi-supervised learning can be used for model training, which combines the advantages of supervised learning and unsupervised learning, and effectively reduces the labeling cost of faulty equipment data.

[0011] In an optional embodiment, determining the label data according to the classification result includes: converting the classification result into a posterior probability value using a data fitting method; and labeling the third feature data according to the posterior probability value and the classification threshold to obtain the label data corresponding to the third feature data. In the above scheme, the classification result can be post-processed using a data fitting method and converted into a posterior probability value, thereby providing confidence for the label of the data.

[0012] In a second aspect, an embodiment of the present application provides a health detection device for industrial equipment, comprising: an acquisition module, used to acquire first feature data corresponding to the industrial equipment to be detected; wherein the first feature data is determined based on vibration data corresponding to the industrial equipment to be detected; an input module, used to input the first feature data into a health detection model to obtain a health detection result output by the health detection model; wherein the health detection model is obtained by performing unsupervised learning and semi-supervised learning training on an initial model, and the initial model is a first-class support vector machine.

[0013] In the above scheme, the health detection model can be used directly based on the first feature data corresponding to the industrial equipment to be detected to achieve the purpose of detecting the health status of the industrial equipment to be detected, wherein the data-driven development method based on artificial intelligence (AI) has a high development efficiency, thereby solving the problem of low development efficiency of traditional expert systems. In addition, in the process of obtaining the health detection model of the catenary, the unsupervised learning method and the semi-supervised learning method are combined to improve the confidence of the label, thereby reducing the cost of health checks on industrial equipment.

[0014] In an optional implementation, the acquisition module is specifically used to: acquire the vibration data corresponding to the industrial equipment to be detected; perform data cleaning on the vibration data to obtain the time domain data corresponding to the industrial equipment to be detected; perform Fourier transform on the time domain data to obtain the frequency domain data and time-frequency domain data corresponding to the industrial equipment to be detected; determine the first feature data based on the time domain data, the frequency domain data and the time-frequency domain data. In the above scheme, in the process of extracting the corresponding time domain data based on the vibration data corresponding to the industrial equipment to be detected, the high-order eigenvalues ​​of the waveform or spectrum can be used, thereby avoiding the problem of feature redundancy and large amount of calculation caused by using the original waveform or spectrum with too large dimensions, thereby improving the calculation efficiency and effect of the algorithm.

[0015] In an optional embodiment, the acquisition module is also used to: respectively extract the time domain features corresponding to the time domain data, the frequency domain features corresponding to the frequency domain data, and the time and frequency domain features corresponding to the time and frequency domain data; the time domain features, the frequency domain features, and the time and frequency domain features are fused to obtain the first feature data. In the above scheme, in the process of extracting the corresponding time domain data based on the vibration data corresponding to the industrial equipment to be detected, the multi-dimensional statistical features of the time domain signal, the frequency domain signal, and the time and frequency domain signal can be fused. Compared with the scheme of using only time domain or frequency domain signal features, a more comprehensive data consideration is provided, thereby improving the accuracy of the health detection model. In addition, by extracting high-dimensional features of data of three signal types, namely, time domain, frequency domain, and time and frequency domain, the information contained in the feature dimension can be made more comprehensive.

[0016] In an optional embodiment, the health detection device for industrial equipment further includes: a training module for training the initial model using the following steps to obtain the health detection model: obtaining the second feature data corresponding to the healthy industrial equipment; training the initial model using the second feature data to obtain an intermediate model; obtaining the third feature data and label data corresponding to the sample industrial equipment; wherein the sample industrial equipment is an industrial equipment with an unknown health status; training the intermediate model using the third feature data and the label data to obtain the health detection model. In the above scheme, in the process of training the initial model, firstly, an unsupervised learning method can be used to train the initial model using the second feature data corresponding to the healthy industrial equipment, and then a semi-supervised learning method can be used to use the initial model to predict the results of the unlabeled data, and the unlabeled data can be labeled according to the posterior probability value, thereby effectively reducing the labeling cost and providing more data support to help the model capture richer data distribution characteristics, thereby improving the generalization ability of the model.

[0017] In an optional implementation, the training module is specifically used to: obtain vibration sample data corresponding to the sample industrial equipment; determine the third feature data according to the vibration sample data; input the third feature data into the initial model to obtain a classification result corresponding to the third feature data; and determine the label data according to the classification result. In the above scheme, semi-supervised learning can be used for model training, which combines the advantages of supervised learning and unsupervised learning, and effectively reduces the labeling cost of faulty equipment data.

[0018] In an optional implementation, the training module is further used to: convert the classification result into a posterior probability value using a data fitting method; annotate the third feature data according to the posterior probability value and the classification threshold to obtain the label data corresponding to the third feature data. In the above scheme, the classification result can be post-processed using a data fitting method and converted into a posterior probability value, thereby providing confidence for the label of the data.

[0019] In a third aspect, an embodiment of the present application provides a computer program product, including computer program instructions, which, when read and executed by a processor, execute the health detection method for industrial equipment as described in the first aspect.

[0020] In a fourth aspect, an embodiment of the present application provides an electronic device, comprising: a processor, a memory and a bus; the processor and the memory communicate with each other through the bus; the memory stores computer program instructions that can be executed by the processor, and the processor calls the computer program instructions to execute the health detection method for industrial equipment as described in the first aspect.

[0021] In a fifth aspect, an embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer program instructions. When the computer program instructions are executed by a computer, the computer executes the health detection method for industrial equipment as described in the first aspect.

[0022] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the following specifically cites the embodiments of the present application and describes them in detail with reference to the attached drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments of the present application will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0024] Figure 1 A flowchart of a health detection method for industrial equipment provided in an embodiment of the present application;

[0025] Figure 2 A structural block diagram of a health detection device for industrial equipment provided in an embodiment of the present application;

[0026] Figure 3 A structural block diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0027] The technical solutions in the embodiments of the present application will be described below in conjunction with the drawings in the embodiments of the present application.

[0028] Please refer to Figure 1 , Figure 1 A flowchart of a health detection method for industrial equipment provided in an embodiment of the present application, the health detection method for industrial equipment may include the following steps:

[0029] Step S101: obtaining first characteristic data corresponding to the industrial equipment to be detected; wherein the first characteristic data is determined based on vibration data corresponding to the industrial equipment to be detected.

[0030] Step S102: input the first feature data into the health detection model to obtain the health detection result output by the health detection model; wherein the health detection model is obtained by performing unsupervised learning and semi-supervised learning training on the initial model, and the initial model is a first-class support vector machine.

[0031] Specifically, the first characteristic data in step S101 can be determined based on vibration data corresponding to the industrial equipment to be detected. The industrial equipment to be detected refers to the industrial equipment currently being tested for health status, wherein the industrial equipment can be various equipment used in industry, such as centrifugal pumps and centrifugal fans.

[0032] Vibration is a common phenomenon. For industrial equipment, most failures are closely related to mechanical movement or vibration. When the internal parts of the equipment undergo subtle changes such as wear, misalignment, looseness, or poor sealing, the energy of the vibration will change accordingly. Therefore, by monitoring the vibration data of industrial equipment, potential failures of industrial equipment can be discovered in a timely manner, thereby realizing the detection of the health status of industrial equipment.

[0033] As an implementation method, the above vibration data can be collected by a vibration sensor deployed on the industrial equipment, wherein the vibration sensor can be an acceleration sensor, a displacement sensor, or other sensors. As another implementation method, the sensor deployed on the industrial equipment can set a collection cycle to regularly collect vibration data of the industrial equipment, wherein the collection frequency and collection duration can be set according to demand, and the collected data can be packaged and uploaded to a database for storage.

[0034] It should be noted that the embodiments of the present application do not specifically limit the specific implementation of the first characteristic data, and those skilled in the art can make appropriate adjustments according to actual conditions. For example, the first characteristic data may include one or more of time domain data, frequency domain data, and time-frequency domain data; or, the first characteristic data may include one or more of maximum value, maximum absolute value, minimum value, mean value, frequency root mean square, average frequency, frequency variance, etc.

[0035] In addition, the specific implementation method of obtaining the first characteristic data corresponding to the industrial equipment to be detected is also not specifically limited in the embodiment of the present application, and those skilled in the art can make appropriate adjustments according to actual conditions. For example, the first characteristic data sent by an external device can be received; or the data collection function of the sensor can be started to collect vibration data and determine the first characteristic data; or the first characteristic data stored locally or in the cloud can be read.

[0036] In the above step S102, the health detection model is obtained by performing unsupervised learning and semi-supervised learning training on the initial model. By inputting the first feature data obtained in step S102 into the above health detection model, the health detection result output by the above health detection model can be obtained. As an implementation method, the above health monitoring results can be packaged and uploaded to a database for storage.

[0037] The initial model may be a class-one support vector machine (class-one support vector machine), which is an anomaly detection model for identifying anomalies or outliers in a data set. As an implementation, the initial model may construct a classifier that divides data into two categories: normal samples and abnormal samples, where the abnormal samples are samples that are significantly different from normal samples.

[0038] In the above scheme, the health detection model can be used directly based on the first feature data corresponding to the industrial equipment to be detected to achieve the purpose of detecting the health status of the industrial equipment to be detected, wherein the data-driven development method based on artificial intelligence (AI) has a high development efficiency, thereby solving the problem of low development efficiency of traditional expert systems. In addition, in the process of obtaining the health detection model of the catenary, the unsupervised learning method and the semi-supervised learning method are combined to improve the confidence of the label, thereby reducing the cost of health checks on industrial equipment.

[0039] Further, based on the above embodiment, the above step S101 may specifically include the following steps:

[0040] Step 1), obtaining vibration data corresponding to the industrial equipment to be detected.

[0041] Step 2), clean the vibration data to obtain the time domain data corresponding to the industrial equipment to be detected.

[0042] Step 3), perform Fourier transform on the time domain data to obtain frequency domain data and time-frequency domain data corresponding to the industrial equipment to be detected.

[0043] Step 4), determining first feature data based on the time domain data, the frequency domain data, and the time-frequency domain data.

[0044] Specifically, in the above step 1), the above vibration data can be collected by a vibration sensor deployed on the industrial equipment, wherein the vibration sensor can be an acceleration sensor, a displacement sensor, or other sensors. As another implementation, the sensor deployed on the industrial equipment can set a collection cycle to regularly collect vibration data of the industrial equipment, wherein the collection frequency and collection duration can be set according to demand, and the collected data can be packaged and uploaded to a database for storage.

[0045] It should be noted that the specific implementation method for obtaining the vibration data corresponding to the industrial equipment to be detected is also not specifically limited in the embodiments of the present application, and those skilled in the art can make appropriate adjustments according to actual conditions. For example, vibration data sent by an external device can be received; or the data collection function of the sensor can be started to collect vibration data; or vibration data pre-stored locally or in the cloud can be read.

[0046] In the above step 2), the vibration data can be cleaned to obtain cleaned time domain data. It should be noted that the specific implementation method of the above data cleaning is also not specifically limited in the embodiment of the present application, and those skilled in the art can make appropriate adjustments according to actual conditions. For example, data cleaning can include filtering shutdown values, filtering false signals, filtering abnormal values, etc.

[0047] It is understood that if the acquired data is an analog signal, it can also be converted into a digital signal and saved in an appropriate file format (such as CSV, Excel, etc.).

[0048] In the above step 3), the time domain data may be subjected to Fourier transformation, thereby obtaining the frequency domain data and the time-frequency domain data corresponding to the industrial equipment to be detected.

[0049] In the above step 4), the first characteristic data can be determined based on the time domain data obtained in the above step 2), and the frequency domain data and time-frequency domain data obtained in the above step 3). It should be noted that the embodiment of the present application also does not specifically limit the specific implementation method of determining the first characteristic data, and those skilled in the art can make appropriate adjustments according to actual conditions. For example, the first characteristic data can be obtained by fusing time domain data, frequency domain data, and time-frequency domain data; or, the first characteristic data can be obtained by extracting time domain data, frequency domain data, and special additions to time-frequency domain data, etc.

[0050] In the above scheme, in the process of extracting the corresponding time domain data based on the vibration data corresponding to the industrial equipment to be detected, the high-order eigenvalues ​​of the waveform or spectrum can be used, thereby avoiding the problem of feature redundancy and large amount of calculation caused by using the original waveform or spectrum with too large dimensions, thereby improving the computational efficiency and effect of the algorithm.

[0051] Further, based on the above embodiment, the step of determining the first characteristic data according to the time domain data, the frequency domain data and the time-frequency domain data may specifically include the following steps:

[0052] Step 1), respectively extracting the time domain features corresponding to the time domain data, the frequency domain features corresponding to the frequency domain data, and the time-frequency domain features corresponding to the time-frequency domain data.

[0053] Step 2), the time domain features, frequency domain features and time-frequency domain features are fused to obtain the first feature data.

[0054] Specifically, in the above step 1), the time domain features can be obtained by extracting corresponding feature values ​​from the waveform data of the time domain data, wherein the time domain features may include one or more features such as maximum value, maximum absolute value, minimum value, mean value, peak-to-peak value, absolute mean value, root mean square value, root amplitude, standard deviation, kurtosis, skewness, margin index, waveform index, pulse index, peak index, etc.

[0055] Frequency domain features can be obtained by using fast Fourier transform to convert the time domain vibration waveform into frequency domain data and calculating its corresponding feature values, wherein the frequency domain features may include one or more features such as center frequency, root mean square frequency, average frequency, frequency variance, etc.

[0056] The time-frequency domain features can be obtained by performing wavelet transform on the time-domain data, decomposing the vibration signal into components with different frequencies and time resolutions, and then calculating the percentage of frequency band energy of each layer of decomposition as the extracted time-frequency domain features.

[0057] In the above step 2), the first feature data can be obtained by fusing the time domain features, the frequency domain features, and the time-frequency domain features. It should be noted that the specific implementation method of the above feature fusion is also not specifically limited in the embodiment of the present application, and those skilled in the art can make appropriate adjustments according to actual conditions. For example, the above feature fusion can be performed by splicing or weighted summation.

[0058] In the above scheme, in the process of extracting the corresponding time domain data based on the vibration data corresponding to the industrial equipment to be detected, the multi-dimensional statistical features of the time domain signal, the frequency domain signal and the time-frequency domain signal can be integrated. Compared with the scheme of using only the time domain or frequency domain signal features, it provides a more comprehensive data consideration and thus can improve the accuracy of the health detection model. In addition, by extracting high-dimensional features of data of three signal types, namely, the time domain, the frequency domain and the time-frequency domain, the information contained in the feature dimension can be made more comprehensive.

[0059] Furthermore, based on the above embodiment, the initial model can be trained using the following steps to obtain a health detection model:

[0060] Step 1), obtaining second characteristic data corresponding to the healthy industrial equipment.

[0061] Step 2), using the second feature data to train the initial model to obtain an intermediate model.

[0062] Step 3), obtaining third characteristic data and label data corresponding to the sample industrial equipment; wherein the sample industrial equipment is an industrial equipment whose health status is unknown.

[0063] Step 4), use the third feature data and label data to train the intermediate model to obtain a health detection model.

[0064] Specifically, the second characteristic data in the above step 1) can be determined based on the vibration data corresponding to the healthy industrial equipment. The above healthy industrial equipment refers to industrial equipment with a normal health status, wherein the above industrial equipment can be various equipment used in industry such as centrifugal pumps and centrifugal fans.

[0065] As an implementation mode, the above-mentioned vibration data may be collected by a vibration sensor deployed on the industrial equipment, wherein the vibration sensor may be an acceleration sensor, a displacement sensor or other sensors.

[0066] It should be noted that the embodiments of the present application do not specifically limit the specific implementation of the second characteristic data, and those skilled in the art can make appropriate adjustments according to actual conditions. For example, the second characteristic data may include one or more of time domain data, frequency domain data, and time-frequency domain data; or, the second characteristic data may include one or more of maximum value, maximum absolute value, minimum value, mean value, frequency root mean square, average frequency, frequency variance, etc.

[0067] In addition, the specific implementation method for obtaining the second characteristic data corresponding to the healthy industrial equipment is also not specifically limited in the embodiment of the present application, and those skilled in the art can make appropriate adjustments according to actual conditions. For example, the second characteristic data sent by the external device can be received; or the data collection function of the sensor can be started to collect vibration data and determine the second characteristic data; or the second characteristic data stored locally or in the cloud can be read.

[0068] In the above step 2), the above second feature data can be used to perform unsupervised learning training on the initial model to obtain an intermediate model. It should be noted that the specific implementation method of the unsupervised learning training of the initial model in the embodiment of the present application is not specifically limited, and those skilled in the art can make appropriate adjustments based on the prior art.

[0069] The third characteristic data in step 3) can be determined based on vibration data corresponding to sample industrial equipment. The sample industrial equipment refers to industrial equipment with unknown health status, wherein the industrial equipment can be various equipment used in industry, such as centrifugal pumps and centrifugal fans.

[0070] As an implementation mode, the above-mentioned vibration data may be collected by a vibration sensor deployed on the industrial equipment, wherein the vibration sensor may be an acceleration sensor, a displacement sensor or other sensors.

[0071] It should be noted that the embodiments of the present application do not specifically limit the specific implementation of the third characteristic data, and those skilled in the art can make appropriate adjustments according to actual conditions. For example, the third characteristic data may include one or more of time domain data, frequency domain data, and time-frequency domain data; or, the third characteristic data may include one or more of maximum value, maximum absolute value, minimum value, mean value, frequency root mean square, average frequency, frequency variance, etc.

[0072] In addition, the specific implementation method of obtaining the third characteristic data corresponding to the sample industrial equipment is also not specifically limited in the present application embodiment, and those skilled in the art can make appropriate adjustments according to actual conditions. For example, the third characteristic data sent by an external device can be received; or the data collection function of the sensor can be started to collect vibration data and determine the third characteristic data; or the third characteristic data pre-stored locally or in the cloud can be read, etc.

[0073] The above-mentioned label data refers to the category corresponding to the third characteristic data. In the embodiment of the present application, the label data corresponding to the third characteristic data can be determined by using the intermediate model, and its specific implementation will be introduced in the subsequent embodiments and will not be described here for the time being.

[0074] It should be noted that the embodiments of the present application do not specifically limit the specific implementation methods for obtaining the tag data corresponding to the sample industrial equipment, and those skilled in the art can make appropriate adjustments according to actual conditions. For example, the tag data sent by the external device can be received; or the data collection function of the sensor can be started to collect vibration data and determine the tag data; or the tag data pre-stored locally or in the cloud can be read.

[0075] In the above step 4), the intermediate model is trained using the third feature data and label data obtained in the above step 3) to obtain a health detection model. It should be noted that the specific implementation method of training the intermediate model in the embodiment of the present application is not specifically limited, and those skilled in the art can make appropriate adjustments based on the prior art.

[0076] In the above scheme, in the process of training the initial model, first, an unsupervised learning method can be used to train the initial model using the second feature data corresponding to the health industrial equipment. Then, a semi-supervised learning method can be used to use the initial model to predict the results of the unlabeled data and label the unlabeled data according to the posterior probability value. This can effectively reduce the labeling cost and provide more data support to help the model capture richer data distribution characteristics, thereby improving the generalization ability of the model.

[0077] Further, based on the above embodiment, the step of obtaining the third characteristic data and label data corresponding to the sample industrial equipment may specifically include the following steps:

[0078] Step 1), obtaining vibration sample data corresponding to sample industrial equipment.

[0079] Step 2), determining third characteristic data according to the vibration sample data.

[0080] Step 3), input the third feature data into the initial model to obtain the classification result corresponding to the third feature data.

[0081] Step 4), determine the label data according to the classification results.

[0082] Specifically, in the above step 1), the above vibration sample data can be collected by a vibration sensor deployed on the sample industrial equipment, wherein the vibration sensor can be an acceleration sensor, a displacement sensor, etc. As another implementation, the sensor deployed on the sample industrial equipment can set a collection cycle to regularly collect vibration data of the industrial equipment, wherein the collection frequency and collection duration can be set according to demand, and the collected data can be packaged and uploaded to a database for storage.

[0083] It should be noted that the embodiments of the present application do not specifically limit the specific implementation methods for obtaining vibration sample data corresponding to sample industrial equipment, and those skilled in the art may make appropriate adjustments according to actual conditions. For example, vibration sample data sent by an external device may be received; or a data acquisition function of a sensor may be activated to collect vibration sample data; or vibration sample data pre-stored locally or in the cloud may be read.

[0084] In the above step 2), the vibration sample data can be cleaned to obtain cleaned time domain data. It should be noted that the specific implementation method of the above data cleaning is also not specifically limited in the embodiment of the present application, and those skilled in the art can make appropriate adjustments according to actual conditions. For example, data cleaning can include filtering shutdown values, filtering false signals, filtering abnormal values, etc.

[0085] It is understood that if the acquired data is an analog signal, it can also be converted into a digital signal and saved in an appropriate file format (such as CSV, Excel, etc.).

[0086] By performing Fourier transform on the time domain data, the frequency domain data and time-frequency domain data corresponding to the sample industrial equipment can be obtained. Based on the above time domain data, frequency domain data and time-frequency domain data, the third characteristic data can be determined. It should be noted that the embodiment of the present application does not specifically limit the specific implementation method of determining the third characteristic data, and those skilled in the art can make appropriate adjustments according to actual conditions. For example, the third characteristic data can be obtained by fusing time domain data, frequency domain data and time-frequency domain data; or, the third characteristic data can be obtained by extracting the time domain data, frequency domain data and the special addition of time-frequency domain data, etc.

[0087] In the above step 4), the label data corresponding to the third feature data can be determined according to the classification result obtained in step 3).

[0088] In the above scheme, semi-supervised learning can be used for model training, which combines the advantages of supervised learning and unsupervised learning, and effectively reduces the labeling cost of faulty equipment data.

[0089] Further, based on the above embodiment, the step of determining the label data according to the classification result may specifically include the following steps:

[0090] Step 1), use the data fitting method to convert the classification results into posterior probability values.

[0091] Step 2), labeling the third feature data according to the posterior probability value and the classification threshold to obtain label data corresponding to the third feature data.

[0092] Specifically, the unsupervised learning classification method does not make full use of the existing data set, and there is unlabeled data that is not used. Therefore, in order to improve the classification effect of the model, a semi-supervised learning algorithm can also be used to iteratively optimize the model. That is, use the trained initial model to label the unlabeled samples. And use data fitting methods (such as sigmoid-fitting, etc.) to post-process the classification results and convert them into posterior probability values. Set a classification threshold for the posterior probability. When the posterior probability of a sample belonging to a certain category is greater than the classification threshold, the sample is marked as this category.

[0093] For example, the classification threshold is set to 0.8. If sample a outputs a posterior probability value of 0 for category 0 and a posterior probability value of 1 for category 0 after passing through the intermediate model, then the posterior probability value of category 0 is greater than the classification threshold, and the label of sample a is marked as 0; if sample b outputs a posterior probability value of 0 for category 0 and a posterior probability value of 1 for category 0 after passing through the intermediate model, then the label of sample b is marked as 1. If sample a outputs a posterior probability value of 0 for category 0 and a posterior probability value of 1 for category 0 after passing through the intermediate model, then the posterior empirical value has no label greater than the threshold, and the sample is discarded.

[0094] In the above scheme, the classification results can be post-processed using a data fitting method and converted into posterior probability values, thereby providing confidence for the data labels.

[0095] Please refer to Figure 2 , Figure 2 A structural block diagram of a health detection device for industrial equipment provided in an embodiment of the present application, the health detection device 200 for industrial equipment includes: an acquisition module 201, used to obtain first feature data corresponding to the industrial equipment to be detected; wherein the first feature data is determined based on vibration data corresponding to the industrial equipment to be detected; an input module 202, used to input the first feature data into a health detection model to obtain a health detection result output by the health detection model; wherein the health detection model is obtained by performing unsupervised learning and semi-supervised learning training on an initial model, and the initial model is a first-class support vector machine.

[0096] In the above scheme, the health detection model can be used directly based on the first feature data corresponding to the industrial equipment to be detected to achieve the purpose of detecting the health status of the industrial equipment to be detected, wherein the data-driven development method based on artificial intelligence (AI) has a high development efficiency, thereby solving the problem of low development efficiency of traditional expert systems. In addition, in the process of obtaining the health detection model of the catenary, the unsupervised learning method and the semi-supervised learning method are combined to improve the confidence of the label, thereby reducing the cost of health checks on industrial equipment.

[0097] Further, based on the above embodiment, the acquisition module 201 is specifically used to: acquire the vibration data corresponding to the industrial equipment to be detected; perform data cleaning on the vibration data to obtain time domain data corresponding to the industrial equipment to be detected; perform Fourier transform on the time domain data to obtain frequency domain data and time-frequency domain data corresponding to the industrial equipment to be detected; determine the first feature data based on the time domain data, the frequency domain data and the time-frequency domain data.

[0098] In the above scheme, in the process of extracting the corresponding time domain data based on the vibration data corresponding to the industrial equipment to be detected, the high-order eigenvalues ​​of the waveform or spectrum can be used, thereby avoiding the problem of feature redundancy and large amount of calculation caused by using the original waveform or spectrum with too large dimensions, thereby improving the computational efficiency and effect of the algorithm.

[0099] Furthermore, based on the above embodiment, the acquisition module 201 is also used to: respectively extract the time domain features corresponding to the time domain data, the frequency domain features corresponding to the frequency domain data, and the time-frequency domain features corresponding to the time-frequency domain data; and fuse the time domain features, the frequency domain features, and the time-frequency domain features to obtain the first feature data.

[0100] In the above scheme, in the process of extracting the corresponding time domain data based on the vibration data corresponding to the industrial equipment to be detected, the multi-dimensional statistical features of the time domain signal, the frequency domain signal and the time-frequency domain signal can be integrated. Compared with the scheme of using only the time domain or frequency domain signal features, it provides a more comprehensive data consideration and thus can improve the accuracy of the health detection model. In addition, by extracting high-dimensional features of data of three signal types, namely, the time domain, the frequency domain and the time-frequency domain, the information contained in the feature dimension can be made more comprehensive.

[0101] Furthermore, based on the above embodiment, the health detection device 200 for industrial equipment also includes: a training module, which is used to train the initial model using the following steps to obtain the health detection model: obtain second feature data corresponding to healthy industrial equipment; train the initial model using the second feature data to obtain an intermediate model; obtain third feature data and label data corresponding to sample industrial equipment; wherein the sample industrial equipment is an industrial equipment with an unknown health status; train the intermediate model using the third feature data and the label data to obtain the health detection model.

[0102] In the above scheme, in the process of training the initial model, first, an unsupervised learning method can be used to train the initial model using the second feature data corresponding to the health industrial equipment. Then, a semi-supervised learning method can be used to use the initial model to predict the results of the unlabeled data and label the unlabeled data according to the posterior probability value. This can effectively reduce the labeling cost and provide more data support to help the model capture richer data distribution characteristics, thereby improving the generalization ability of the model.

[0103] Furthermore, based on the above embodiments, the training module is specifically used to: obtain vibration sample data corresponding to the sample industrial equipment; determine the third feature data based on the vibration sample data; input the third feature data into the initial model to obtain a classification result corresponding to the third feature data; and determine the label data based on the classification result.

[0104] In the above scheme, semi-supervised learning can be used for model training, which combines the advantages of supervised learning and unsupervised learning, and effectively reduces the labeling cost of faulty equipment data.

[0105] Furthermore, based on the above embodiment, the training module is also used to: convert the classification result into a posterior probability value using a data fitting method; label the third feature data according to the posterior probability value and the classification threshold to obtain the label data corresponding to the third feature data.

[0106] In the above scheme, the classification results can be post-processed using a data fitting method and converted into posterior probability values, thereby providing confidence for the data labels.

[0107] Please refer to Figure 3 , Figure 3 This is a structural block diagram of an electronic device provided in an embodiment of the present application. The electronic device 300 includes: at least one processor 301, at least one communication interface 302, at least one memory 303 and at least one communication bus 304. Among them, the communication bus 304 is used to realize the direct connection and communication of these components, the communication interface 302 is used to communicate signaling or data with other node devices, and the memory 303 stores machine-readable instructions executable by the processor 301. When the electronic device 300 is running, the processor 301 communicates with the memory 303 through the communication bus 304, and the machine-readable instructions are called by the processor 301 to execute the above-mentioned health detection method of industrial equipment.

[0108] For example, the processor 301 of the embodiment of the present application reads a computer program from the memory 303 through the communication bus 304 and executes the computer program to implement the following method: obtaining first feature data corresponding to the industrial equipment to be detected; wherein the first feature data is determined based on the vibration data corresponding to the industrial equipment to be detected; inputting the first feature data into a health detection model to obtain a health detection result output by the health detection model; wherein the health detection model is obtained by performing unsupervised learning and semi-supervised learning training on an initial model, and the initial model is a first-class support vector machine.

[0109] Among them, the processor 301 includes one or more, which can be an integrated circuit chip with signal processing capabilities. The above-mentioned processor 301 can be a general-purpose processor, including a central processing unit (CPU), a micro control unit (MCU), a network processor (NP) or other conventional processors; it can also be a dedicated processor, including a neural network processor (NPU), a graphics processor (GPU), a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. In addition, when there are multiple processors 301, some of them can be general-purpose processors and the other part can be dedicated processors.

[0110] The memory 303 includes one or more, which may be, but not limited to, a random access memory (RAM), a read only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), etc.

[0111] Understandably, Figure 3 The structure shown is for illustration only. The electronic device 300 may also include Figure 3 More or fewer components as shown, or with Figure 3 Different configurations are shown. Figure 3Each component shown in can be implemented by hardware, software or a combination thereof. In the embodiment of the present application, the electronic device 300 can be, but is not limited to, a physical device such as a desktop computer, a laptop computer, a smart phone, a smart wearable device, a vehicle-mounted device, etc., and can also be a virtual device such as a virtual machine. In addition, the electronic device 300 is not necessarily a single device, but can also be a combination of multiple devices, such as a server cluster, etc.

[0112] The embodiment of the present application also provides a computer program product, including a computer program stored on a computer-readable storage medium, the computer program including computer program instructions, when the computer program instructions are executed by the computer, the computer can execute the steps of the health detection method of industrial equipment in the above embodiment, for example including: Step S101: Obtain the first feature data corresponding to the industrial equipment to be detected; wherein the first feature data is determined based on the vibration data corresponding to the industrial equipment to be detected. Step S102: Input the first feature data into the health detection model to obtain the health detection result output by the health detection model; wherein the health detection model is obtained by performing unsupervised learning and semi-supervised learning training on the initial model, and the initial model is a class of support vector machines.

[0113] An embodiment of the present application further provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer program instructions. When the computer program instructions are executed by a computer, the computer executes the health detection method for industrial equipment described in the aforementioned method embodiment.

[0114] In the embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interfaces, and the indirect coupling or communication connection of the devices or units can be electrical, mechanical or other forms.

[0115] In addition, the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0116] Furthermore, the functional modules in the various embodiments of the present application may be integrated together to form an independent part, or each module may exist separately, or two or more modules may be integrated to form an independent part.

[0117] It should be noted that if the function is implemented in the form of a software function module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can essentially be embodied in the form of a software product, or the part that contributes to the prior art or the part of the technical solution. The computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM) random access memory (RAM), disk or optical disk, and other media that can store program codes.

[0118] In this document, relational terms such as first and second, etc. are used merely to distinguish one entity or operation from another entity or operation, but do not necessarily require or imply any such actual relationship or order between these entities or operations.

[0119] The above description is only an embodiment of the present application and is not intended to limit the protection scope of the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A health detection method for industrial equipment, characterized in that: include: Acquire first characteristic data corresponding to the industrial equipment to be detected; wherein the first characteristic data is determined based on vibration data corresponding to the industrial equipment to be detected; The first feature data is input into a health detection model to obtain a health detection result output by the health detection model; wherein the health detection model is obtained by performing unsupervised learning and semi-supervised learning training on an initial model, and the initial model is a first-class support vector machine.

2. The method for detecting the health of industrial equipment according to claim 1, characterized in that: The obtaining of first characteristic data corresponding to the industrial equipment to be detected includes: Acquire the vibration data corresponding to the industrial equipment to be detected; Performing data cleaning on the vibration data to obtain time domain data corresponding to the industrial equipment to be detected; Performing Fourier transform on the time domain data to obtain frequency domain data and time-frequency domain data corresponding to the industrial equipment to be detected; The first feature data is determined according to the time domain data, the frequency domain data and the time-frequency domain data.

3. The method for detecting the health of industrial equipment according to claim 2, characterized in that: The determining the first feature data according to the time domain data, the frequency domain data and the time-frequency domain data comprises: Respectively extracting the time domain features corresponding to the time domain data, the frequency domain features corresponding to the frequency domain data, and the time-frequency domain features corresponding to the time-frequency domain data; The time domain features, the frequency domain features and the time-frequency domain features are fused to obtain the first feature data.

4. The method for detecting the health of industrial equipment according to any one of claims 1 to 3, characterized in that: Also includes: The initial model is trained using the following steps to obtain the health detection model: Acquire second characteristic data corresponding to the health industrial equipment; Using the second feature data to train the initial model to obtain an intermediate model; Acquire third characteristic data and label data corresponding to sample industrial equipment; wherein the sample industrial equipment is industrial equipment whose health status is unknown; The intermediate model is trained using the third feature data and the label data to obtain the health detection model.

5. The method for detecting the health of industrial equipment according to claim 4, characterized in that: The obtaining of the third characteristic data and label data corresponding to the sample industrial equipment includes: Acquire vibration sample data corresponding to the sample industrial equipment; Determine the third characteristic data according to the vibration sample data; Inputting the third characteristic data into the initial model to obtain a classification result corresponding to the third characteristic data; The label data is determined according to the classification result.

6. The method for detecting the health of industrial equipment according to claim 5, characterized in that: The determining the label data according to the classification result includes: Converting the classification results into posterior probability values ​​using a data fitting method; The third feature data is labeled according to the posterior probability value and the classification threshold to obtain the label data corresponding to the third feature data.

7. A health detection device for industrial equipment, characterized in that: include: An acquisition module, used to acquire first characteristic data corresponding to the industrial equipment to be detected; wherein the first characteristic data is determined based on vibration data corresponding to the industrial equipment to be detected; An input module is used to input the first feature data into a health detection model to obtain a health detection result output by the health detection model; wherein the health detection model is obtained by performing unsupervised learning and semi-supervised learning training on an initial model, and the initial model is a first-class support vector machine.

8. A computer program product, characterized in that The method comprises computer program instructions, and when the computer program instructions are read and executed by a processor, the method for detecting the health of industrial equipment according to any one of claims 1 to 6 is executed.

9. An electronic device, characterized in that: include: processor, memory, and bus; The processor and the memory communicate with each other via the bus; The memory stores computer program instructions that can be executed by the processor, and the processor calls the computer program instructions to execute the health detection method for industrial equipment according to any one of claims 1 to 6.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer program instructions, and when the computer program instructions are executed by a computer, the computer executes the health detection method for industrial equipment according to any one of claims 1 to 6.

Images