Equipment fault early warning method and system based on AI algorithm, and storage medium

Abstract

The invention belongs to the technical field of data processing, and particularly relates to an equipment fault early warning method and system based on an AI algorithm and a storage medium, and the method comprises the steps: collecting multi-source heterogeneous data needing to be collected, extracting normal data in a normal state and abnormal data in an abnormal state from the multi-source heterogeneous data, and storing the normal data and the abnormal data in the normal state; generating a first model through machine learning based on the normal data, executing transition learning on the first model to generate a second model, inputting the real-time data into the first model for prediction to obtain a corresponding prediction value, calculating a deviation degree between the prediction value and the real-time data, and when the deviation degree is greater than a first threshold value, judging that the real-time data is abnormal data, and inputting the real-time data into the second model to obtain an exception type, obtaining standard exception data corresponding to the exception type, calculating the similarity between the standard exception data and the real-time data, and judging whether there is a symptom of occurrence of the corresponding exception type based on the similarity. According to the invention, the accuracy of fault identification can be improved.

Description

Technical Field

[0001] The present invention belongs to the field of data processing technology, and specifically relates to an equipment failure early warning method, system and storage medium based on an AI algorithm. Background Art

[0002] With the rapid development of urban rail transit, the traditional operation and maintenance model can no longer meet the growing operational needs. Traditional fault diagnosis and early warning mainly rely on manual inspection and experience judgment, which has problems such as low efficiency and insufficient accuracy. The existing technology uses artificial intelligence technology to perform fault early warning on the operating status of the equipment, predict the occurrence of faults in advance, reduce the failure rate of the system, and improve the safety and reliability of the system. A similar prior art Chinese patent application with publication number CN117591970A provides a fault early warning method and system for fully automatic assembly equipment, the method comprising: based on the sensor group of the fully automatic assembly equipment, real-time collection of equipment real-time operation data of the fully automatic assembly equipment; the equipment real-time operation data includes multiple equipment operation parameters; according to the equipment real-time operation data and multiple historical equipment operation data, determine the equipment operation data sequence of the fully automatic assembly equipment; input the equipment operation data sequence into the fault prediction model to determine the fault prediction results of the fully automatic assembly equipment for each preset equipment fault type; the fault prediction model includes an LSTM model module; based on each fault prediction result, perform a fault early warning operation. However, this method does not consider the situation that the training model has low accuracy due to insufficient abnormal training data when training the model. Similar prior art also includes a Chinese patent application with publication number CN114065510A, which discloses a fault warning method, device and electronic equipment for a cooling tower fan. The method obtains historical operating data of the cooling tower fan at different measuring points under healthy conditions; based on the historical operating data, the base regressor is trained using a gradient boosting algorithm to obtain a prediction model; the baseline value of the target operating parameter of the cooling tower fan is predicted using the prediction model; and the cooling tower fan is given a fault warning through the baseline value of the target operating parameter and the actual value of the target operating parameter. However, this method also does not consider the situation that the training model has low accuracy due to insufficient abnormal training data when training the prediction model. Therefore, the present invention provides an equipment fault warning method, system and storage medium based on an AI algorithm. Summary of the invention

[0003] In order to perform preventive maintenance on rail transit equipment, it is necessary to obtain enough data to train the model. Before training the model, the collected data needs to be extracted into normal data and abnormal data. Usually the amount of collected data is very large, and it takes a lot of time to extract normal data and abnormal data. Therefore, it is necessary to quickly and accurately extract normal data and abnormal data from the collected data. Secondly, when training the second model, that is, the abnormality recognition model, the frequency of abnormalities in real life is relatively low, so the amount of abnormal data may be insufficient. If the amount of abnormal data is insufficient, the recognition accuracy of the trained model may not be accurate enough. Therefore, it is necessary to generate enough abnormal data to train the abnormality recognition model.

[0004] In order to achieve the above-mentioned purpose of the invention, the present invention provides an equipment failure early warning method based on AI algorithm as described below, which is mainly implemented by executing the following steps: Collecting multi-source heterogeneous data that needs to be collected, the multi-source heterogeneous data includes multiple characteristic values, including basic data of the target device, whole machine status data, component status data, operation data and log data of software controlling the target device, extracting normal data of the target device in a normal operating state and abnormal data when an abnormality occurs from the multi-source heterogeneous data, and collecting the multi-source heterogeneous data that needs to be collected in real time based on a predetermined collection cycle, and the multi-source heterogeneous data collected in real time is called real-time data; Generate a first model through machine learning based on the normal data, the first model can predict the correlation between each data in the normal data under normal conditions, use the abnormal data and the pre-defined abnormal type corresponding to the abnormal data as learning data, perform transition learning on the first model to generate a second model, the second model can identify the abnormal type; Input the real-time data into the first model for prediction to obtain a corresponding prediction value, calculate the degree of deviation between the prediction value and the real-time data, and when the degree of deviation is less than a preset first threshold, judge the real-time data as normal data; when the degree of deviation is greater than the preset first threshold, judge the real-time data as abnormal data, and input the real-time data into the second model to obtain the abnormality type; Obtain standard abnormal data corresponding to the abnormal type, calculate the similarity between the standard abnormal data and the real-time data, and when the similarity is greater than a preset second threshold, judge that the real-time data has signs of the occurrence of the corresponding abnormal type, and issue an early warning based on the abnormal type; when the similarity is less than or equal to the second threshold, define a new abnormal type, issue an early warning based on the new abnormal type, and add the new abnormal type and the corresponding real-time data to the learning data for training the second model to retrain and update the second model.

[0005] As a preferred technical solution of the present invention, extracting normal data in a normal state and abnormal data when an abnormality occurs from the multi-source heterogeneous data includes the following steps: All the collected multi-source heterogeneous data are preprocessed, and the processed multi-source heterogeneous data are called first data. According to the acquisition time of the first data, all the first data are divided into multiple second data at predetermined time intervals, and each second data contains multiple first data. The average value and standard deviation of each characteristic value in the second data are calculated, and each characteristic value is standardized based on the average value and the standard deviation to obtain a standardized value of each characteristic value. It is judged whether the standardized value meets the predetermined condition. If the standardized value meets the predetermined condition, a second judgment is continued. If the standardized value does not meet the predetermined condition, it is judged that the first data corresponding to the standardized value is normal data.

[0006] As a preferred technical solution of the present invention, the predetermined condition is the first condition or the second condition, the first condition is that there is at least one standardized value corresponding to the characteristic value in the first data that is greater than three times the standard deviation, and the second condition is that the number of standardized values ​​in the first data that are greater than two times the standard deviation is greater than a predetermined number. If the standardized value satisfies any one of the first condition and the second condition, it is judged that the standardized value satisfies the predetermined condition.

[0007] As a preferred technical solution of the present invention, the second judgment includes the following steps: Calculate the correlation coefficient of each eigenvalue, combine the correlation coefficients of all eigenvalues ​​in the second data into a correlation coefficient matrix, and then obtain the inverse matrix of the correlation coefficient matrix by calculation, calculate the abnormal value corresponding to the first data based on the inverse matrix and the standardized value, and when the abnormal value is greater than a preset third threshold, judge the first data as abnormal data, otherwise judge the first data as normal data.

[0008] As a preferred technical solution of the present invention, calculating the abnormal value corresponding to the first data based on the inverse matrix and the standardized value includes the following steps: The abnormal value Ab is calculated based on a first formula, where the first formula is: , in, are the standardized values ​​corresponding to the n eigenvalues ​​in the first data, and CM is the inverse matrix.

[0009] As a preferred technical solution of the present invention, executing transition learning to generate a second model includes the following steps: Based on the learning data, transition learning is performed on the first model to generate the second model, part of the data is extracted from the abnormal data as test data and input into the second model, the detection performance of the second model is evaluated and an evaluation value is obtained, if the evaluation value is less than a preset third threshold, a simulation model of the target device is created, a variety of abnormal scenarios are set based on the simulation model to make the simulation model abnormal, and experimental data under abnormal conditions are generated, the experimental data are collected, and simulation data of the abnormal conditions under real conditions is generated based on the experimental data, the simulation data is added to the learning data, and this step is repeated to retrain the second model to improve the detection performance of the second model until the evaluation value is greater than or equal to the third threshold.

[0010] As a preferred technical solution of the present invention, calculating the degree of deviation between the predicted value and the real-time data comprises the following steps: The deviation degree DL is calculated using a second formula, wherein the second formula is: , Wherein, n is the total number of eigenvalues ​​in the real-time data, is the ith actual value in the real-time data, is the predicted value corresponding to the i-th actual value.

[0011] As a preferred technical solution of the present invention, each of the characteristic values ​​is standardized based on the mean value and the standard deviation, comprising the following steps: The third formula is used to obtain the standardized value S of each eigenvalue. The third formula is: , Among them, X is a certain eigenvalue, AV is the average value of the eigenvalue, and D is the standard deviation.

[0012] The present invention also provides an equipment failure early warning system based on an AI algorithm, comprising the following modules: A data collection unit is used to collect multi-source heterogeneous data that needs to be collected, wherein the multi-source heterogeneous data includes multiple characteristic values, including basic data of a target device, whole device status data, component status data, operation data, and log data of software that controls the target device, extract normal data in a normal state and abnormal data when an abnormality occurs from the multi-source heterogeneous data, and further collect the multi-source heterogeneous data that needs to be collected in real time based on a predetermined collection cycle, and the multi-source heterogeneous data collected in real time is referred to as real-time data; a model generating unit, configured to generate a first model through machine learning based on the normal data, wherein the first model can predict the correlation between each data in the normal data under a normal state, and use the abnormal data and the predefined abnormal type corresponding to the abnormal data as learning data, perform transition learning on the first model to generate a second model, wherein the second model can identify the abnormal type; an abnormality judgment unit, used for inputting the real-time data into the first model for prediction to obtain a corresponding predicted value, calculating the degree of deviation between the predicted value and the real-time data, and judging the real-time data as normal data when the degree of deviation is less than a preset first threshold, and judging the real-time data as abnormal data when the degree of deviation is greater than the preset first threshold, and inputting the real-time data into the second model to obtain the abnormality type; The type determination unit is used to obtain standard abnormal data corresponding to the abnormal type, calculate the similarity between the standard abnormal data and the real-time data, and determine whether the similarity is greater than or equal to a preset second threshold. If the similarity is greater than the second threshold, it is determined that the real-time data has signs of the occurrence of the corresponding abnormal type, and an early warning prompt is issued based on the abnormal type.

[0013] The present invention also provides a storage medium, which stores program instructions, wherein when the program instructions are executed, the device where the storage medium is located is controlled to execute any one of the above-mentioned AI algorithm-based equipment failure warning methods.

[0014] Compared with the prior art, the beneficial effects of the present invention are at least as follows: When extracting normal data and abnormal data from multi-source heterogeneous data, the present invention accurately extracts normal data and abnormal data through two calculations, reduces the calculation load under the premise of ensuring the extraction accuracy, and then generates a first model based on the normal data, the first model can generate a predicted value corresponding to the real-time data of the device under normal circumstances, and a second model based on the abnormal data. In the process of generating the second model, considering that the recognition accuracy of the second model will be affected when there are not enough abnormal data, more simulated abnormal data are generated to ensure the recognition accuracy of the second model, improve the recognition performance and recognition accuracy of the second model, and accurately judge whether the target device has signs of abnormality based on the first model and the second model, so as to help relevant personnel take accurate preventive maintenance measures. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 A flowchart of the steps of an equipment failure early warning method based on an AI algorithm of the present invention; Figure 2 This is a structural diagram of the equipment failure early warning system based on AI algorithm of the present invention. DETAILED DESCRIPTION

[0016] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0017] It is understood that the terms "first", "second", etc. used in this application may be used herein to describe various elements, but unless otherwise specified, these elements are not limited by these terms. These terms are only used to distinguish a first element from another element. For example, without departing from the scope of this application, a first xx script may be referred to as a second xx script, and similarly, a second xx script may be referred to as a first xx script.

[0018] The present invention provides Figure 1 The device fault early warning method based on AI algorithm is mainly implemented by executing the following steps: Step S1, collect multi-source heterogeneous data that need to be collected, the multi-source heterogeneous data includes multiple characteristic values, including basic data of the target device, whole machine status data, component status data, operation data and log data of the software that controls the target device, extract normal data of the target device under normal operating state and abnormal data when an abnormality occurs from the multi-source heterogeneous data, and collect the multi-source heterogeneous data that needs to be collected in real time based on a predetermined collection period, and the multi-source heterogeneous data collected in real time is called real-time data.

[0019] Specifically, in order to provide early warning of equipment failure, it is necessary to collect various data related to the target equipment. These data are collectively referred to as multi-source heterogeneous data. Multi-source heterogeneous data includes multiple characteristic values. Taking the target equipment as a shielding door as an example, the basic data includes the opening and closing speed, voltage resistance level and sealing performance of the shielding door. The whole machine status data includes the operating mode, working voltage and current, temperature, operating time and performance indicators of the shielding door. The component status data includes the operating parameters, performance indicators, wear and tear, fault history, electrical parameters, vibration data, sound data, temperature data and replacement of each component related to the subway shielding door. The operating data includes the number of times the shielding door is opened and closed, the opening and closing time, the operating speed and power consumption, etc. The log data of the software that controls the target device includes system status, exceptions and error information, etc. After obtaining these multi-source heterogeneous data, in order to train the prediction model, it is necessary to extract normal data under normal conditions and abnormal data under abnormal conditions. Through accurate and fast extraction methods, more accurate training data is provided for subsequent training models. The specific extraction method will be explained in detail later. Real-time data is also collected according to the predetermined collection cycle, and the real-time data is input into the subsequent trained model to predict whether the equipment has signs of failure.

[0020] Step S2: Generate a first model through machine learning based on normal data. The first model can predict the correlation between each data in the normal data under normal conditions. The abnormal data and the abnormal types corresponding to the pre-defined abnormal data are used as learning data. Transition learning is performed on the first model to generate a second model. The second model can identify the abnormal type.

[0021] Specifically, in order to obtain a more accurate prediction model, a first model is first generated through machine learning based on normal data. The first model can analyze the correlation between various data when the equipment is operating normally. That is to say, the first model can output the predicted value of each data under the normal operating state of the equipment based on the input data. Then, on the basis of the normal model, the extracted abnormal data is used for over-learning to generate a second model. When training the second model, the corresponding abnormal type is defined based on the abnormal data, and training is performed based on the abnormal data and the abnormal type so that the second model can identify the abnormal type. When there are not enough abnormal data, in order to improve the recognition accuracy of the second model, more abnormal data are generated to train the second model, thereby generating a second model with higher recognition performance. The specific process of generating the second model will be explained in detail later.

[0022] Step S3: input the real-time data into the first model for prediction to obtain the corresponding predicted value, calculate the degree of deviation between the predicted value and the real-time data, and when the degree of deviation is less than the preset first threshold, judge the real-time data as normal data; when the degree of deviation is greater than the preset first threshold, judge the real-time data as abnormal data, and input the real-time data into the second model to obtain the abnormal type.

[0023] Specifically, in order to determine whether there are any signs of an abnormality at present, the collected real-time data is first input into the first model to obtain the predicted value corresponding to the real-time data, and then the degree of deviation between the two is calculated based on the predicted value and the real-time data. The specific calculation formula for the degree of deviation is calculated using the subsequent second formula. When the degree of deviation is less than the preset first threshold, it means that the real-time data and the data that should be generated when the equipment is operating normally are not much different, so it is judged that there is no sign of failure of the equipment at this time. When the degree of deviation is greater than the first threshold, it means that the real-time data and the data that should be generated when the equipment is operating normally are significantly different, and it is judged that the equipment may have signs of failure at this time. In order to further determine the signs of failure, the real-time data is input into the pre-trained second model for further judgment and obtain the corresponding abnormality type.

[0024] Step S4, obtaining standard abnormal data corresponding to the abnormal type, calculating the similarity between the standard abnormal data and the real-time data, and when the similarity is greater than a preset second threshold, judging that the real-time data has signs of the corresponding abnormal type, and giving an early warning based on the abnormal type; when the similarity is less than or equal to the second threshold, defining a new abnormal type, giving an early warning based on the new abnormal type, and adding the new abnormal type and the corresponding real-time data to the learning data for training the second model to retrain and update the second model.

[0025] Specifically, after obtaining the abnormal type, the standard abnormal data corresponding to the abnormal type is also obtained. The standard abnormal data refers to the data generated by the target device when the corresponding abnormal type occurred in history. The similarity between the standard abnormal data and the real-time data is calculated. When the similarity is greater than the preset second threshold, it means that the current real-time data and the standard abnormal data have a large similarity, which can further indicate that the possibility of the abnormal situation corresponding to the abnormal type occurring is greater. An early warning is given based on the abnormal type to remind relevant staff that the abnormal situation of the above-mentioned abnormal type may be about to occur. The relevant staff can take precise preventive maintenance measures based on the prompted abnormal type. When the similarity is less than or equal to the second threshold, it means that the abnormal type currently predicted by the second model may be a new abnormal type that has not occurred before. Therefore, by defining a new abnormal type, and adding the new abnormal type and the corresponding real-time data to the learning data for training the second model, the second model is updated. Afterwards, an early warning is given based on the new abnormal type to remind relevant staff to take corresponding preventive measures.

[0026] When extracting normal data and abnormal data from multi-source heterogeneous data, the present invention accurately extracts normal data and abnormal data through two calculations, reduces the calculation load under the premise of ensuring the extraction accuracy, and then generates a first model based on the normal data, the first model can generate a predicted value corresponding to the real-time data of the device under normal circumstances, and a second model based on the abnormal data. In the process of generating the second model, considering that the recognition accuracy of the second model will be affected when there are not enough abnormal data, more simulated abnormal data are generated to ensure the recognition accuracy of the second model, improve the recognition performance and recognition accuracy of the second model, and accurately judge whether the target device has signs of abnormality based on the first model and the second model, so as to help relevant personnel take accurate preventive maintenance measures.

[0027] Furthermore, extracting normal data in a normal state and abnormal data when an abnormality occurs from multi-source heterogeneous data includes the following steps: All collected multi-source heterogeneous data are preprocessed, and the processed multi-source heterogeneous data are called first data. According to the acquisition time of the first data, all the first data are divided into multiple second data at predetermined time intervals, and each second data contains multiple first data. The average value and standard deviation of each eigenvalue in the second data are calculated, and each eigenvalue is standardized based on the average value and the standard deviation to obtain a standardized value of each eigenvalue. It is judged whether the standardized value meets the predetermined condition. If the standardized value meets the predetermined condition, a second judgment is performed. If the standardized value does not meet the predetermined condition, it is judged that the first data corresponding to the standardized value is normal data.

[0028] Specifically, in order to extract normal data and abnormal data from multi-source heterogeneous data more accurately and quickly, the multi-source heterogeneous data is first preprocessed. Preprocessing refers to cleaning all the collected multi-source heterogeneous data to remove noise and outliers. The processed multi-source heterogeneous data is called first data. Then, based on the acquisition time of the first data and the predetermined time interval, the first data is segmented into multiple second data. For example, the first data is acquired every 10 minutes, and a total of data within the past 1000 minutes is collected. Then, there are 100 first data. Taking the predetermined time interval of 100 minutes as an example, the first data is segmented into 10 parts, each of which includes 10 first data. The second data is A={A1, A2, A3, ...A10}, A1, A2...A10 are all first data, and each first data includes multiple Eigenvalues ​​A1={a1, a1, ...an} are obtained. Based on the second data, the average value AV and the standard deviation D of each eigenvalue in the second data are calculated. Then, each eigenvalue is standardized based on the average value and the standard deviation to obtain a standardized value of each eigenvalue. The standardization method will be explained in detail later. Then, it is determined whether the standardized value meets a predetermined condition. If the predetermined condition is met, it indicates that the first data may be abnormal data. In order to more accurately determine whether the first data is normal data, a second judgment is made. The specific method of the second judgment will be explained in detail later. If the standardized value does not meet the predetermined condition, it indicates that the first data is normal data. The first data judged to be normal data does not need to be judged for the second time. The above method can reduce the amount of calculation and speed up the judgment.

[0029] Furthermore, the predetermined condition is the first condition or the second condition, the first condition is that there is at least one characteristic value in the first data whose corresponding standardized value is greater than three times the standard deviation, and the second condition is that the number of standardized values ​​in the first data greater than two times the standard deviation is greater than a predetermined number. If the standardized value satisfies any one of the first condition and the second condition, it is judged that the standardized value satisfies the predetermined condition.

[0030] Specifically, by using the standardized data for judgment, it is possible to quantitatively and easily obtain the number of abnormal characteristic values ​​compared with the standard data. When the first condition is met, it means that there are data in the first data with a large difference between one or more characteristic values ​​and the standardized value. The existence of data with a large difference indicates that the possibility of an abnormality is higher, so the corresponding first data is judged as abnormal data. When the second condition is met, it means that there are a large number of abnormal characteristic values ​​in the first data. Although the difference is relatively small, the possibility of judging the existence of an abnormality is also high due to the large number. Therefore, when the standard deviation meets any one of the first and second conditions, it is judged that the standardized value meets the predetermined condition, and the corresponding first data is judged as abnormal data. Otherwise, the corresponding first data is judged as normal data.

[0031] Further, a second judgment is performed, including the following steps: Calculate the correlation coefficient of each eigenvalue, combine the correlation coefficients of all eigenvalues ​​in the second data into a correlation coefficient matrix, and then obtain the inverse matrix of the correlation coefficient matrix by calculation. Calculate the abnormal value corresponding to the first data based on the inverse matrix and the standardized value. When the abnormal value is greater than a preset third threshold, judge the first data as abnormal data; otherwise, judge the first data as normal data.

[0032] Specifically, by screening out data that may be abnormal in the first judgment, the second judgment can focus on the judgment of these data without calculating all the data, reducing the overall calculation load. At the same time, the results of the first calculation are used as a basis for more in-depth analysis and verification to improve the accuracy of the judgment. The elements of the inverse matrix reflect the correlation between the eigenvalues ​​in the original data. When calculating the abnormal value, the inverse matrix takes into account the correlation between the eigenvalues, so as to more accurately calculate the distance between the first data and other first data belonging to the same second data, and use the distance as an indicator to determine whether each first data is an abnormal data. The specific calculation formula for calculating the abnormal value of the corresponding first data based on the inverse matrix and the standardized value will be introduced later.

[0033] Further, calculating the abnormal value corresponding to the first data based on the inverse matrix and the standardized value includes the following steps: The abnormal value Ab is calculated based on the first formula, which is: , in, are the standardized values ​​corresponding to the n eigenvalues ​​in the first data, and CM is the inverse matrix.

[0034] Specifically, based on the above formula, the data is converted into standardized data by using the inverse matrix of the correlation coefficient matrix, which not only simplifies the complexity of the calculation but also improves the calculation efficiency.

[0035] Further, performing transition learning to generate a second model includes the following steps: Based on the learning data, transition learning is performed on the first model to generate a second model, part of the data is extracted from the abnormal data as test data and input into the second model, the detection performance of the second model is evaluated and an evaluation value is obtained, if the evaluation value is less than a preset third threshold, a simulation model of the target device is created, a variety of abnormal scenarios are set based on the simulation model to make the simulation model abnormal, and experimental data under abnormal conditions are generated, the experimental data is collected, and simulation data under abnormal conditions is generated based on the experimental data, the simulation data is added to the learning data, and this step is repeated to retrain the second model to improve the detection performance of the second model until the evaluation value is greater than or equal to the third threshold.

[0036] Specifically, in order to improve the recognition accuracy of the second model, after the second model is generated, some abnormal data is selected as test data to input into the second model, the detection performance of the second model is evaluated and an evaluation value is obtained. The evaluation value can be the detection accuracy of the second threshold. When the evaluation value is less than the preset third threshold, it means that the recognition accuracy of the currently trained second model is low, which may be caused by insufficient abnormal data. Therefore, a simulation model of the target device is created. The simulation model can simulate the working conditions of the target device under real conditions. Based on the simulation model, a variety of abnormal scenarios are set to make the simulation model abnormal, and experimental data under abnormal conditions are generated. Then, based on the experimental data, simulated data of abnormal conditions occurring under real conditions are generated, and the simulated data is added to the learning data for training the second model. Then, this step is repeated to retrain the second model to improve the detection performance of the second model until the evaluation value of the second model is greater than or equal to the third threshold. The detection performance and recognition accuracy of the second model are improved by the above method, so that the second model can more quickly identify the abnormal types occurring in real-time data, help relevant staff quickly understand abnormal conditions, and take relevant preventive measures.

[0037] Furthermore, the degree of deviation between the predicted value and the real-time data is calculated, including the following steps: The second formula is used to calculate the degree of deviation DL, which is: , Where n is the total number of eigenvalues ​​in the real-time data, is the ith actual value in the real-time data, is the predicted value corresponding to the i-th actual value.

[0038] Specifically, the degree of deviation between the predicted value and the real-time data is calculated based on the above formula, and a preliminary judgment can be made based on the degree of deviation whether there are potential signs of failure.

[0039] Furthermore, each eigenvalue is standardized based on the mean value and the standard deviation, including the following steps: The third formula is used to obtain the standardized value S of each eigenvalue. The third formula is: , Among them, X is a certain eigenvalue, AV is the average value of the eigenvalue, and D is the standard deviation.

[0040] According to another aspect of the embodiment of the present invention, Figure 2 As shown, an equipment failure early warning system based on an AI algorithm is also provided, including a data collection unit, a model generation unit, an abnormality judgment unit and a type determination unit, which are used to implement an equipment failure early warning method based on an AI algorithm as described above. The specific functions of each module are as follows: A data collection unit is used to collect multi-source heterogeneous data that needs to be collected, the multi-source heterogeneous data includes multiple characteristic values, including basic data of the target device, whole machine status data, component status data, operation data and log data of software controlling the target device, extract normal data in a normal state and abnormal data when an abnormality occurs from the multi-source heterogeneous data, and collect the multi-source heterogeneous data that needs to be collected in real time based on a predetermined collection cycle, and the multi-source heterogeneous data collected in real time is called real-time data; A model generation unit, configured to generate a first model through machine learning based on normal data, wherein the first model can predict the correlation between each data in the normal data under a normal state, and use the abnormal data and the abnormal type corresponding to the predefined abnormal data as learning data, perform transition learning on the first model to generate a second model, wherein the second model can identify the abnormal type; an abnormality judgment unit, used for inputting the real-time data into the first model for prediction to obtain a corresponding predicted value, calculating the degree of deviation between the predicted value and the real-time data, and judging the real-time data as normal data when the degree of deviation is less than a preset first threshold, and judging the real-time data as abnormal data when the degree of deviation is greater than the preset first threshold, and inputting the real-time data into the second model to obtain the abnormality type; The type determination unit is used to obtain standard abnormal data corresponding to the abnormal type, calculate the similarity between the standard abnormal data and the real-time data, and determine whether the similarity is greater than or equal to a preset second threshold. If the similarity is greater than the second threshold, it is determined that the real-time data has signs of the corresponding abnormal type, and an early warning prompt is issued based on the abnormal type.

[0041] According to another aspect of an embodiment of the present invention, a storage medium is also provided, wherein the storage medium stores program instructions, wherein when the program instructions are executed, the device where the storage medium is located is controlled to execute any one of the above-mentioned equipment fault warning methods based on AI algorithms.

[0042] In summary, the present invention provides an equipment fault early warning method, system and storage medium based on an AI algorithm, the method includes collecting multi-source heterogeneous data that needs to be collected, extracting normal data in a normal state and abnormal data when an abnormality occurs from the multi-source heterogeneous data, generating a first model through machine learning based on normal data, performing transition learning on the first model to generate a second model, inputting real-time data into the first model for prediction to obtain a corresponding predicted value, calculating the degree of deviation between the predicted value and the real-time data, judging the real-time data as abnormal data when the degree of deviation is greater than a first threshold, inputting the real-time data into the second model to obtain the abnormal type, obtaining the standard abnormal data corresponding to the abnormal type, calculating the similarity between the standard abnormal data and the real-time data, and judging whether there is a sign of the corresponding abnormal type based on the similarity. The present invention can improve the accuracy of fault identification.

[0043] It should be understood that, although each step in the flow chart of each embodiment of the present invention is shown in sequence according to the indication of the arrow, these steps are not necessarily performed in sequence according to the order indicated by the arrow. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be performed in other orders. Moreover, at least a portion of the steps in each embodiment may include a plurality of sub-steps or a plurality of stages, and these sub-steps or stages are not necessarily performed at the same time, but can be performed at different times, and the execution order of these sub-steps or stages is not necessarily performed in sequence, but can be performed in turn or alternately with at least a portion of other steps or sub-steps or stages of other steps.

[0044] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the above-mentioned program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0045] The technical features of the above-mentioned embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the technical features in the above-mentioned embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0046] The above-mentioned embodiments only express several implementation methods of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

[0047] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A device failure early warning method based on AI algorithm, characterized in that: The method comprises the following steps: Collecting multi-source heterogeneous data that needs to be collected, the multi-source heterogeneous data includes multiple characteristic values, including basic data of the target device, whole machine status data, component status data, operation data and log data of software controlling the target device, extracting normal data of the target device in a normal operating state and abnormal data when an abnormality occurs from the multi-source heterogeneous data, and collecting the multi-source heterogeneous data that needs to be collected in real time based on a predetermined collection cycle, and the multi-source heterogeneous data collected in real time is called real-time data; Generate a first model through machine learning based on the normal data, the first model can predict the correlation between each data in the normal data under normal conditions, use the abnormal data and the pre-defined abnormal type corresponding to the abnormal data as learning data, perform transition learning on the first model to generate a second model, the second model can identify the abnormal type; Input the real-time data into the first model for prediction to obtain a corresponding prediction value, calculate the degree of deviation between the prediction value and the real-time data, and when the degree of deviation is less than a preset first threshold, judge the real-time data as normal data; when the degree of deviation is greater than the preset first threshold, judge the real-time data as abnormal data, and input the real-time data into the second model to obtain the abnormality type; Obtain standard abnormal data corresponding to the abnormal type, calculate the similarity between the standard abnormal data and the real-time data, and when the similarity is greater than a preset second threshold, judge that the real-time data has signs of the occurrence of the corresponding abnormal type, and issue an early warning based on the abnormal type; when the similarity is less than or equal to the second threshold, define a new abnormal type, issue an early warning based on the new abnormal type, and add the new abnormal type and the corresponding real-time data to the learning data for training the second model to retrain and update the second model.

2. The method according to claim 1, characterized in that Extracting normal data in a normal state and abnormal data when an abnormality occurs from the multi-source heterogeneous data includes the following steps: All the collected multi-source heterogeneous data are preprocessed, and the processed multi-source heterogeneous data are called first data. According to the acquisition time of the first data, all the first data are divided into multiple second data at predetermined time intervals, and each second data contains multiple first data. The average value and standard deviation of each characteristic value in the second data are calculated, and each characteristic value is standardized based on the average value and the standard deviation to obtain a standardized value of each characteristic value. It is judged whether the standardized value meets the predetermined condition. If the standardized value meets the predetermined condition, a second judgment is continued. If the standardized value does not meet the predetermined condition, it is judged that the first data corresponding to the standardized value is normal data.

3. The method according to claim 2, characterized in that The predetermined condition is the first condition or the second condition, the first condition is that there is at least one standardized value corresponding to the characteristic value in the first data that is greater than three times the standard deviation, the second condition is that the number of standardized values ​​in the first data that are greater than two times the standard deviation is greater than a predetermined number, and the standardized value that meets any one of the first condition and the second condition is judged to meet the predetermined condition.

4. The method according to claim 2, characterized in that: The second judgment includes the following steps: Calculate the correlation coefficient of each eigenvalue, combine the correlation coefficients of all eigenvalues ​​in the second data into a correlation coefficient matrix, and then obtain the inverse matrix of the correlation coefficient matrix by calculation, calculate the abnormal value corresponding to the first data based on the inverse matrix and the standardized value, and when the abnormal value is greater than a preset third threshold, judge the first data as abnormal data, otherwise judge the first data as normal data.

5. The method according to claim 4, characterized in that Calculating an abnormal value corresponding to the first data based on the inverse matrix and the standardized value includes the following steps: The abnormal value Ab is calculated based on a first formula, where the first formula is: ， in, are the standardized values ​​corresponding to the n eigenvalues ​​in the first data, and CM is the inverse matrix.

6. The method according to claim 1, characterized in that Performing transition learning to generate a second model includes the following steps: Based on the learning data, transition learning is performed on the first model to generate the second model, part of the data is extracted from the abnormal data as test data and input into the second model, the detection performance of the second model is evaluated and an evaluation value is obtained, if the evaluation value is less than a preset third threshold, a simulation model of the target device is created, a variety of abnormal scenarios are set based on the simulation model to make the simulation model abnormal, and experimental data under abnormal conditions are generated, the experimental data are collected, and simulation data of the abnormal conditions under real conditions is generated based on the experimental data, the simulation data is added to the learning data, and this step is repeated to retrain the second model to improve the detection performance of the second model until the evaluation value is greater than or equal to the third threshold.

7. The method according to claim 1, characterized in that Calculating the degree of deviation between the predicted value and the real-time data comprises the following steps: The deviation degree DL is calculated using a second formula, wherein the second formula is: ， Wherein, n is the total number of eigenvalues ​​in the real-time data, is the ith actual value in the real-time data, is the predicted value corresponding to the i-th actual value.

8. The method according to claim 2, characterized in that: The method comprises the following steps: performing a standardization process on each of the characteristic values ​​based on the mean value and the standard deviation: The third formula is used to obtain the standardized value S of each eigenvalue. The third formula is: ， Among them, X is a certain eigenvalue, AV is the average value of the eigenvalue, and D is the standard deviation.

9. An equipment failure early warning system based on an AI algorithm, used to implement an equipment failure early warning method based on an AI algorithm as described in any one of claims 1 to 8, characterized in that: Includes the following modules: A data collection unit is used to collect multi-source heterogeneous data that needs to be collected, wherein the multi-source heterogeneous data includes multiple characteristic values, including basic data of a target device, whole device status data, component status data, operation data, and log data of software that controls the target device, extract normal data of the target device in a normal operating state and abnormal data when an abnormality occurs from the multi-source heterogeneous data, and further collect the multi-source heterogeneous data that needs to be collected in real time based on a predetermined collection period, and the multi-source heterogeneous data collected in real time is referred to as real-time data; a model generating unit, configured to generate a first model through machine learning based on the normal data, wherein the first model can predict the correlation between each data in the normal data under a normal state, and use the abnormal data and the predefined abnormal type corresponding to the abnormal data as learning data, perform transition learning on the first model to generate a second model, wherein the second model can identify the abnormal type; an abnormality judgment unit, used for inputting the real-time data into the first model for prediction to obtain a corresponding predicted value, calculating the degree of deviation between the predicted value and the real-time data, and judging the real-time data as normal data when the degree of deviation is less than a preset first threshold, and judging the real-time data as abnormal data when the degree of deviation is greater than the preset first threshold, and inputting the real-time data into the second model to obtain the abnormality type; A type determination unit is used to obtain standard abnormal data corresponding to the abnormal type, calculate the similarity between the standard abnormal data and the real-time data, and when the similarity is greater than a preset second threshold, determine that the real-time data has signs of the occurrence of the corresponding abnormal type, and issue an early warning based on the abnormal type; when the similarity is less than or equal to the second threshold, define a new abnormal type, issue an early warning based on the new abnormal type, and add the new abnormal type and the corresponding real-time data to a training data set for training the second model to train and update the second model.

10. A storage medium, characterized in that: The storage medium stores program instructions, wherein when the program instructions are executed, the device where the storage medium is located is controlled to execute the device fault warning method based on AI algorithm described in any one of claims 1 to 8.

Images