Equipment fault diagnosis model updating method and device, electronic equipment and storage medium
By collecting and labeling vibration data during the operation of the equipment, and judging and updating the equipment fault diagnosis model, the problem of model identification and update of new fault types is solved, and the effect of performance maintenance and resource conservation is achieved.
Patent Information
- Application Number
- CN202510148410.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-02-11
- Publication Date
- 2025-05-30
AI Technical Summary
When faced with new fault types, existing equipment fault diagnosis models cannot be identified and updated, resulting in a degradation of model performance and the update process consumes a lot of calculation costs and time.
By collecting vibration data during the operation of the equipment, if the initial model cannot predict the fault type, it is labeled as the target data and determine whether the preset update conditions are met. If so, the model is updated with the predictive performance of the initial model as a constraint.
The equipment fault diagnosis model is used to identify and update new fault types, avoid catastrophic forgetting, save calculation costs and time, and ensure the uninterrupted progress of fault diagnosis tasks.
Smart Images
Figure CN120067841A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of equipment monitoring, and in particular, to a method, a device, an electronic device, and a storage medium for updating a device fault diagnosis model. Background Art
[0002] In recent years, with the increasing complexity and intelligence of electromechanical equipment, in order to ensure the operation accuracy of electromechanical equipment, the demand for fault diagnosis of equipment has become stronger and stronger. Currently, most scholars explore fault diagnosis methods based on machine learning and deep learning and have achieved remarkable results.
[0003] However, most of the research focuses on static data sets. Electromechanical equipment may continuously feedback various different data, which may include new fault types not involved in previous model training. At this time, the model cannot identify the new fault types, so the model needs to be updated; when updating the model, directly using the data of the new fault type to update the model will cause catastrophic forgetting, and the performance of the model on the old data will decrease significantly; if the new and old data are mixed and the model is retrained, it will consume a large amount of computational cost and time cost. Summary of the Invention
[0004] In order to solve the above technical problems, the present invention provides a method for updating a device fault diagnosis model to solve the problems of the device fault diagnosis model.
[0005] In a first aspect, the present invention provides a method for updating a device fault diagnosis model, where the device is an electromechanical device, and the method includes:
[0006] During the operation of the device, vibration data of a preset component of the device is collected at a preset period, and the vibration data is input into an initial model;
[0007] If the initial model cannot predict the fault type of the vibration data, the vibration data is used as target data;
[0008] A fault label is set for the target data;
[0009] It is determined whether the initial model meets a preset update condition;
[0010] If so, the initial model is updated with the target data and the corresponding fault label under the constraint of the prediction performance of the initial model;
[0011] If not, the step of collecting vibration data of a preset component in the device at a preset period and inputting it into the initial model during the operation of the device is returned and executed.
[0012] In a second aspect, the present invention provides a device for updating a device fault diagnosis model, including:
[0013] A data acquisition and processing module, configured to collect vibration data of a preset component of the device at a preset period during the operation of the device, and input the vibration data into an initial model;
[0014] A target data determination module, configured to use the vibration data as target data if the initial model cannot predict the fault type of the vibration data;
[0015] A fault label setting module, configured to set a fault label for the target data;
[0016] An update judgment module, configured to judge whether the initial model meets a preset update condition; if so, execute the content of the model update module, if not, execute the content of the data acquisition and processing module;
[0017] A model update module, configured to update the initial model by using the target data and the corresponding fault labels with the prediction performance of the initial model as a constraint.
[0018] In a third aspect, the present invention provides an electronic device, which includes:
[0019] At least one processor; and
[0020] A memory communicatively connected to the at least one processor; wherein,
[0021] The memory stores a computer program executable by the at least one processor, and when the computer program is executed by the at least one processor, the at least one processor is enabled to execute the device fault diagnosis model update method according to the first aspect of the present invention.
[0022] In a fourth aspect, the present invention provides a computer-readable storage medium, which stores computer instructions for causing a processor to implement the device fault diagnosis model update method according to the first aspect of the present invention when executed.
[0023] For the device fault diagnosis model update method provided by the embodiments of the present invention, during the operation of the device, vibration data of a preset component of the device is collected at a preset period and input into an initial model. If the initial model cannot predict the fault type of the vibration data, the vibration data is used as target data; a fault label is set for the target data; it is judged whether the initial model meets a preset update condition; if so, the initial model is updated by using the target data and the corresponding fault labels with the prediction performance of the initial model as a constraint; if not, the step of collecting vibration data of a preset component in the device at a preset period and inputting it into the initial model during the operation of the device is returned for execution.
[0024] During the operation of the device, target data whose fault type cannot be recognized is collected and training data is obtained through marking. When the initial model meets the update condition, the initial model is updated with the new training data, which enables the initial model to learn to recognize new fault types and eliminates the need to retrain the model with old data, saving a large amount of computational and time costs. On the other hand, when updating the initial model, the prediction performance of the initial model is used as a constraint to avoid catastrophic forgetting caused by only updating the model with new data. In addition, the model does not need to be updated offline throughout the process, ensuring the uninterrupted progress of the fault diagnosis task.
[0025] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present invention, nor is it used to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. Brief Description of the Drawings
[0026] To more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the accompanying drawings required for the description of the embodiments. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.
[0027] Figure 1 It is a flowchart of a method for updating a device fault diagnosis model provided in Embodiment 1 of the present invention;
[0028] Figure 2 It is a flowchart of a method for updating a device fault diagnosis model provided in Embodiment 2 of the present invention;
[0029] Figure 3 It is a schematic diagram of the model incremental update training process provided in Embodiment 2 of the present invention;
[0030] Figure 4 It is a schematic structural diagram of a device fault diagnosis model update device provided in Embodiment 3 of the present invention;
[0031] Figure 5 It is a schematic structural diagram of an electronic device provided in Embodiment 4 of the present invention. Detailed Embodiments
[0032] To enable those skilled in the art to better understand the solution of the present invention, the following will clearly and completely describe the technical solution in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the scope of protection of the present invention.
[0033] Embodiment 1
[0034] Figure 1 FIG. is a flowchart of a method for updating an equipment fault diagnosis model provided in Embodiment 1 of the present invention. This embodiment is applicable to the situation of updating an equipment fault diagnosis model. The equipment is an electromechanical equipment, specifically including a motor, a driving wheel, etc. For example, the equipment can be an escalator, a vertical lift elevator, an X-ray security inspection machine, etc.
[0035] This method can be executed by an equipment fault diagnosis model updating device. The equipment fault diagnosis model updating device can be implemented in the form of hardware and / or software, and the equipment fault diagnosis model updating device can be configured in an electronic device.
[0036] As Figure 1 shown, the method for updating the equipment fault diagnosis model includes:
[0037] S101. During the operation of the equipment, collect vibration data of preset components of the equipment at a preset period, and input the vibration data into the initial model.
[0038] The preset components are usually key components during the operation of the equipment. Exemplarily, when the equipment is an escalator, the preset components can be a gearbox, a main driving wheel, and a step chain tensioning wheel. At least one vibration sensor is arranged at the positions where these key components are located, and relevant vibration data can be collected through the vibration sensor.
[0039] In an alternative embodiment, the training process of the initial model can be:
[0040] Obtain training data, where the training data includes vibration data and corresponding actual fault labels;
[0041] Input the vibration data into the constructed model to obtain predicted fault labels;
[0042] Determine the loss value according to the difference between the actual fault label and the predicted fault label;
[0043] Judge whether the loss value is greater than a preset loss threshold;
[0044] If so, adjust the parameters of the model based on the loss value;
[0045] Otherwise, return to the step of inputting the vibration data into the constructed model to obtain the predicted fault label.
[0046] It should also be noted that the preset period is a preset time period, for example, it can be 3 - 5 seconds. When collecting vibration data using the preset period, each vibration data is a collection of data points within a continuous time period.
[0047] S102. If the initial model cannot predict the fault type of the vibration data, then use the vibration data as the target data.
[0048] The initial model performs a fault diagnosis task on the input vibration data and outputs a fault diagnosis result. When the model can identify the sample, it outputs the specific fault category, otherwise the output result is an unknown category.
[0049] If the initial model cannot predict the fault type of the vibration data, it indicates that a new fault type may have occurred, that is, the current vibration data may be the data corresponding to the new fault type. Therefore, the vibration data is used as the target data. It should be noted here that the fault type can also include the case of no fault.
[0050] S103. Set a fault label for the target data.
[0051] Among them, the fault label is the fault type. At this time, the target data and the corresponding fault label can be saved in the database, and the fault label can be specifically set through the manual terminal.
[0052] S104. Determine whether the initial model reaches the preset update condition.
[0053] If so, execute S105; if not, return to execute S101.
[0054] Specifically, the preset update condition is that the prediction accuracy of the initial model is less than the preset accuracy threshold. Then, when the initial model performs the diagnosis task, the prediction accuracy of the initial model will be continuously counted. The calculation method of the prediction accuracy can be: counting the proportion of the data of the vibration data that the initial model can predict as the prediction accuracy. When the prediction accuracy is greater than or equal to the preset accuracy threshold, it means that the initial model can successfully predict the fault types of most vibration data or rarely has unknown fault types. When the prediction accuracy is less than the preset accuracy threshold, it means that the proportion of the number of vibration data that the initial model cannot predict also reaches a certain ratio. At this time, the initial model can be updated.
[0055] S105. Update the initial model using the target data and the corresponding fault label with the prediction performance of the initial model as a constraint.
[0056] Constrained by the prediction performance of the initial model, the updated initial model can retain the prediction performance of the initial model before the update. The specific constraint method can be: during the model update process, compare the output results of the updated model with those of the model before the update to prevent the updated model from forgetting old knowledge, that is, avoid the updated model losing the prediction performance of the model before the update.
[0057] Since the target data is data that cannot be recognized by the current initial model, updating the initial model with the target data and the corresponding fault labels can enable the updated model to have the performance of recognizing new fault types, achieving the purpose of updating the initial model. Moreover, updating the initial model only with the target data can reduce the data used for training the model and improve the training speed and efficiency.
[0058] It should be noted that the target data here is the new data saved since the last model update, and the data volume is usually small, without the need to consume a large amount of computing resources for training the model, nor a large amount of internal resources for storing data.
[0059] It should also be noted that the models mentioned in the present invention are all device fault diagnosis models.
[0060] The device fault diagnosis model update method provided by the embodiments of the present invention, during the operation of the device, collects the vibration data of the preset components of the device at a preset period and inputs it into the initial model. If the initial model cannot predict the fault type of the vibration data, the vibration data is used as the target data; set fault labels for the target data; determine whether the initial model reaches the preset update condition; if so, update the initial model with the target data and the corresponding fault labels constrained by the prediction performance of the initial model; if not, return to execute the step of collecting the vibration data of the preset components in the device at a preset period and inputting it into the initial model during the operation of the device.
[0061] Collecting target data whose fault types cannot be recognized during the operation of the device and obtaining training data through labeling. When the initial model reaches the update condition, updating the initial model with the new training data can enable the initial model to learn to recognize new fault types, and there is no need to retrain the model with old data, saving a large amount of computing costs and time costs; on the other hand, when updating the initial model, being constrained by the prediction performance of the initial model can avoid catastrophic forgetting caused by only using new data to update the model. In addition, the model does not need to be updated offline throughout the process, ensuring the uninterrupted progress of the fault diagnosis task.
[0062] Embodiment 2
[0063] Figure 2The flowchart of a method for updating an equipment fault diagnosis model provided in the second embodiment of the present invention. The embodiment of the present invention is optimized on the basis of the above-mentioned first embodiment. For example Figure 2 As shown, the method for updating the equipment fault diagnosis model includes:
[0064] S201. During the operation of the equipment, collect the vibration data of the preset components of the equipment at a preset period, and input the vibration data into the initial model.
[0065] In an optional embodiment, before inputting the vibration data into the initial model, it further includes: judging whether the vibration data is abnormal data according to the change amplitude of the vibration data; if so, deleting the vibration data; if not, then executing the step of inputting the vibration data into the initial model.
[0066] Due to improper installation of the sensor or sensor failure, data with abnormal fluctuations will be collected. Such data with abnormal fluctuations cannot correctly reflect the true state of the equipment, resulting in a significant increase in the false alarm rate of the fault diagnosis model. To avoid this situation, abnormal data is screened out before inputting the vibration data into the initial model.
[0067] On the one hand, judging whether the vibration data is abnormal data according to the change amplitude of the vibration data can be carried out by presetting a vibration threshold for screening. When there are data points in the vibration data whose vibration amplitude exceeds the vibration threshold, it can be considered that the vibration data is abnormal.
[0068] On the other hand, judging whether the target data is abnormal data according to the change amplitude of the target data includes: for each vibration data, perform non-overlapping sliding on the vibration data with a window of preset length to obtain a plurality of subsequences; calculate the average value, standard deviation and effective value of each subsequence to obtain three sets of statistical data; calculate the fluctuation coefficient for each of the three sets of statistical data, and obtain the preset fluctuation coefficient threshold corresponding to each set of statistical data; if there is a fluctuation coefficient greater than the preset fluctuation coefficient threshold, it is determined that the vibration data is abnormal data. Among them, the effective value is the root mean square value.
[0069] Assume that the number of data points of the vibration data is N, the data length of the subsequence is n, N is an integer multiple of n, and the vibration data is [x 1 , x 2 , x 3 , …, x N , the first subsequence is [x 1 , x 2 , x 3 , …, x n , the second subsequence is [x n+1 , x n+2 , x n+3 , …, x 2n, and so on. The three sets of statistical data obtained are the mean data set mean_list, the standard deviation data set std_list, and the root mean square data set rms_list. Assuming the number of subsequences is M, then the number of data in each data set is also M. The mean, standard deviation, and root mean square can reflect the data characteristics of the subsequences in different aspects. Therefore, these data characteristics are used to analyze the fluctuation of the entire target data.
[0070] Calculate the fluctuation coefficient (Fluctuation Coefficient, fc) for each of the three sets of statistical data. The fluctuation coefficient is a dimensionless index, a statistic used to measure the degree of data variation, and it is usually used to measure the fluctuation or variability of a set of data.
[0071] Specifically, the calculation method of the fluctuation coefficient is as follows: Subtract the minimum value from the maximum value in each set of statistical values, and then divide by the standard deviation of this set of statistical values. The smaller the value of the fluctuation coefficient, the more stable the data.
[0072] The formula for calculating the fluctuation coefficient is as follows:
[0073]
[0074] Set a threshold for each of the three fluctuation coefficients. When one of the fluctuation coefficients is greater than the corresponding threshold, the vibration data is considered abnormal data, and no fault diagnosis is performed and the vibration data is deleted; otherwise, it is normal data, and then it can be input into the fault diagnosis model (initial model) for judgment. Among them, after deleting the abnormal target data, it is also possible to return to execute S201 to continue collecting data.
[0075] Optionally, the parameters in the standard deviation calculation formula of the subsequence include the mean of the subsequence. Calculating the standard deviation of each subsequence includes: calculating the mean of the vibration data; using the mean of the vibration data as the mean in the standard deviation calculation formula of the subsequence; calculating the standard deviation of each subsequence based on the standard deviation calculation formula.
[0076] The formula for calculating the standard deviation σ is:
[0077]
[0078] Among them, n is the number of data in the subsequence, and x i is the i-th data in the subsequence, and μ is the mean.
[0079] The standard deviation of each subsequence is calculated using the average value of the entire vibration data, which can ensure that the calculation of the standard deviation is related to the fluctuation of the entire vibration data sequence, avoiding the situation where in a certain subsequence, the standard deviation is small but the deviation degree of this subsequence from the entire vibration data sequence is large. Especially when the window length of the sliding window is small, this kind of situation is more likely to occur.
[0080] According to the abnormal waveform data screening conditions, the abnormal waveforms caused by improper installation or sensor failure are screened out, and this part of abnormal data will not be input into the fault diagnosis model (the current initial model), which can prevent interference with the fault diagnosis results.
[0081] S202. If the initial model cannot predict the fault type of the vibration data, then the vibration data is used as the target data.
[0082] S203. Set a fault label for the target data.
[0083] S204. Determine whether the initial model meets the preset update conditions.
[0084] If so, execute S205; if not, return to execute S201.
[0085] S202 - S204 is similar to S102 - S104 in the first embodiment, and the specific description can refer to the relevant description in the first embodiment.
[0086] S205. Divide the target data and the corresponding fault labels into a training set and a validation set according to a preset ratio.
[0087] Exemplarily, the data ratio of the training set and the validation set can be 8:2. The training set is used to train the performance of the model, and the validation set is used to verify the performance of the model.
[0088] S206. Copy the initial model to obtain an incremental learning model and a copy model respectively.
[0089] S207. Using the prediction performance of the copy model as a constraint, train the incremental learning model with the training set to update the model parameters of the incremental learning model.
[0090] The copy model is the current initial model. Before the first update training, the model parameters of the copy model, the incremental learning model, and the initial model are the same. The update training only updates the model parameters of the incremental learning model. Therefore, after the update training, the model parameters of the copy model and the initial model still remain the same, while the model parameters of the incremental learning model have changed.
[0091] Specifically, the initial model includes a fully connected layer and an output layer. Figure 3A schematic diagram of the model incremental update training process provided by the present invention is as follows Figure 3 As shown in the figure, taking the prediction performance of the replica model as a constraint, a training set is used to train the incremental learning model to update the model parameters of the incremental learning model, including:
[0092] Input the training set into the replica model to obtain the first matrix in the fully connected layer of the replica model; use the training set to update and train the incremental learning model, and obtain the second matrix in the fully connected layer of the incremental learning model and the predicted label in the output layer; calculate the contrast loss based on the first matrix and the corresponding second matrix; calculate the category loss based on the predicted label and the corresponding fault label; update the model parameters of the incremental learning model based on the contrast loss and the category loss.
[0093] Among them, the contrast loss is used to add constraints to the model update process to prevent forgetting of old knowledge when learning new knowledge, and the category loss is used to guide the model to learn the recognition ability of new category data. Specifically, the contrast loss and the category loss can be determined by calculating the cross-entropy loss function, semantic cosine value, etc.
[0094] S208. Use the validation set to verify whether the updated incremental learning model meets the preset accuracy condition.
[0095] After training, the updated incremental learning model is obtained. Input the validation set into the model to verify the accuracy of the model. Here, the accuracy refers to the classification accuracy of the model on the test set. When the accuracy is greater than or equal to the set accuracy threshold, it indicates that the model update is completed, and then execute S209 to replace the current fault diagnosis model with the incrementally updated model; when the accuracy is less than the set accuracy threshold, it indicates that the model update fails to meet the standard, and then execute S207 to continue training the model. Among them, the accuracy threshold set when verifying the accuracy of the updated model and the accuracy threshold set when verifying whether the model meets the preset update condition can be the same or different, and the present invention does not limit this.
[0096] Specifically, the process of using the validation set to verify the incremental learning model is mainly as follows: Input the validation set into the incremental learning model to obtain the predicted label, compare the predicted label with the actual label to determine whether the predicted label is correct. After all the validation set data is input, count the classification accuracy of the incremental learning model, that is, the model accuracy, and compare this model accuracy with the preset accuracy threshold. If it is greater than or equal to the accuracy threshold, it meets the preset accuracy condition, otherwise it does not meet the preset accuracy condition.
[0097] S209. Replace the current initial model with the incremental learning model.
[0098] The embodiments of the present invention propose a complete incremental learning training process that conforms to the actual situation. The model can not only complete the fault diagnosis task, but also has the ability to screen unknown type data, reducing the workload of manually calibrating unknown data types; a screening mechanism for abnormal waveform data is proposed, significantly reducing the false alarm rate and improving the diagnostic accuracy and reliability of the system; by establishing an incremental learning model copy, the forgetting of old knowledge by the model during new data training is avoided, reducing the training cost of the model and the memory requirement for data storage.
[0099] Embodiment III
[0100] Figure 4 FIG. is a schematic structural diagram of a device fault diagnosis model updating device provided by Embodiment III of the present invention. As Figure 4 shown, the device fault diagnosis model updating device includes:
[0101] A data acquisition and processing module 100, configured to collect vibration data of a preset component of the device at a preset period during the operation of the device, and input the vibration data into an initial model;
[0102] A target data determination module 200, configured to use the vibration data as target data if the initial model cannot predict the fault type of the vibration data;
[0103] A fault label setting module 300, configured to set a fault label for the target data;
[0104] An update judgment module 400, configured to judge whether the initial model reaches a preset update condition; if so, execute the content of the model update module 500, if not, execute the content of the data acquisition and processing module 100;
[0105] A model update module 500, configured to update the initial model with the target data and the corresponding fault label under the constraint of the prediction performance of the initial model.
[0106] Optionally, the data acquisition and processing module 100 includes:
[0107] An abnormal data judgment sub-module, configured to judge whether the vibration data is abnormal data according to the change amplitude of the vibration data; if so, execute the content of the data deletion sub-module, if not, execute the content of the data input sub-module;
[0108] A data deletion sub-module, configured to delete the vibration data;
[0109] A data input sub-module, configured to input the vibration data into the initial model.
[0110] Optionally, the abnormal data judgment sub-module includes:
[0111] A subsequence acquisition unit, configured to perform non-overlapping sliding on the vibration data by using a window with a preset length for each piece of the vibration data, so as to obtain a plurality of subsequences;
[0112] A statistical data acquisition unit, configured to calculate the average value, standard deviation and effective value of each of the subsequences, so as to obtain three sets of statistical data;
[0113] A fluctuation coefficient calculation unit, configured to calculate the fluctuation coefficient for each of the three sets of the statistical data respectively, and obtain a preset fluctuation coefficient threshold corresponding to each set of the statistical data;
[0114] An abnormal data determination unit, configured to determine that the vibration data is abnormal data if there is a fluctuation coefficient greater than the preset fluctuation coefficient threshold.
[0115] Optionally, the parameters in the standard deviation calculation formula of the subsequence include the average value of the subsequence, and the fluctuation coefficient calculation unit includes:
[0116] A vibration data average value calculation component, configured to calculate the average value of the vibration data;
[0117] An average value replacement component, configured to use the average value of the vibration data as the average value in the standard deviation calculation formula of the subsequence;
[0118] A standard deviation calculation component, configured to calculate the standard deviation of each of the subsequences based on the standard deviation calculation formula.
[0119] Optionally, the model update module 500 includes:
[0120] A data grouping sub-module, configured to divide the target data and the corresponding fault labels into a training set and a validation set according to a preset ratio;
[0121] A model copying sub-module, configured to copy the initial model to obtain an incremental learning model and a copy model respectively;
[0122] A model update sub-module, configured to train the incremental learning model by using the training set with the prediction performance of the copy model as a constraint, so as to update the model parameters of the incremental learning model;
[0123] An accuracy judgment sub-module, configured to verify whether the updated incremental learning model meets a preset accuracy condition by using the validation set; if so, execute the content of the model replacement sub-module, and if not, execute the content of the model update sub-module;
[0124] A model replacement sub-module, configured to replace the current initial model with the incremental learning model.
[0125] Optionally, the initial model includes a fully connected layer and an output layer. The model update sub-module includes:
[0126] A first input-output unit for inputting the training set into the replica model to obtain a first matrix in the fully connected layer of the replica model;
[0127] A second input-output unit for updating and training the incremental learning model using the training set, obtaining a second matrix in the fully connected layer of the incremental learning model, and obtaining predicted labels in the output layer;
[0128] A contrast loss calculation unit for calculating a contrast loss based on the first matrix and the corresponding second matrix;
[0129] A class loss calculation unit for calculating a class loss based on the predicted labels and the corresponding fault labels;
[0130] A model parameter update unit for updating the model parameters of the incremental learning model based on the contrast loss and the class loss.
[0131] Optionally, the preset update condition is that the prediction accuracy of the initial model is less than a preset accuracy threshold.
[0132] The device fault diagnosis model update device provided by the embodiments of the present invention can execute the device fault diagnosis model update method provided by any embodiment of the present invention, and has corresponding functional modules and beneficial effects for executing the method.
[0133] Embodiment 4
[0134] Figure 5 FIG. shows a schematic structural diagram of an electronic device 40 that can be used to implement the embodiments of the present invention. The electronic device is intended to represent various forms of digital computers, such as, laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as, personal digital processing, cellular phones, smart phones, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present invention described and / or claimed herein.
[0135] As Figure 5As shown, the electronic device 40 includes at least one processor 41 and a memory communicatively connected to the at least one processor 41, such as a read-only memory (ROM) 42, a random access memory (RAM) 43, etc. The memory stores a computer program executable by the at least one processor. The processor 41 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 42 or the computer program loaded from the storage unit 48 into the random access memory (RAM) 43. In the RAM 43, various programs and data required for the operation of the electronic device 40 can also be stored. The processor 41, the ROM 42, and the RAM 43 are connected to each other via a bus 44. An input / output (I / O) interface 45 is also connected to the bus 44.
[0136] Multiple components in the electronic device 40 are connected to the I / O interface 45, including: an input unit 46, such as a keyboard, a mouse, etc.; an output unit 47, such as various types of displays, speakers, etc.; a storage unit 48, such as a magnetic disk, an optical disc, etc.; and a communication unit 49, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 49 allows the electronic device 40 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.
[0137] The processor 41 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the processor 41 include but are not limited to a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 41 executes the various methods and processes described above, such as the device fault diagnosis model update method.
[0138] In some embodiments, the device fault diagnosis model update method can be implemented as a computer program tangibly embodied in a computer-readable storage medium, such as the storage unit 48. In some embodiments, part or all of the computer program can be loaded and / or installed onto the electronic device 40 via the ROM 42 and / or the communication unit 49. When the computer program is loaded into the RAM 43 and executed by the processor 41, one or more steps of the device fault diagnosis model update method described above can be executed. Alternatively, in other embodiments, the processor 41 can be configured to execute the device fault diagnosis model update method by any other appropriate means (e.g., by means of firmware).
[0139] The various embodiments of the systems and techniques described above in this specification can be implemented in digital electronic circuitry, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on a chip (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include: being implemented in one or more computer programs that are executable and / or interpretable on a programmable system including at least one programmable processor, which can be a special-purpose or general-purpose programmable processor that receives data and instructions from, and transmits data and instructions to, a storage system, at least one input device, and at least one output device.
[0140] The computer programs for implementing the methods of the present invention can be written in any combination of one or more programming languages. These computer programs can be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus, such that the computer programs, when executed by the processor, cause the functions / operations specified in the flowchart and / or block diagram to be implemented. The computer programs can be executed entirely on the machine, partly on the machine, as a stand-alone software package partly on the machine and partly on a remote machine or entirely on the remote machine or server.
[0141] In the context of the present invention, a computer-readable storage medium can be a tangible medium that can contain, or store a computer program for use by or in connection with an instruction execution system, apparatus, or device. The computer-readable storage medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. Alternatively, the computer-readable storage medium can be a machine-readable signal medium. More specific examples of the machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.
[0142] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or a trackball) through which the user can provide input to the electronic device. Other kinds of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).
[0143] The systems and techniques described herein can be implemented in a computing system including backend components (e.g., as a data server), or a computing system including middleware components (e.g., an application server), or a computing system including frontend components (e.g., a user computer having a graphical user interface or a web browser through which the user can interact with an implementation of the systems and techniques described herein), or a computing system including any combination of such backend components, middleware components, or frontend components. The components of the system can be interconnected by digital data communication in any form or medium (e.g., a communication network). Examples of communication networks include: local area network (LAN), wide area network (WAN), blockchain network, and the Internet.
[0144] A computing system can include a client and a server. The client and the server are generally far from each other and typically interact through a communication network. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or a cloud host, which is a host product in the cloud computing service system and solves the defects of difficult management and weak business scalability existing in traditional physical hosts and VPS services.
[0145] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps recited in the present invention can be executed in parallel, sequentially, or in a different order, as long as the desired results of the technical solution of the present invention can be achieved, and no limitation is made herein.
[0146] The above specific embodiments do not constitute a limitation on the protection scope of the present invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.
Claims
1. A method for updating a device fault diagnosis model, characterized in that: The device is an electromechanical device, and the method comprises: During the operation of the equipment, vibration data of preset components of the equipment are collected at a preset period, and the vibration data are input into the initial model; If the initial model cannot predict the fault type of the vibration data, taking the vibration data as target data; Setting a fault tag for the target data; Determining whether the initial model meets a preset update condition; If so, updating the initial model using the prediction performance of the initial model as a constraint and the target data and the corresponding fault label; If not, return to the step of collecting vibration data of preset components in the equipment at a preset period during the operation of the equipment and inputting the data into the initial model.
2. The method according to claim 1, characterized in that Before inputting the vibration data into the initial model, it also includes: Determining whether the vibration data is abnormal data according to the change amplitude of the vibration data; If yes, deleting the vibration data; If not, the step of inputting the vibration data into an initial model is performed.
3. The method according to claim 2, characterized in that The determining whether the vibration data is abnormal data according to the change amplitude of the vibration data includes: For each of the vibration data, a window of a preset length is used to slide on the vibration data without overlap to obtain multiple subsequences; Calculate the mean value, standard deviation and effective value of each subsequence to obtain three sets of statistical data; Calculating fluctuation coefficients for the three groups of statistical data respectively, and obtaining a preset fluctuation coefficient threshold corresponding to each group of statistical data; If the fluctuation coefficient is greater than a preset fluctuation coefficient threshold, the vibration data is determined to be abnormal data.
4. The method according to claim 3, characterized in that The parameters in the formula for calculating the standard deviation of the subsequences include the average value of the subsequences, and calculating the standard deviation of each subsequence includes: Calculating an average value of the vibration data; Using the average value of the vibration data as the average value in the standard deviation calculation formula of the subsequence; The standard deviation of each of the subsequences is calculated based on the standard deviation calculation formula.
5. The method according to claim 1, characterized in that The initial model is updated using the target data and the corresponding fault labels with the prediction performance of the initial model as a constraint, including: Dividing the target data and the corresponding fault labels into a training set and a validation set according to a preset ratio; The initial model is copied to obtain an incremental learning model and a copy model respectively; Taking the prediction performance of the replica model as a constraint, the incremental learning model is trained using the training set to update the model parameters of the incremental learning model; Using the verification set to verify whether the updated incremental learning model meets the preset accuracy condition; If so, replace the current initial model with the incremental learning model; If not, return to the step of training the incremental learning model using the training set with the prediction performance of the replica model as a constraint.
6. The method according to claim 5, characterized in that The initial model includes a fully connected layer and an output layer. The incremental learning model is trained using the training set with the prediction performance of the replica model as a constraint to update the model parameters of the incremental learning model, including: Inputting the training set into the replica model, and obtaining a first matrix at the fully connected layer of the replica model; The incremental learning model is updated and trained using the training set, and a second matrix is obtained at a fully connected layer of the incremental learning model, and a predicted label is obtained at an output layer; Calculate contrast loss based on the first matrix and the corresponding second matrix; Calculating a class loss based on the predicted label and the corresponding fault label; Model parameters of the incremental learning model are updated based on the contrastive loss and the category loss.
7. The method according to any one of claims 1 to 6, characterized in that: The preset updating condition is that the prediction accuracy of the initial model is less than a preset accuracy threshold.
8. A device for updating a fault diagnosis model of an equipment, characterized in that: include: A data collection and processing module, used to collect vibration data of preset components of the equipment at a preset period during the operation of the equipment, and input the vibration data into the initial model; a target data determination module, configured to use the vibration data as target data if the initial model cannot predict the fault type of the vibration data; A fault label setting module, used for setting a fault label for the target data; An update judgment module, used to judge whether the initial model meets the preset update conditions; If yes, the content of the model update module is executed, if no, the content of the data acquisition and processing module is executed; The model updating module is used to update the initial model by using the target data and the corresponding fault labels, taking the prediction performance of the initial model as a constraint.
9. An electronic device, characterized in that: The electronic device comprises: at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the device fault diagnosis model updating method according to any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the device fault diagnosis model updating method according to any one of claims 1 to 7 when executed.