Intelligent operation and maintenance strategy making method and device and related system

Abstract

The invention provides an intelligent operation and maintenance strategy making method and device and a related system. The method comprises the steps that a first operation data set of target equipment in a historical time period and a second operation data set of the target equipment in an operation and maintenance time period are acquired; determining a first fault type data set according to the first operation data set; determining a first synthetic data set according to the first operation data set and the first fault type data set; determining an operation fault data set according to the first operation data set, the first fault type data set and the first synthetic data set; training a fault prediction model through the operation fault data set to obtain a fault prediction model, and inputting the second operation data set into the fault prediction model to obtain a fault prediction result; and generating an operation and maintenance strategy according to the fault prediction result, and scheduling corresponding operation and maintenance personnel to perform equipment operation and maintenance through the operation and maintenance strategy. By executing the method, high efficiency and reliability in the operation and maintenance process of the equipment can be improved.

Description

Technical Field

[0001] The present application relates to the technical field of power equipment maintenance, and particularly to a method, device and related system for formulating an intelligent operation and maintenance strategy. Background Art

[0002] The operation and maintenance strategy is a key factor to ensure the efficient and safe operation of equipment. By formulating a scientific maintenance plan, optimizing the allocation of equipment management and maintenance resources, it can maximize the reliability and service life of equipment. A reasonable operation and maintenance strategy can effectively reduce the occurrence of equipment failures, reduce equipment downtime, improve production efficiency, and reduce unnecessary maintenance expenses. The current equipment operation and maintenance strategies mainly include time-based maintenance strategies and condition-based maintenance strategies. The time-based maintenance strategy relies on preset time intervals for regular maintenance, and the condition-based maintenance strategy judges whether maintenance is required by setting thresholds through real-time monitoring of equipment status data such as temperature and vibration.

[0003] However, when faced with complex equipment and a highly dynamic working environment, the effectiveness and adaptability of the current equipment operation and maintenance strategies are significantly insufficient. The time-based maintenance strategy does not consider the actual operating status of the equipment, which is prone to over-maintenance and resource waste, thus increasing the operation and maintenance costs; usually, equipment failures do not occur only when a certain threshold is reached, and the occurrence of failures is often gradual, and the condition-based maintenance strategy is prone to overlooking or misjudging potential failures.

[0004] Therefore, in the operation and maintenance of equipment, how to improve the efficiency and reliability of operation and maintenance urgently needs to be solved. Summary of the Invention

[0005] The embodiments of the present application provide a method, device and related system for formulating an intelligent operation and maintenance strategy, which realizes improving the efficiency and reliability of equipment operation and maintenance in equipment operation and maintenance.

[0006] In a first aspect, the embodiments of the present application provide a method for formulating an intelligent operation and maintenance strategy, which is applied to a server, and the method includes:

[0007] Obtain a first operation data set of a target device within a historical time period, and a second operation data set of the target device within an operation and maintenance time period; the start time of the operation and maintenance time period is later than the end time of the historical time period;

[0008] Determine the fault type corresponding to the target device according to the first operation data set to obtain a first fault type data set;

[0009] Determine a first composite data set according to the first operation data set and the first fault type data set;

[0010] Determine the operating fault dataset of the target device according to the first operating dataset, the first fault type dataset, and the first synthetic dataset;

[0011] Train a preset fault prediction model with the operating fault dataset to obtain a fault prediction model;

[0012] Input the second operating dataset into the fault prediction model to obtain a fault prediction result;

[0013] Generate an operation and maintenance repair strategy for the target device according to the fault prediction result, so as to dispatch corresponding operation and maintenance personnel for device operation and maintenance repair through the operation and maintenance repair strategy.

[0014] In a second aspect, an embodiment of the present application provides a device for formulating an intelligent operation and maintenance repair strategy, which is applied to a server. The device includes:

[0015] An acquisition unit, configured to acquire a first operating dataset of a target device within a historical time period, and a second operating dataset of the target device within an operation and maintenance time period; the start time of the operation and maintenance time period is later than the end time of the historical time period;

[0016] A determination unit, configured to determine the fault type corresponding to the target device according to the first operating dataset to obtain a first fault type dataset;

[0017] The determination unit is further configured to determine a first synthetic dataset according to the first operating dataset and the first fault type dataset;

[0018] The determination unit is further configured to determine the operating fault dataset of the target device according to the first operating dataset, the first fault type dataset, and the first synthetic dataset;

[0019] A calculation unit, configured to train a preset fault prediction model with the operating fault dataset to obtain a fault prediction model;

[0020] The calculation unit is configured to input the second operating dataset into the fault prediction model to obtain a fault prediction result;

[0021] A control unit, configured to generate an operation and maintenance repair strategy for the target device according to the fault prediction result, so as to dispatch corresponding operation and maintenance personnel for device operation and maintenance repair through the operation and maintenance repair strategy.

[0022] In a third aspect, an embodiment of the present application provides a server, including a processor, a memory, a communication interface, and one or more programs. Among them, the above one or more programs are stored in the above memory and are configured to be executed by the above processor. The above programs include instructions for executing the steps in any of the methods in the first aspect of the embodiments of the present application.

[0023] In a fourth aspect, an embodiment of the present application provides a system for formulating an intelligent operation and maintenance inspection strategy. Among them, the above system for formulating an intelligent operation and maintenance inspection strategy is used to execute some or all of the steps described in any of the methods in the first aspect of the embodiments of the present application.

[0024] In a fifth aspect, an embodiment of the present application provides a computer-readable storage medium. Among them, the above computer-readable storage medium stores a computer program for electronic data exchange. The above computer program causes a computer to execute some or all of the steps described in any of the methods in the first aspect of the embodiments of the present application.

[0025] In a sixth aspect, an embodiment of the present application provides a computer program product. Among them, the above computer program product includes a non-transitory computer-readable storage medium storing a computer program. The above computer program is operable to cause a computer to execute some or all of the steps described in any of the methods in the first aspect of the embodiments of the present application. This computer program product can be a software installation package.

[0026] By implementing the embodiments of the present application, the following technical effects can be achieved.

[0027] A method for formulating an intelligent operation and maintenance inspection strategy described in this application is applied to a server. By obtaining a first operation data set of a target device within a historical time period and a second operation data set of the target device within an operation and maintenance time period, where the start time of the operation and maintenance time period is later than the end time of the historical time period; then, determining the fault type corresponding to the target device according to the first operation data set to obtain a first fault type data set; determining a first synthetic data set according to the first operation data set and the first fault type data set; determining an operation fault data set of the target device according to the first operation data set, the first fault type data set, and the first synthetic data set; then, training a preset fault prediction model through the operation fault data set to obtain a fault prediction model; inputting the second operation data set into the fault prediction model to obtain a fault prediction result; finally, generating an operation and maintenance inspection strategy for the target device according to the fault prediction result, so as to dispatch corresponding operation and maintenance personnel for equipment operation and maintenance inspection through the operation and maintenance inspection strategy. In this way, on the one hand, by analyzing the historical data and real-time data of the equipment to predict possible future faults of the equipment, maintenance can be carried out in advance, thus avoiding sudden equipment failures. Compared with traditional time-based maintenance strategies and condition-based maintenance strategies, the efficiency of equipment operation and maintenance inspection is improved; on the other hand, by innovatively using a generative adversarial network model to generate new fault data according to the historical operation fault data of the equipment and training the model, the generalization performance of the fault prediction model is improved, so as to improve the reliability of model prediction and further improve the reliability of equipment operation and maintenance inspection. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the accompanying drawings required for the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings in the following description are some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0029] Figure 1 is an architecture diagram of a system for formulating an intelligent operation and maintenance inspection strategy provided by an embodiment of the present application;

[0030] Figure 2 is a schematic structural diagram of a server provided by an embodiment of the present application;

[0031] Figure 3 is a schematic flow diagram of a method for formulating an intelligent operation and maintenance inspection strategy provided by an embodiment of the present application;

[0032] Figure 4 is a schematic flow diagram of another method for formulating an intelligent operation and maintenance inspection strategy provided by an embodiment of the present application;

[0033] Figure 5 It is a schematic diagram of the scenario of a system for formulating an intelligent operation and maintenance inspection strategy provided by an embodiment of the present application;

[0034] Figure 6 It is a display diagram of the interface of a system for formulating an intelligent operation and maintenance inspection strategy provided by an embodiment of the present application;

[0035] Figure 7 It is a display diagram of the interface of another system for formulating an intelligent operation and maintenance inspection strategy provided by an embodiment of the present application;

[0036] Figure 8 It is a block diagram of the functional units of another device for formulating an intelligent operation and maintenance inspection strategy provided by an embodiment of the present application. Detailed implementation manners

[0037] In order to enable those skilled in the art to better understand the solution of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0038] The terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally further includes steps or units not listed, or optionally further includes other steps or units inherent to these processes, methods, products or devices.

[0039] It should be understood that the term "and / or" in this article is only a description of the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B may represent: A exists alone, A and B exist simultaneously, and B exists alone. In addition, the character " / " in this article indicates that the associated objects before and after are in an "or" relationship. The "multiple" mentioned in the embodiments of the present application refers to two or more.

[0040] The "at least one (item)" or its similar expression in the embodiments of the present application refers to any combination of these items, including any combination of single item (s) or plural items (s), meaning one or more, and multiple means two or more. For example, at least one (item) of a, b, or c can represent the following seven cases: a, b, c, a and b, a and c, b and c, a, b, and c. Each of a, b, and c can be an element or a set containing one or more elements.

[0041] The "connection" that appears in the embodiments of the present application refers to various connection methods such as direct connection or indirect connection to achieve communication between devices, and the embodiments of the present application do not make any limitations on this.

[0042] Referring to "embodiment" in this context means that the specific features, structures, or characteristics described in connection with the embodiment can be included in at least one embodiment of the present application. The phrase appearing in various positions in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment mutually exclusive with other embodiments. Those skilled in the art explicitly and implicitly understand that the embodiments described herein can be combined with other embodiments.

[0043] First, the relevant terms involved in the present application are explained as follows:

[0044] Generative Adversarial Networks: Generative Adversarial Networks (GAN) is an unsupervised deep learning method. It consists of a Generator and a Discriminator. These two networks compete with each other. By continuously improving their respective capabilities, they finally generate data close to "real" data, achieving the effect that the generated data is almost indistinguishable from the real data.

[0045] In the face of complex devices and highly dynamic working environments, there are significant deficiencies in the effectiveness and adaptability of the device operation and maintenance inspection strategies. During the process of device operation and maintenance inspection, problems such as insufficient reliability and low efficiency occur.

[0046] To solve the problems of insufficient efficiency and reliability in operation and maintenance, the embodiments of the present application provide a method, device and related system for formulating an intelligent operation and maintenance strategy, which are applied to a server. By obtaining a first operation data set of a target device within a historical time period and a second operation data set of the target device within an operation and maintenance time period, where the start time of the operation and maintenance time period is later than the end time of the historical time period; then, determining the fault type corresponding to the target device according to the first operation data set to obtain a first fault type data set; determining a first synthetic data set according to the first operation data set and the first fault type data set; determining an operation fault data set of the target device according to the first operation data set, the first fault type data set and the first synthetic data set; then, training a preset fault prediction model through the operation fault data set to obtain a fault prediction model; inputting the second operation data set into the fault prediction model to obtain a fault prediction result; finally, generating an operation and maintenance strategy for the target device according to the fault prediction result, so as to dispatch corresponding operation and maintenance personnel for equipment operation and maintenance through the operation and maintenance strategy. In this way, on the one hand, by analyzing the historical data and real-time data of the equipment, predicting possible future faults of the equipment, and being able to perform maintenance in advance, thus avoiding sudden equipment failures, compared with traditional time-based maintenance strategies and condition-based maintenance strategies, the efficiency of equipment operation and maintenance is improved; on the other hand, by innovatively using a generative adversarial network model to generate new fault data according to the historical operation fault data of the equipment and training the model, the generalization performance of the fault prediction model is improved, so as to improve the reliability of model prediction and further improve the reliability of equipment operation and maintenance.

[0047] The following combines Figure 1 to illustrate the architecture of a system for formulating an intelligent operation and maintenance strategy in the embodiments of the present application. Figure 1 FIG. 1 is an architecture diagram of a system for formulating an intelligent operation and maintenance strategy provided by the embodiments of the present application. The system 100 for formulating an intelligent operation and maintenance strategy includes: an equipment health status monitoring module 110, a fault prediction module 120, a maintenance scheduling module 130, a data processing module 140, and an operating device 150.

[0048] Among them, the device health status monitoring module 110 is used to collect multi-dimensional status data of the operating device 150 in real time, and monitor parameters such as temperature, vibration, pressure, and current through sensors deployed at key parts of the operating device 150. In addition, the device health status monitoring module 110 also detects the health status of device components in the operating device 150, dynamically monitors the working conditions of the operating device 150, identifies abnormal parameter fluctuations, and provides original data for fault prediction, etc. The functions are not limited here. In actual operation, the device health status monitoring module 110 establishes a real-time data link with the operating device 150 to ensure the timeliness and accuracy of data, which is used as the basic data input of the system.

[0049] Among them, the fault prediction module 120 receives the monitored multi-dimensional status data from the device health status monitoring module 110, analyzes and processes it using machine learning technology to achieve early fault prediction. The module has multiple built-in machine learning models, analyzes data features and matches models, judges potential fault hazards, predicts the probability and time of fault occurrence, generates prediction results, and provides a decision-making basis for maintenance scheduling. Specifically, the fault prediction module 120 analyzes the operating status data of the operating device 150 to identify the possible fault types, occurrence time, and remaining useful life (RUL) of the operating device 150.

[0050] Among them, the maintenance scheduling module 130 receives the prediction information from the fault prediction module 120. The prediction information includes at least one of the fault type, fault occurrence time, and remaining useful life, which is not limited here. Combining the operating requirements and production plan of the operating device 150, it formulates a maintenance scheduling plan. Based on the health status of the device and the fault prediction results, it prioritizes maintenance tasks and reasonably allocates maintenance personnel, tools, and resources. The maintenance scheduling module 130 ensures that maintenance activities are carried out at the most appropriate time by analyzing the urgency of fault occurrence and the remaining useful life of the device, minimizing device downtime and reducing maintenance costs.

[0051] Among them, the data processing module 140 is used for data processing work in the intelligent operation and maintenance repair strategy formulation system. On the one hand, it cleans, denoises, and standardizes the original data of the device health status monitoring module 110 to improve the quality; on the other hand, it stores the prediction data of the fault prediction module 120 and the execution data of the maintenance scheduling module 130 to form an operation and maintenance log. In addition, the data processing module 140 also supports secondary data analysis, mines historical operation and maintenance data to optimize the fault prediction model and maintenance strategy, and constructs a data-driven intelligent operation and maintenance repair closed loop.

[0052] Among them, the operating device 150 is used as the system monitoring object, covering key devices in the power system. The device health status monitoring module 110 obtains its operation data, which is processed by the data processing module 140 and then analyzed and predicted by the fault prediction module 120, and then the maintenance scheduling is executed through the maintenance scheduling module 130. Each module forms a data interaction and business collaboration link, realizing the full-process intelligent operation and maintenance from status monitoring, fault prediction, maintenance scheduling to data optimization, improving the device reliability and operation and maintenance efficiency, reducing production losses, providing a scientific and intelligent operation and maintenance inspection strategy plan for the operating device 150, and ensuring the stable operation of the device and the efficient and continuous operation of industrial production. At the same time, an incremental learning mechanism is introduced, which runs through the system for formulating the intelligent operation and maintenance inspection strategy of Zhen Gege. By continuously updating and adjusting the fault prediction model, it can be optimized in real time as the device operation state changes. During the device operation, the system continuously obtains new real-time data, and the incremental learning mechanism will adjust the prediction model according to the new data. In this way, the fault prediction model can adapt to the changes in the device state over time, provide more accurate prediction results, and further improve the decision-making quality of the maintenance scheduling module.

[0053] In a possible embodiment, when the operating device 150 is in an industrial production operation state, the device health status monitoring module 110 collects key part data in real time. For example, in motor-type mechanical equipment, the device health status monitoring module 110 collects bearing temperature through a temperature sensor, obtains the vibration amplitude and frequency of the device through a vibration sensor, monitors the hydraulic system pressure value through a pressure sensor, forms multi-dimensional operation data, and transmits it to the data processing module 140 in real time. The data processing module 140 starts the data processing process, uses a filtering algorithm to eliminate abnormal jump values in the temperature data, fills in the missing period of the vibration data through linear interpolation, and then normalizes parameters such as pressure, and outputs a standardized data set. The fault prediction module 120 obtains the processed data set and analyzes it based on the trained fault prediction model. Then, after receiving the result of the fault prediction module 120, the maintenance scheduling module 130 performs resource scheduling in combination with production arrangements. Given that the remaining operating time of the device meets the requirements of the current production batch, the maintenance scheduling module 130 plans the maintenance task to the next production downtime window. Subsequently, it schedules operation and maintenance personnel with mechanical maintenance qualifications, synchronously retrieves maintenance resources such as spare parts and vibration detectors, and generates a detailed work order including maintenance time nodes, operation steps, and quality acceptance standards. In the maintenance execution stage, the operation and maintenance personnel perform maintenance on the operating device 150 according to the work order, completing component replacement and equipment debugging. After the maintenance is completed, the device health status monitoring module 110 collects the device operation data again and transmits it to the data processing module 140. The data processing module 140 integrates the data before and after maintenance into a historical operation and maintenance file for the fault prediction module 120 to optimize the model. The fault prediction module 120 uses an incremental learning algorithm to integrate new data into the training set and update the model parameters to improve the accuracy of fault prediction for similar devices.

[0054] It can be seen that through the above architecture of a system for formulating an intelligent operation and maintenance inspection strategy, the efficiency and reliability of operation and maintenance inspection can be greatly improved.

[0055] The following will be combined with Figure 2 to illustrate the server in the embodiments of the present application. Figure 2 is a schematic structural diagram of a server provided by an embodiment of the present application. As Figure 2 shown, the server 200 includes one or more processors 210, a memory 220, a communication interface 230, and one or more programs 221. The processor 210 is communicatively connected to the memory 220 and the communication interface 230 through an internal communication bus.

[0056] Among them, the processor 210 can be a Central Processing Unit (CPU), a general-purpose processor, a Digital Signal Processor (DSP), an Application-Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute various exemplary logical blocks, units, and circuits described in connection with the disclosure of this application. The processor can also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, and so on. The communication unit can be a communication interface, a transceiver, a transceiver circuit, etc., and the storage unit can be a memory.

[0057] Among them, the memory 220 can be a volatile memory or a non-volatile memory, or can include both volatile and non-volatile memories. Among them, the non-volatile memory can be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory can be a random access memory (RAM), which is used as an external cache. By way of example but not limitation, many forms of random access memory (RAM) are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchlink dynamic random access memory (SLDRAM), and direct rambus random access memory (DR RAM).

[0058] Among them, the one or more programs 221 are stored in the above-mentioned memory 220 and are configured to be executed by the above-mentioned processor 210. The one or more programs 221 include instructions for executing any step in the following embodiments of a data processing method.

[0059] It can be understood that the server 200 may include more or fewer structural elements than those in the above structural block diagram. For example, it includes a power module, physical buttons, a Wi-Fi module, a speaker, a Bluetooth module, sensors, a display module, etc., which are not limited herein. It can be understood that the server 200 can carry an architecture of a system for formulating an intelligent operation and maintenance inspection strategy as Figure 1 described.

[0060] After understanding the software and hardware architecture of the embodiments of the present application, the following combines Figure 3 to describe a method for formulating an intelligent operation and maintenance inspection strategy in the embodiments of the present application. Figure 3 is a schematic flowchart of a method for formulating an intelligent operation and maintenance inspection strategy provided by the embodiments of the present application, which specifically includes the following steps:

[0061] Step S310, obtain a first operation data set of a target device within a historical time period, and a second operation data set of the target device within an operation and maintenance time period; the start time of the operation and maintenance time period is later than the end time of the historical time period.

[0062] Among them, the first operation data set is a multi-dimensional operation data set generated by the target device during the historical operation stage. This data set includes the operation parameters of the key components of the target device, such as the bearing temperature data collected by the temperature sensor, the device vibration frequency and amplitude data obtained by the vibration sensor, the hydraulic system pressure value monitored by the pressure sensor, and electrical parameters such as current and rotational speed. These data are collected in real time by sensors deployed on the operating device, reflecting the historical operating conditions of the target device and providing a data basis for analyzing the historical operating laws of the device and identifying potential failure modes.

[0063] Among them, the second operation data set is the multi-dimensional operation state data of the target device in the current stage. This data set not only includes basic operation parameters similar to the first operation data set, but also includes the device state data before and after the implementation of the operation and maintenance operations. For example, after the maintenance is performed, the device health status monitoring module will re-collect the operation data to verify the maintenance effect and judge whether the device performance has returned to normal. At the same time, the second operation data set can reflect the impact of the operation and maintenance measures on the device operation state and provide a basis for optimizing the subsequent operation and maintenance strategies.

[0064] Specifically, data collection is carried out by multiple sensor nodes for real-time monitoring and data acquisition at key parts of the device. These sensors can provide various operating parameters of the device, such as temperature, pressure, vibration, load, current, etc. The data collected in real time by the sensors usually exists in a time series form and needs to go through a strict preprocessing process before it can be used as input data for the fault diagnosis and prediction model. In the data preprocessing process, the first step is to remove outliers, fill in missing values, and standardize the data. The raw data collected by the sensors may have abnormal fluctuations, resulting in data deviating from the actual situation. Therefore, it is also necessary to denoise the data and remove outliers. Among them, the detection of outliers can adopt statistical methods, which can be the outlier detection method based on the mean and standard deviation, or the three-sigma principle (if the data point deviates from the mean by more than 3 times the standard deviation, then this point is considered an outlier) to remove outliers, etc. This is not limited here. The purpose of removing outliers is to ensure the quality of the data set and make it conform to normal physical behavior. In addition, it is also necessary to fill in the missing values. In practical applications, sensor failures or communication problems often lead to the loss of device data. Especially in large-scale device monitoring, the processing of missing values is crucial. The filling methods include mean filling, interpolation methods (such as linear interpolation, spline interpolation), and model-based filling methods, which are not limited here. Finally, the data is standardized. Sensor data usually has different dimensions and units. For example, temperature is expressed in degrees Celsius, while pressure may be in pascals. To avoid the interference of unit differences on subsequent calculations and modeling, it is necessary to standardize the data so that the input data has no bias.

[0065] Step S320, determine the fault type corresponding to the target device according to the first operating data set, and obtain the first fault type data set.

[0066] Among them, the raw data collected by the device sensors needs to be preprocessed and meaningful features need to be extracted. For example, the collected sensor data includes multiple time series data such as temperature, pressure, vibration, and rotational speed. The feature extraction process is to transform these data and extract the most representative features for fault diagnosis. Next, a machine learning algorithm is used to train the fault diagnosis model. For example, through deep learning models such as convolutional neural networks and long short-term memory networks, or machine learning algorithms such as support vector machines and random forests, feature mining and pattern matching are performed on multi-dimensional parameters such as temperature, vibration, and pressure in the first operating data set, and combined with the preset fault type knowledge base to identify the fault type of the target device.

[0067] Among them, the first fault type dataset is a systematic induction of the historical fault types of the target device, including key information such as fault type codes, fault names, characteristic parameter thresholds for triggering faults, and fault occurrence probabilities. For example, if the model analysis finds that the vibration data shows periodic impact characteristics and the temperature continues to rise, and it matches the "gear wear in the gearbox" fault type, the fault type and its corresponding characteristic parameter range will be recorded in the dataset to form a structured fault type record.

[0068] In a possible embodiment, determining the fault type corresponding to the target device during operation according to the first operation dataset to obtain the first fault type dataset specifically includes the following steps:

[0069] 321. Perform data preprocessing on the first operation dataset to obtain a target dataset;

[0070] 322. Perform feature selection on the target dataset to obtain a fault feature set;

[0071] 323. Input the fault feature set into a preset machine learning model, and output the fault components of the target device through the machine learning model to obtain a fault component set;

[0072] 324. Extract the operation parameters corresponding to each component in the fault component set from the first operation dataset to obtain a plurality of operation parameters;

[0073] 325. Extract the fault types corresponding to the plurality of operation parameters from a preset operation parameter and fault type database to obtain at least one fault type, and determine the first fault type dataset according to the at least one fault type.

[0074] Among them, the data preprocessing includes operations such as denoising, normalization, and missing value filling. Here is an example for illustration: Taking the vibration data of rotating machinery as an example, the wavelet threshold denoising algorithm is used to eliminate high-frequency noise, the Z-score normalization is used to unify the dimensions, and the K-nearest neighbor interpolation method is used to fill in the missing data to ensure data integrity and consistency. The preset machine learning model adopts a fusion architecture of a convolutional neural network (CNN) and a long short-term memory network (LSTM). The CNN is responsible for extracting the local spatial features of the vibration signal, and the LSTM captures the temporal dependence relationship of the temperature parameters. The accuracy of fault component recognition is improved through multi-modal feature fusion. It is also possible to use traditional machine learning methods such as support vector machines and random forests, which are not limited here.

[0075] Specifically, first, the raw data collected from the device sensors needs to be preprocessed and meaningful features need to be extracted. Assume that the collected sensor data includes multiple time-series data such as temperature, pressure, vibration, and rotational speed. The feature extraction process is to transform this data and extract the most representative features for fault diagnosis. For example, for time-domain feature extraction, for each set of time-series data x(t) (such as temperature, pressure, vibration), the following time-domain features can be calculated:

[0076] Mean value:

[0077] Standard deviation:

[0078] Maximum value: max(x) = max(x(t));

[0079] Minimum value: min(x) = min(x(t)),

[0080] where t represents the time index, x(t) represents the time-series data at time t, and N represents the number of samples of the time-series data x(t);

[0081] For the extraction of frequency-domain features, the fast Fourier transform (FFT) is used to convert the time-domain signal into the frequency domain and extract frequency features. Among them, the frequency components are obtained through Fourier transform to get the frequency spectrum X(f) of the signal:

[0082]

[0083] where f is the frequency, X(f) represents the spectrum value at frequency f, N is the number of sampling points, x(t) represents the value of the time-domain signal at time t, t is the discrete time index, the discrete time index t represents the position of each sampling point in the time domain, and i is the imaginary unit.

[0084] Next, through the time-domain and frequency-domain features extracted above, data input can be provided for subsequent machine learning algorithms. The purpose of feature selection is to retain the features that are important for fault diagnosis and eliminate redundant features. Feature selection methods include correlation analysis and principal component analysis (PCA), which are not limited here. A machine learning algorithm is used to train a fault diagnosis model. For example, support vector machine and random forest are selected as classification algorithms, and a regression algorithm is combined for position prediction. Among them, the device fault dataset is D = {(x i , y i ), where x i is the feature vector of the i-th sample, and y iIt is a fault type label. Machine learning methods (such as support vector machines, random forests, etc.) are used for training to obtain a fault type recognition model. For example, in the random forest method, a random forest is an ensemble learning method composed of multiple decision trees. For each tree, different random subsets of the data set will be used for training, and finally the classification result is determined through a voting mechanism. For sample x i , the prediction process of the random forest is as follows:

[0085] y pred = mode( T ree1(x i ), T ree2(x i ),..., T ree M (x i ))

[0086] where y pred represents the final prediction result, Tree1, Tree2,…Tree M represents M decision tree models, x i is the feature vector of the i-th sample, and mode is the voting function.

[0087] In addition, it also includes predicting the fault location. Fault location recognition is usually a regression problem, aiming to predict the location where the fault occurs. Assume that the label of the fault location is p i , which represents the component number or location where the device fails. Regression models (such as linear regression, support vector regression (SVR), etc.) can be used for prediction, which is not limited here. For example, using support vector regression, a hyperplane is constructed to fit the data, and the objective function of this SVR is:

[0088]

[0089] where w is the weight vector of the regression model, b is the bias vector, ∈ i is the prediction error, C is the penalty parameter, C is used to balance the regularization term and the error, N represents the total number of training samples, ||w|| 2 is the regularization term, and the output of SVR is the predicted location p i p i represents the fault occurrence area or component number of the device.

[0090] Among them, model training uses labeled historical data, and the data processed through feature engineering is used as input for supervised learning. During the supervised learning process, the algorithm calculates the loss function based on the input features and labels, and adjusts the model parameters through backpropagation or other optimization methods to minimize the prediction error.

[0091] For a clearer illustration, an example is given. First, the sliding window technique is called to segment the continuously running data to form a sample sequence of a fixed length. For example, the vibration data every 30 seconds is segmented into a sequence of 1000 sampling points, and each sequence corresponds to a time window. Then, based on the mutual information method, the correlation between each feature and the known fault label is calculated, and the features with a mutual information value greater than 0.5 are selected as the fault feature set. The fault feature set is input into the trained CNN-LSTM model, and the model outputs the fault probability distribution of each component. A threshold of 0.7 is set to select the components with a fault probability higher than the threshold to form a fault component set. Then, for the bearing components in the fault component set, the corresponding vibration acceleration, temperature, and rotational speed parameters are extracted, and the mapping relationship is stored in a decision tree structure according to the preset operating parameters and fault type database. For example, when the vibration acceleration exceeds 5g, the temperature is higher than 85°C, and the rotational speed fluctuation exceeds 10%, the "bearing overload" fault type is triggered. By using fuzzy logic to process the combined conditions of multiple parameters, the problem of fuzzy parameter threshold boundaries is solved, and the robustness of fault type matching is improved.

[0092] Step S330: Determine a first synthetic dataset according to the first operating dataset and the first fault type dataset.

[0093] Among them, the first synthetic dataset is formed by fusing historical operating data with synthetic fault data generated by a generative adversarial network. This dataset solves the limitations of model training in small sample scenarios by expanding the original fault samples. During the generation process, physical model constraint regularization is introduced to ensure that the synthetic data conforms to the physical laws of equipment operation.

[0094] In a possible embodiment, the determining the first synthetic dataset according to the first operating dataset and the first fault type dataset specifically includes the following steps:

[0095] 331. Obtain the input format of the preset generative adversarial network model to get a first input format;

[0096] 332. Based on the first operating dataset, determine the operating data corresponding to each fault in the first fault type dataset to obtain a fault-operating association dataset;

[0097] 333. Convert the fault-operating association dataset according to the first input format to obtain a first conversion dataset;

[0098] 334. Optimize the generative adversarial network model according to a preset first regularization formula to obtain a target generative adversarial network model;

[0099] 335. Input the first conversion dataset into the target generative adversarial network model to obtain the first synthetic dataset.

[0100] Among them, the generative adversarial network model can be an improved deep convolutional generative adversarial network architecture, and its input format is defined as a time series data sequence with a fixed length. For example, in the scenario of rotating machinery fault diagnosis, the first input format requires each sample to be a vibration acceleration signal of 1024 time steps, the data range is normalized to [-1, 1], and auxiliary parameters such as temperature and rotational speed are added as conditional inputs. This format is intercepted from continuous operation data through the sliding window technique to ensure the temporal correlation of the generated data.

[0101] Among them, in the context of device fault data generation, the goal of the improved generative adversarial network model is to generate synthetic data similar to the original fault data, and these synthetic data will be used as the input for subsequent training models to make up for the shortage of actual device data. The generator (G), the goal of the generator is to sample from a latent space and generate a set of device fault data x gen , whose goal is to approximate the distribution of real device fault data x as much as possible real ; the discriminator (D), the task of the discriminator is to judge whether the input data comes from the real data set. It optimizes itself based on the comparison between the generated synthetic data and the real data, and outputs a probability value D(x), indicating whether the input data is real data. During the training process, the generator continuously generates more real data, while the discriminator continuously improves its ability to distinguish between real data and generated data, and finally reaches an equilibrium state. The optimization goals of the generator and the discriminator are:

[0102]

[0103] Among them, E represents expectation, x represents real data samples, P data (x) is the probability distribution of real data samples, z represents a random noise vector, P z (z) represents the probability distribution of the noise, G(z) represents the synthetic data generated by the generator based on the noise z, D(x) represents the discriminant result of the discriminator for the real data x, D(G(z)) represents the discriminant result of the discriminator for the generated data G(z), V(D, G) represents the value function, and this value function is used to measure the adversarial effect between the discriminator and the generator. Among them, the input of GAN is the random noise z, and the output is the synthetic fault data x gen .

[0104] Since the fault data not only needs to conform to statistical laws but also follow the physical laws of the equipment, physical model constraints are introduced into the GAN. The purpose of physical model constraints is to ensure that the generated synthetic data is consistent with the actual operating behavior of the equipment at the physical level. During the operation of the equipment, the occurrence of faults is affected by many physical factors, such as temperature, pressure, vibration, etc. These physical factors have clear physical laws, and the generated data must satisfy these laws. Therefore, physical model constraint regularization is introduced to ensure that the generated data can reflect the physical characteristics of the real equipment. The physical model constraint regularization formula is as follows:

[0105]

[0106] Among them, L physical (x gen ) represents the physical constraint loss function, f i represents the extraction function of the i-th physical characteristic, f i (x gen ) is the output of the generated data in the i-th physical model, f i (x true ) is the output of the real equipment data in the i-th physical model, ∈ is the allowable error tolerance, and N represents the number of physical characteristics.

[0107] By introducing the physical model constraint regularization method, the generated data not only satisfies the statistical characteristics but also is consistent with the actual equipment data at the physical level, which can effectively improve the reliability and practical application value of the synthetic data. Then, the generated synthetic fault data will be used as input to expand the training set to solve the problem of scarce actual fault data. Through the model trained by the generative adversarial network, the generator can generate synthetic data that conforms to the equipment fault mode from the latent space. These synthetic data are similar to the real equipment fault data, which can not only reflect the operating state of the equipment but also satisfy the physical constraints of the equipment. The generated data can not only help train the fault prediction model but also enhance the fault identification ability of the intelligent operation and maintenance system. Especially in the case of low fault occurrence frequency or extreme working conditions, the generated data can effectively make up for the data deficiency and improve the generalization ability of the model. In addition, the generation of synthetic fault data not only helps to fill the data gap but also provides a simulation of unknown fault modes. The generated data can better adapt to the complex working conditions and fault types of the equipment, helping the intelligent operation and maintenance system to accurately diagnose potential faults and optimize the maintenance strategy.

[0108] In a possible embodiment, optimizing the generative adversarial network model according to the preset first regularization formula to obtain the target generative adversarial network model specifically includes the following steps:

[0109] 3341. Obtain the first device parameter set of the target device;

[0110] 3342. Determine the physical characteristics corresponding to each device parameter in the first device parameter set, obtaining n physical characteristics; n is an integer greater than 1;

[0111] 3343. Generate n physical models based on the n physical characteristics; each physical model in the n physical models represents the physical operating characteristics of a certain aspect of the target device;

[0112] 3344. Determine the objective function of the generative adversarial network model according to the first regularization formula and the n physical models;

[0113] 3345. Optimize the generative adversarial network model based on a preset optimization algorithm and the objective function to obtain a target generative adversarial network model.

[0114] Among them, the first device parameter set includes the design parameters and operating parameters of the target device. For example, for rotating machinery, the design parameters include mass, stiffness, damping, gear modulus, etc., and the operating parameters include rotational speed, load, ambient temperature, etc. Parameter acquisition combines the extraction of device nameplate data and real-time collection by sensors to ensure the accuracy and integrity of the parameters.

[0115] Among them, the mapping relationship between the physical characteristics and the device parameters is determined by prior knowledge. For example: the mass parameter corresponds to Newton's laws of motion; the stiffness parameter corresponds to Hooke's law; the temperature parameter corresponds to the first law of thermodynamics; the current parameter corresponds to Ohm's law, etc., which will not be elaborated here. Through this mapping relationship, the device parameters are transformed into quantifiable physical constraint conditions.

[0116] Specifically, first read the design parameters from the device file database and obtain the operating parameters in real time, and merge them to form the first device parameter set. For example, the parameter set of a certain motor includes: mass m = 50 kg, stator winding resistance R = 0.5 Ω, operating rotational speed ω = 1500 rpm, ambient temperature T = 25 °C. Then, establish the physical characteristic association for each parameter. For example: the vibration model is a multi-degree-of-freedom vibration differential equation established based on Newton's second law:

[0117] Mu(t)+Cu(t)+Ku(t)=F(t)

[0118] Among them, M is the mass matrix; C is the damping matrix; K is the stiffness matrix of the material; F(t) is the excitation force vector, u(t) represents the displacement vector, u(t) is a function of time t; F(t) is the external force vector, and F(t) is the excitation force acting on the system that changes with time t.

[0119] In a possible embodiment, determining the physical characteristics corresponding to each device parameter in the first device parameter set to obtain n physical characteristics specifically includes the following steps:

[0120] A1. Classify all device parameters in the first device parameter set to obtain m device types; m is an integer greater than 1; the device types are used to represent the functional types of the target device;

[0121] A2. According to the preset physical characteristic mapping relationship and the first device parameter set, determine the physical characteristics corresponding to each device type in the multiple device types to obtain n physical characteristics; the physical characteristic mapping relationship includes the association relationship between the device parameters under different device types and the corresponding physical characteristics.

[0122] Among them, the classification process is based on the functional attributes and physical connotations of the device parameters. For example, the parameters can be divided into power transmission types, control and regulation types, status monitoring types, etc. The power transmission type covers parameters such as rotational speed and torque, which characterize the power output and transmission characteristics of the device; the control and regulation type includes parameters such as voltage, current, and frequency, which reflect the operation control logic of the device; the status monitoring type involves parameters such as temperature, vibration amplitude, and pressure, which are used to reflect the real-time operation status of the device. By analyzing the correlation between parameters through a clustering algorithm, the systematic classification of parameters is realized.

[0123] Among them, the physical characteristic mapping relationship is constructed based on historical operation and maintenance data and domain knowledge. For example, for the power transmission type of device type, the preset mapping relationship specifies that the rotational speed parameter corresponds to the device operating speed characteristic, and the torque parameter is associated with the power output intensity characteristic. When executing, the system traverses the first device parameter set one by one, and matches the physical characteristics for the parameters of each device type according to the mapping relationship table.

[0124] Step S340, determining the operation fault data set of the target device according to the first operation data set, the first fault type data set, and the first synthetic data set.

[0125] Among them, the operation fault data set is a multi-modal data set containing the historical faults, synthetic faults of the target device and their corresponding operation parameters. This data set solves the problem of scarce fault data by fusing real data and synthetic data, and improves the generalization ability of the model to complex fault modes. The data fusion follows the principle of spatio-temporal alignment to ensure the comparability of synthetic data and real data in terms of time stamps and working conditions.

[0126] Among them, the time series signals of the first operation data set and the generated signals of the first synthetic data set are spliced according to the time series, and then the statistical characteristics and frequency domain characteristics of parameters such as vibration and temperature are extracted. Then, combined with the label information of the first fault type data set, a triple mapping relationship of fault type - characteristic parameter - time stamp is constructed.

[0127] Specifically, adjust the timestamp of the synthetic data to the time window corresponding to before the actual fault occurs. For example, align the generated bearing fault data to 10 minutes before the actual fault occurs. Then, use the Local Outlier Factor algorithm to detect and remove abnormal samples in the synthetic data to ensure data distribution consistency. Decompose the vibration signal into Intrinsic Mode Functions by Empirical Mode Decomposition, and extract the energy entropy of the Intrinsic Mode Functions as supplementary features. Then, associate the fault labels of the first fault type dataset with the fused operation data to form a labeled operation fault dataset.

[0128] Step S350, train a preset fault prediction model through the operation fault dataset to obtain a fault prediction model.

[0129] Among them, the preset fault prediction model adopts a multi-modal deep learning architecture to capture the spatial local features and time series dependencies of the equipment operation data. The model input is multi-dimensional time series data such as vibration, temperature, and pressure, and the output is the Remaining Useful Life (RUL) and the probability distribution of fault types.

[0130] Specifically, when training the fault prediction model, the input data comes from the generated synthetic fault data and the actual historical fault data. Based on the historical operation state data and synthetic data of the target equipment, a fault prediction model is constructed, which can accurately predict the RUL and fault types of the equipment. Among them, the training of the fault prediction model is mainly divided into two parts: Remaining Useful Life prediction and fault type prediction. Different tasks will select different algorithms and models. For Remaining Useful Life prediction, RUL refers to the time or number of cycles that the equipment is expected to continue working in the current operating state. RUL prediction is usually a regression problem, and the goal is to predict the numerical value of the remaining life of the equipment based on the historical data of the equipment. In RUL prediction, the regression model based on deep learning can be an LSTM network or a Gated Recurrent Unit (GRU), which is not limited here. LSTM is suitable for processing time series data, such as equipment operation data. The health state of the equipment changes over time, so LSTM can capture the long-term dependence and short-term volatility of the equipment state. The input data includes the historical operation data of the equipment (such as sensor data of temperature, pressure, vibration, etc.) and the synthetic data generated by the Generative Adversarial Network. The structure of the LSTM network generally includes multiple LSTM layers and fully connected layers, and the LSTM layer can capture the dynamic changes of the equipment state over time. The final output of the network is the predicted value of the remaining useful life of the equipment. The loss function of RUL prediction is usually the Mean Squared Error (MSE), which is defined as the average of the squared differences between the predicted value and the actual value:

[0131]

[0132] where, l RUL represents the loss function for RUL prediction, represents the predicted remaining useful life of the i-th sample, RUL i represents the true remaining useful life of the i-th sample, and N is the total number of samples.

[0133] By minimizing the loss function, the network continuously optimizes its parameters and learns how to extract features from the historical data and synthetic data of the device, so as to predict the remaining useful life of the device.

[0134] For the fault type prediction model, fault type prediction is a classification problem. The goal is to predict the possible fault types of the device based on the historical data and current state of the device. In fault type prediction, a deep neural network (DNN) or CNN is used. When dealing with time series data, CNN extracts local features through convolutional layers and can effectively identify abnormal patterns of the device. When training the model, the input data comes from historical fault data and synthetic fault data generated by a generative adversarial network. The synthetic data enhances the training set and makes up for the scarcity of actual fault data. To ensure the accuracy of the fault prediction model, methods such as cross-validation are used for model evaluation and selection. In hybrid data training, historical fault data and synthetic data jointly participate in training. Since the synthetic data can simulate various fault patterns and extreme situations, the model can learn more comprehensive fault patterns and improve the generalization ability of the model.

[0135] Step S360, input the second operation data set into the fault prediction model to obtain a fault prediction result.

[0136] Among them, the second operation data set is the real-time operation data of the target device during the operation and maintenance phase, including multi-dimensional time series signals such as vibration, temperature, and pressure. The fault prediction model adopts a multi-task learning architecture and synchronously outputs the RUL and fault type probability distribution of the target device. During model inference, the continuous time series data is segmented into sample sequences of fixed length (such as one sample every 10 seconds) through the sliding window technique, and working condition parameters (such as load and speed) are appended as auxiliary inputs to improve the prediction accuracy.

[0137] In addition, an incremental learning technique is also used to adjust and optimize the fault prediction model in real time to cope with new fault patterns or changes that may occur during the operation of the device. The operating state of the device is dynamically changing. Due to changes in the external environment, device aging, or different operating conditions, the fault pattern may change. In device fault prediction, it is assumed that there is a preliminarily trained fault prediction model M old , and new fault data D new ={(x i ,y i )} is received, where xi is the device feature vector, y i is the fault type or remaining useful life, and the existing model is incrementally updated based on new data. First, new data is acquired, and new fault data is collected in real time during the operation of the device. The new data D new will be used to update the model. Then, the existing model is trained through an incremental learning algorithm to update the model parameters. In this implementation, an incremental learning method based on gradient descent is adopted. The updated model M new is represented by the following formula:

[0138]

[0139] where M new represents the updated model, M old represents the preliminarily trained fault prediction model, η is the learning rate, and η is used to control the step size of model update. represents the gradient of the parameter θ, is the gradient of the loss function L with respect to the model parameter θ, represents the direction of optimizing the existing model on the new data.

[0140] In model evaluation, the newly incrementally learned model is evaluated to ensure that the performance of the model on new data is improved. To avoid excessive computational overhead, incremental learning often combines a memory mechanism, that is, not all historical data is used in each update, but only a part of the key data is retained for training. It can be the nearest neighbor storage method, which retains the historical samples closest to the current data. The purpose of the memory mechanism is to improve computational efficiency and reduce the dependence on a large amount of historical data. It can also be the sliding window technique. Through the sliding window, the system only retains the most recent N samples (data window). Whenever a new data sample (x new , y new ) arrives, the system updates the data window and uses the new data sample to perform incremental learning on the model.

[0141] Assume that a linear regression model is used for fault prediction, and the predicted output of the model is:

[0142]

[0143] where, represents the predicted value, w represents the weight vector, w T is the transpose of the weight vector w, x is the input feature vector, and b is the bias term.

[0144] Through incremental learning, after the model receives new data each time, it updates the parameters according to the following steps. First, calculate the gradients, that is, calculate the gradients of the loss function with respect to the model parameters w and b. Then, based on the calculated gradients, use the gradient descent method to update the model parameters. Through this process, the fault prediction model can quickly update and optimize the parameters every time it receives new data, thus ensuring that it can always adapt to the changes in the device state. For the system for formulating intelligent operation and maintenance repair strategies, the implementation of incremental learning can respond to the changes in the device state in real time during actual operation, update the fault prediction model in a timely manner, and make the prediction results more accurate. In addition, incremental learning is more efficient when dealing with large-scale data sets, and can effectively reduce the calculation and storage costs.

[0145] Step S370, generate the operation and maintenance repair strategy for the target device according to the fault prediction result, so as to dispatch the corresponding operation and maintenance personnel to carry out device operation and maintenance repair through the operation and maintenance repair strategy.

[0146] Among them, the generation of the operation and maintenance repair strategy is based on a multi-objective optimization framework, comprehensively considering factors such as the remaining useful life (RUL) of the device, maintenance cost, production plan constraints, etc., which is used to weigh the importance of different optimization objectives and can be flexibly adjusted according to device types, production scenarios, etc. The optimization objective function is defined as:

[0147] min(α·Cost + β·Downtime + γ·Risk)

[0148] Among them, α, β, and γ are weight coefficients, and α + β + γ = 1. Cost represents the maintenance cost, Downtime represents the downtime, and Risk represents the risk value of fault occurrence.

[0149] Based on the multi-objective optimization framework, the optimal maintenance plan is solved by particle swarm optimization or genetic algorithm, which is not limited here. Taking the particle swarm optimization algorithm as an example, each particle represents a maintenance plan, including decision variables such as maintenance time, maintenance content, and resource allocation. The algorithm iteratively updates the particle positions, calculates the objective function values of each plan, and gradually approaches the global optimal solution. The genetic algorithm simulates the biological evolution mechanism, performs selection, crossover, and mutation operations on the population of maintenance plans, and screens out the plans with high fitness. Through these intelligent algorithms, the optimal operation and maintenance repair strategy that takes into account cost, time, and risk can be quickly searched under complex constraint conditions, realizing the efficient allocation of device maintenance resources, reducing production losses, improving the management level of the entire device life cycle, and providing decision support for the reliable operation and maintenance of industrial devices.

[0150] In a possible embodiment, the generating the operation and maintenance repair strategy for the target device according to the fault prediction result specifically includes the following steps:

[0151] 371. Determine the areas to be repaired for the target device according to the fault prediction result, and obtain multiple areas to be repaired;

[0152] 372. Determine multiple maintenance plans for the target device according to the multiple maintenance areas;

[0153] 373. Determine the operation and maintenance strategy according to the multiple maintenance plans;

[0154] Among them, the determining of the multiple maintenance plans for the target device according to the multiple maintenance areas includes:

[0155] 3721. Determine the fault prediction probability corresponding to the target area to be repaired, and obtain the target fault prediction probability; the target area to be repaired is any one of the multiple areas to be repaired;

[0156] 3722. When the target fault prediction probability is greater than or equal to a preset threshold, obtain the target maintenance plan for the target area to be repaired from the preset standard maintenance process library; the standard maintenance process library includes standard maintenance plans formulated according to the areas of the equipment.

[0157] Among them, the determination of the area to be repaired is based on the output result of the fault prediction model. The standard maintenance process library stores maintenance plans in a hierarchical structure. Taking an industrial robot as an example, the library includes maintenance plans for joint areas (including disassembly steps and torque parameters), maintenance plans for control cabinet areas (including circuit detection processes and component replacement standards), etc. Each plan includes elements such as process flowcharts, tool lists, spare part models, and quality acceptance standards. The preset threshold is usually set to 0.7 and can be dynamically adjusted according to the importance of the equipment and the scenario.

[0158] Specifically, first analyze the structured data of the fault prediction result. For example, the fault prediction result of a certain motor shows that the probability of "abnormal stator winding temperature" is 0.88, and the probability of "excessive bearing vibration" is 0.75. Through the association between the equipment 3D model and the fault characteristics, the stator winding corresponds to the stator area of the motor, and the bearing corresponds to the drive end area, thus determining two areas to be repaired. Then, extract the fault prediction probability for each area to be repaired. Then, compare 0.88 of the stator area with the preset threshold of 0.7. Since the condition is met, the stator winding maintenance plan is retrieved from the standard maintenance process library. The plan includes: power outage and voltage verification, cut off the power supply and verify no voltage; winding disassembly, remove the stator winding according to the drawings; insulation detection, use a megohmmeter to measure the insulation resistance of the winding; repair of the fault point, locally reinforce the aging insulation layer or replace the winding; reinstallation and testing, reinstall the winding and test the no-load current.

[0159] In a possible embodiment, the determining of the operation and maintenance strategy according to the multiple maintenance plans specifically includes the following steps:

[0160] 3731. Obtain the maintenance components corresponding to each maintenance plan among the multiple maintenance plans, and obtain a plurality of maintenance components; the area to be maintained includes at least one maintenance component;

[0161] 3732. Determine the health status scores corresponding to the multiple maintenance components, and obtain a plurality of health status scores;

[0162] 3733. Sort the multiple health status scores in ascending order to obtain a health status score sequence; each maintenance component in the health status score sequence corresponds to a health status score;

[0163] 3734. Determine the maintenance plan corresponding to each maintenance component in the health status score sequence to obtain a target maintenance plan sequence;

[0164] 3735. Determine the operation and maintenance strategy according to the order of the target maintenance plan sequence.

[0165] Among them, the system first obtains real-time operation data from the equipment monitoring system, including key parameters such as temperature, vibration, pressure, current, etc., and combines the historical maintenance records and the analysis results of the predictive maintenance model. According to the results, it schedules maintenance personnel or resources for maintenance processing. In the maintenance strategy, it is necessary to ensure that high-priority tasks are executed first. At the same time, under the same priority, select the task with the shortest expected maintenance time for allocation. When new emergency tasks occur or the status of maintenance resources changes, the system recalculates the task priorities in real time and adjusts the scheduling strategy to ensure the maximization of resource utilization and the minimization of maintenance response time. During the task execution process, the system continuously tracks the maintenance progress. If the task is delayed or the maintenance fails, the system will re-evaluate the priorities and dynamically adjust the personnel and resource allocation to ensure the efficient completion of the maintenance work.

[0166] Specifically, the system parses the structured data of the maintenance plan. Then, it obtains the real-time data (such as vibration, temperature) of the maintenance components and the historical operation and maintenance records, processes the data, and calculates the health status score of each component through weighted calculation. The lower the score, the worse the health status. Among them, calculate the health status score (Health Score, HS) of the equipment. The calculation of the health status score is based on the degree of abnormality of the sensor data, and weighted summation is performed after normalization:

[0167]

[0168] Among them, HS represents the health status score, T represents the current temperature, V represents the current voltage, A represents the current current, T nominal represents the nominal temperature value (design standard temperature value), V nominalrepresents the nominal voltage value (design standard voltage value), A nominal represents the nominal current value (design standard current value), T max , V max , A max respectively represent the maximum allowable temperature, the maximum allowable voltage, and the maximum allowable current, T min , V min , A min respectively represent the minimum allowable temperature, the minimum allowable voltage, and the minimum allowable current. w1, w2, and w3 represent weight coefficients, and w1 + w2 + w3 = 1.

[0169] After the calculation is completed, according to the set interval, the health score is divided into different levels. For example: HS < 40, normal (no repair required); 40 ≤ HS < 70, slightly abnormal (low priority); 70 ≤ HS < 90, severely abnormal (medium priority); HS ≥ 90, that is, fault repaired (high priority). At the same time, the urgency level of the task (Urgency Score, US) is calculated based on production impact, safety risk, and repair difficulty:

[0170] US = w4 × P + w5 × S + w6 × D

[0171] where P represents production impact, S represents safety risk, D represents repair difficulty, and w4, w5, and w6 are the corresponding weight factors, and w4 + w5 + w6 = 1.

[0172] Combining the health status score and the urgency score, the priority of the task (Priority Score, PS) is calculated comprehensively:

[0173] PS = α × HS + β × US

[0174] where α and β are hyperparameters, and α and β are used to balance the impact of the health status and the task urgency.

[0175] Finally, according to the calculated task priority, the system dynamically adjusts the task queue to ensure that the highest-priority tasks are always at the front, and schedules and optimizes low-priority tasks to avoid over-occupying resources.

[0176] In the maintenance resource matching section, for the determined task queue, the system dynamically matches maintenance personnel and equipment resources according to the task type and priority. The allocation of maintenance personnel is calculated based on Skill Match (SM), current Workload (WL), geographical location (Distance, D), and historical experience (Experience, E). Among them, the Match Score (MS) is calculated by the matching degree between the skills of the maintenance personnel and the task requirements. The current Workload (WL) is normalized by the current number of maintenance tasks. The geographical location (Distance, D) is normalized by calculating the Euclidean distance between the task location and the current location of the maintenance personnel. The historical experience (Experience, E) is calculated by the historical maintenance success rate and the experience in handling similar tasks. In this embodiment, the specific values of the parameters are queried by querying the match score database. Among them, the specific formula for calculating the Match Score (MS) is:

[0177] MS = γ1×SM + γ2×(1 - WL) + γ3×(1 - D) + γ4×E

[0178] Among them, MS represents the match score, WL represents the current workload, D represents the geographical location, E represents the historical experience, and γ1, γ2, γ3, γ4 are hyperparameters respectively.

[0179] The allocation of maintenance equipment resources depends on the inventory situation. The system first checks whether the spare parts required for the maintenance task are sufficient. The task scheduling first ensures that high-priority tasks are executed first. At the same time, under the same priority, the task with the shortest expected maintenance time is selected for allocation:

[0180]

[0181] Among them, CurrentStock represents the current inventory quantity, RequiredStock represents the inventory demand quantity, and Availability represents the inventory availability rate. Among them, if Availability < 1, that is, the current inventory quantity is less than the inventory demand quantity, an automatic procurement process is triggered, and the task priority is adjusted or the task execution is delayed according to the inventory situation.

[0182] It should be noted that the health status scoring system supports dynamic updates. For example, if a certain type of fault occurs frequently, the weights of its related indicators are automatically increased. The sorted maintenance plan sequence can be optimized in combination with the production plan, such as arranging the maintenance tasks with longer time consumption during the production gap. In addition, the method for formulating the intelligent operation and maintenance inspection strategy can also be used and viewed by the user through the terminal, and the terminal includes a display interface.

[0183] For the sake of easy understanding, seeFigure 4 , Figure 4 is a flowchart of another method for formulating an intelligent operation and maintenance strategy provided in an embodiment of the present application. Figure 4 This paper describes the process of formulating intelligent operation and maintenance strategies. First, the temperature, vibration, pressure and other sensors deployed at the key parts of the equipment are used to collect the equipment operation parameters in real time. Then, the collected operation status data is preprocessed. In view of the problems that data in industrial scenarios are easily polluted by noise and missing due to transmission interruption, the filtering algorithm is used to remove outliers, and the interpolation method is used to fill the missing data. Then, the dimension is unified through standardization to improve the data quality. The preprocessed data is analyzed by machine learning methods to identify the type of equipment failure. Then, GAN is introduced to generate synthetic fault data. This process effectively solves the problem of insufficient actual fault data and expands the diversity of training data samples. Subsequently, the fault prediction model is trained based on the synthetic fault data and historical data for the RUL and fault type of the target equipment components. The model integrates the physical model constraints and deep learning algorithms to achieve accurate prediction of the time and type of equipment failure. In addition, the incremental learning mechanism is used to adjust the fault prediction model in real time. As the equipment runs, new data is continuously generated. Incremental learning enables the model to dynamically update parameters. If the equipment condition changes or ages, the model automatically incorporates new data for training, just like a "continuously evolving brain" to maintain adaptability to changes in equipment status and ensure long-term stability of prediction accuracy. Finally, based on the fault prediction results and equipment health status, the optimization algorithm is used to optimize the timing and content of maintenance, and combined with the maintenance strategy optimization results, maintenance personnel and equipment resources are dispatched. According to the maintenance task requirements, personnel with corresponding skills (such as mechanical maintenance workers, electrical engineers) are matched, and spare parts (such as bearings, insulating materials) and tools (such as torque wrenches, testers) are dispatched.

[0184] For easier understanding, see Figure 5 , Figure 5It is a schematic diagram of the scenario of a system for formulating intelligent operation and maintenance inspection strategies. This system constructs an intelligent operation and maintenance closed-loop of "monitoring - analysis - prediction - scheduling" through the collaborative operation of the equipment health status monitoring module, fault prediction module, maintenance scheduling module, and data processing module, realizing the comprehensive monitoring of the equipment operation status and the scientific formulation of inspection strategies. In the equipment health status monitoring module, the operation status and health level of the equipment are monitored in real time. Through sensors such as temperature, vibration, and pressure deployed at key parts of the equipment, core data such as bearing temperature, vibration amplitude of the gearbox, and pressure of the hydraulic system are continuously collected. For example, when a rotating machine is running, the module obtains vibration signals in real time and dynamically captures abnormal vibration trends, providing raw data for fault analysis. In the fault prediction module, based on the equipment historical data and real-time monitoring data, the fault type, occurrence time, and remaining service life are predicted. The module integrates algorithms such as CNN and LSTM to deeply analyze data features. For example, bearing wear is identified through the frequency domain characteristics of vibration data, and the remaining time of the fault is predicted by combining temperature and load parameters, realizing the early prediction of potential faults. In the maintenance scheduling module, according to the fault prediction results, the maintenance task scheduling and resource allocation are optimized. The module combines the production plan and the status of maintenance resources (personnel, spare parts, tools), and uses algorithms such as particle swarm optimization to generate the optimal strategy. In the data processing module, the collected data is denoised, cleaned, analyzed, and stored. In the input stage, it is filtered to remove noise, interpolated to fill in missing values, and standardized; in the analysis stage, data features are extracted for the fault prediction module to use; at the same time, historical operation, fault, and maintenance records are stored to form operation and maintenance data.

[0185] For ease of understanding, please refer to Figure 6 , Figure 6 It is an interface display diagram of an intelligent operation and maintenance inspection strategy provided by an embodiment of the present application. It can be seen that on the display interface of the terminal, through structured presentation, it provides key basis for equipment inspection and operation and maintenance strategy optimization. This interface includes main functions such as operation and maintenance inspection time, fault details list, etc. Among them, the operation and maintenance inspection time column records the specific time of the inspection operation in the format of "xx year xx month xx day xx:xx", which is used to trace the time line of equipment fault handling, analyze the fault occurrence frequency and interval rules, and provide time-dimensional data for equipment health status assessment; and as part of the work record of the operation and maintenance team, it is used to assess the fault response timeliness and improve the operation and maintenance efficiency. The fault details list presents specific faults in tabular form. Among the two groups of records of "Error3 motor temperature anomaly" and "Error7 temperature sensor anomaly", the fault code and text description are integrated. The recording form of the fault codes "Error3" and "Error7" facilitates the quick retrieval and statistics of fault types by the equipment management system, and also helps the operation and maintenance personnel to match similar cases in the historical fault library and reuse existing solutions, realizing the orderly management of equipment fault information.

[0186] For ease of understanding, please refer toFigure 7 , Figure 7 This is the interface display diagram of another intelligent operation and maintenance inspection strategy formulation system provided by the embodiments of the present application. It can be seen that this interface display presents the core plan of equipment maintenance in a structured form, covering the setting of maintenance time and the specific operation process. It is a standardized document for guiding maintenance personnel to perform maintenance tasks, aiming to ensure the orderly and efficient development of maintenance work and improve the standardization and accuracy of equipment fault handling. Among them, this interface includes: the core part of the start time of maintenance and specific steps. The start time of maintenance is a key node for the time management of maintenance tasks, and the specific steps part provides specific operation and maintenance inspection strategies for users to schedule around fault handling.

[0187] It can be seen that through an intelligent operation and maintenance inspection strategy formulation method, by analyzing the historical data and real-time data of equipment, predicting possible future faults of the equipment, and being able to perform maintenance in advance, sudden equipment failures can be avoided, thereby improving the efficiency of operation and maintenance inspection; in addition, by innovatively using a generative adversarial network model to generate new fault data based on the historical operation fault data of the equipment and training the model, the generalization performance of the fault prediction model is improved, the reliability of model prediction is ensured, and further the reliability of equipment operation and maintenance inspection is improved.

[0188] The above mainly introduces the solutions of the embodiments of the present application from the perspective of the execution process on the method side. It can be understood that in order for the server to implement the above functions, it includes the corresponding hardware structure and / or software module for executing each function. Those skilled in the art should easily realize that, combined with the units and algorithm steps of each example described in the embodiments provided in this article, the present application can be implemented in the form of hardware or a combination of hardware and computer software.

[0189] The embodiments of the present application can divide the server into functional units according to the above method examples. For example, each functional unit can be divided corresponding to each function, or two or more functions can be integrated into one processing unit. The above integrated unit can be implemented in the form of hardware or in the form of a software functional unit. It should be noted that the division of units in the embodiments of the present application is illustrative, only a logical function division, and there may be other division methods in actual implementation.

[0190] In the case of dividing each functional module corresponding to each function, Figure 8 This is the block diagram of the functional unit composition of an intelligent operation and maintenance inspection strategy formulation device provided by the embodiments of the present application. The intelligent operation and maintenance inspection strategy formulation device 800 includes:

[0191] An acquisition unit 810, configured to acquire a first operation data set of a target device within a historical time period, and a second operation data set of the target device within an operation and maintenance time period; a start time of the operation and maintenance time period is later than an end time of the historical time period;

[0192] A determination unit 820, configured to determine a fault type corresponding to the target device according to the first operation data set, and obtain a first fault type data set;

[0193] The determination unit 820 is further configured to determine a first synthetic data set according to the first operation data set and the first fault type data set;

[0194] The determination unit 820 is further configured to determine an operation fault data set of the target device according to the first operation data set, the first fault type data set, and the first synthetic data set;

[0195] A calculation unit 830, configured to train a preset fault prediction model through the operation fault data set to obtain a fault prediction model;

[0196] The calculation unit 830 is configured to input the second operation data set into the fault prediction model to obtain a fault prediction result;

[0197] A control unit 840, configured to generate an operation and maintenance repair strategy for the target device according to the fault prediction result, so as to dispatch corresponding operation and maintenance personnel for device operation and maintenance repair through the operation and maintenance repair strategy.

[0198] In a possible embodiment, in terms of determining the fault type corresponding to the target device during operation according to the first operation data set and obtaining a first fault type data set, the determination unit 820 is specifically configured to:

[0199] Perform data preprocessing on the first operation data set to obtain a target data set;

[0200] Perform feature selection on the target data set to obtain a fault feature set;

[0201] Input the fault feature set into a preset machine learning model, and output a fault component of the target device through the machine learning model to obtain a fault component set;

[0202] Extract operation parameters corresponding to each component in the fault component set from the first operation data set to obtain a plurality of operation parameters;

[0203] Extract fault types corresponding to the plurality of operation parameters from a preset operation parameter and fault type database to obtain at least one fault type, and determine the first fault type data set according to the at least one fault type.

[0204] In a possible embodiment, the determining unit 820 is specifically configured to determine a first synthetic data set according to the first operation data set and the first fault type data set as follows:

[0205] Obtain the input format of a preset generative adversarial network model to obtain a first input format;

[0206] Based on the first operation data set, determine the operation data corresponding to each fault in the first fault type data set to obtain a fault operation association data set;

[0207] Convert the fault operation association data set according to the first input format to obtain a first conversion data set;

[0208] Optimize the generative adversarial network model according to a preset first regularization formula to obtain a target generative adversarial network model;

[0209] Input the first conversion data set into the target generative adversarial network model to obtain the first synthetic data set.

[0210] In a possible embodiment, the determining unit 820 is specifically configured to optimize the generative adversarial network model according to a preset first regularization formula to obtain a target generative adversarial network model as follows:

[0211] Obtain a first device parameter set of the target device;

[0212] Determine the physical characteristics corresponding to each device parameter in the first device parameter set to obtain n physical characteristics; n is an integer greater than 1;

[0213] Generate n physical models according to the n physical characteristics; each physical model in the n physical models represents a physical operation characteristic of a certain aspect of the target device;

[0214] Determine the objective function of the generative adversarial network model according to the first regularization formula and the n physical models;

[0215] Optimize the generative adversarial network model based on a preset optimization algorithm and the objective function to obtain a target generative adversarial network model.

[0216] In a possible embodiment, the determining unit 820 is specifically configured to determine the physical characteristics corresponding to each device parameter in the first device parameter set to obtain n physical characteristics as follows:

[0217] Classify all device parameters in the first device parameter set to obtain m device types; m is an integer greater than 1; the device types are used to represent the functional types of the target device;

[0218] Determine the physical characteristics corresponding to each device type among the multiple device types according to the preset physical characteristic mapping relationship and the first device parameter set, and obtain n physical characteristics; the physical characteristic mapping relationship includes the association relationship between the device parameters under different device types and the corresponding physical characteristics.

[0219] In a possible embodiment, the control unit 840 is specifically configured to generate an operation and maintenance repair strategy for the target device according to the fault prediction result:

[0220] Determine the areas to be repaired for the target device according to the fault prediction result, and obtain a plurality of areas to be repaired;

[0221] Determine a plurality of repair plans for the target device according to the plurality of repair areas;

[0222] Determine the operation and maintenance repair strategy according to the plurality of repair plans;

[0223] Among them, determining the plurality of repair plans for the target device according to the plurality of repair areas includes:

[0224] Determine the fault prediction probability corresponding to the target area to be repaired, and obtain the target fault prediction probability; the target area to be repaired is any one of the plurality of areas to be repaired;

[0225] When the target fault prediction probability is greater than or equal to a preset threshold, obtain the target repair plan for the target area to be repaired from the preset standard repair process library; the standard repair process library includes standard repair plans formulated according to the areas of the device.

[0226] In a possible embodiment, the control unit 840 is specifically configured to determine the operation and maintenance repair strategy according to the plurality of repair plans:

[0227] Obtain the repair components corresponding to each repair plan among the plurality of repair plans, and obtain a plurality of repair components; the area to be repaired includes at least one repair component;

[0228] Determine the health status scores corresponding to the plurality of repair components, and obtain a plurality of health status scores;

[0229] Sort the plurality of health status scores in ascending order to obtain a health status score sequence; each repair component in the health status score sequence corresponds to a health status score;

[0230] Determine the repair plan corresponding to each repair component in the health status score sequence, and obtain a target repair plan sequence;

[0231] The operation and maintenance strategy is determined according to the order of the target maintenance plan sequence.

[0232] It should be noted that the specific functional implementation of the intelligent operation and maintenance strategy formulation device 800 is shown in the above Figure 3 The description of a method for formulating an intelligent operation and maintenance strategy is shown, for example, the acquisition unit 810 is used to implement the relevant content of executing S310, and the calculation unit 830 is used to implement the relevant content of executing S350 and S360, which will not be described in detail. Each unit or module in a device 800 for formulating an intelligent operation and maintenance strategy can be separately or completely combined into one or several other units or modules to form, or one (some) of the units or modules can be further divided into multiple functionally smaller units or modules to form, which can achieve the same operation without affecting the realization of the technical effects of the embodiments of the present invention. The above-mentioned units or modules are divided based on logical functions. In actual applications, the function of one unit (or module) is implemented by multiple units (or modules), or the function of multiple units (or modules) is implemented by one unit (or module).

[0233] It can be seen that the present embodiment describes a device for formulating an intelligent operation and maintenance strategy. By analyzing the historical data and real-time data of the equipment, possible future failures of the equipment are predicted, and maintenance can be performed in advance to avoid sudden equipment failures, thereby improving the efficiency of operation and maintenance and ensuring the reliability of model predictions, thereby improving the reliability of equipment operation and maintenance.

[0234] An embodiment of the present application also provides a system for formulating an intelligent operation and maintenance strategy, wherein the system for formulating an intelligent operation and maintenance strategy is used to execute part or all of the steps of any method recorded in the above-mentioned embodiment of the method for formulating an intelligent operation and maintenance strategy, and the above-mentioned system for formulating an intelligent operation and maintenance strategy can be applied to a server.

[0235] An embodiment of the present application also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program for electronic data exchange, wherein the computer program enables a computer to execute part or all of the steps of any method recorded in the above method embodiments, and the above computer includes a server.

[0236] The present application also provides a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program, and the computer program is operable to cause a computer to execute some or all of the steps of any method described in the method embodiment. The computer program product may be a software installation package, and the computer includes a server.

[0237] It should be noted that, for the above-described embodiments, for the sake of simple description, they are all expressed as a series of action combinations. Those skilled in the art should be aware that the present application is not limited by the described order of actions, because some steps in the embodiments of the present application can be performed in other orders or simultaneously. In addition, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions, steps, modules or units involved are not necessarily essential to the embodiments of the present application.

[0238] In the above embodiments, the descriptions of the embodiments of the present application each have their own focuses. For the parts not detailed in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0239] Those of ordinary skill in the art can understand all or part of the processes in the methods of the above embodiments. These processes can be completed by relevant hardware instructed by a computer program. The program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the above method embodiments. The aforementioned storage media include: ROM or random access memory (RAM), magnetic disks, or optical discs and other media that can store program codes.

[0240] Those skilled in the art should be able to realize that in the above one or more examples, the functions described in the embodiments of the present application can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the processes or functions described in the embodiments of the present application are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from a website, computer, server, or data center to another website, computer, server, or data center in a wired manner (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or wirelessly (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that the computer can access or a data storage device such as a server or data center that includes one or more integrated available media. The available medium can be a magnetic medium (such as a floppy disk, hard disk, magnetic tape), an optical medium (such as a digital video disc (DVD)), or a semiconductor medium (such as a solid state disk (SSD)), etc.

[0241] Each device and product described in the above embodiments, and each module / unit included therein, can be a software module / unit, a hardware module / unit, or can be partially a software module / unit and partially a hardware module / unit. For example, for each device and product applied to or integrated into a chip, each module / unit included therein can be implemented in a hardware manner such as a circuit, or at least some of the modules / units can be implemented in the form of a software program that runs on a processor integrated inside the chip, and the remaining (if any) part of the modules / units can be implemented in a hardware manner such as a circuit; for each device and product applied to or integrated into a chip module, each module / unit included therein can be implemented in a hardware manner such as a circuit, and different modules / units can be located in the same component (such as a chip, a circuit module, etc.) or different components of the chip module, or at least some of the modules / units can be implemented in the form of a software program that runs on a processor integrated inside the chip module, and the remaining (if any) part of the modules / units can be implemented in a hardware manner such as a circuit; for each device and product applied to or integrated into a terminal device, each module / unit included therein can be implemented in a hardware manner such as a circuit, and different modules / units can be located in the same component (such as a chip, a circuit module, etc.) or different components inside the terminal device, or at least some of the modules / units can be implemented in the form of a software program that runs on a processor integrated inside the terminal device, and the remaining (if any) part of the modules / units can be implemented in a hardware manner such as a circuit.

[0242] The specific embodiments described above further elaborate on the objectives, technical solutions, and beneficial effects of the embodiments of the present application. It should be understood that the above is only the specific embodiments of the embodiments of the present application and is not used to limit the protection scope of the embodiments of the present application. Any modifications, equivalent replacements, improvements, etc. made on the basis of the technical solutions of the embodiments of the present application shall be included in the protection scope of the embodiments of the present application.

Claims

1. A method for formulating an intelligent operation and maintenance inspection strategy, characterized in that Applied to a server, the method includes: Obtain a first operation data set of a target device within a historical time period, and a second operation data set of the target device within an operation and maintenance time period; the start time of the operation and maintenance time period is later than the end time of the historical time period; Determine the fault type corresponding to the target device according to the first operation data set to obtain a first fault type data set; Determine a first synthetic data set according to the first operation data set and the first fault type data set; Determine an operation fault data set of the target device according to the first operation data set, the first fault type data set, and the first synthetic data set; Train a preset fault prediction model through the operation fault data set to obtain a fault prediction model; Input the second operation data set into the fault prediction model to obtain a fault prediction result; Generate an operation and maintenance repair strategy for the target device according to the fault prediction result, so as to dispatch corresponding operation and maintenance personnel for device operation and maintenance repair through the operation and maintenance repair strategy.

2. The method according to claim 1, wherein The determining the fault type corresponding to the target device during operation according to the first operation data set to obtain a first fault type data set includes: Perform data preprocessing on the first operation data set to obtain a target data set; Perform feature selection on the target data set to obtain a fault feature set; Input the fault feature set into a preset machine learning model, and output the faulty components of the target device through the machine learning model to obtain a faulty component set; Extract the operation parameters corresponding to each component in the faulty component set from the first operation data set to obtain a plurality of operation parameters; Extract the fault types corresponding to the plurality of operation parameters from a preset operation parameter and fault type database to obtain at least one fault type, and determine the first fault type data set according to the at least one fault type.

3. The method according to claim 1, characterized in that, The determining a first synthetic data set according to the first operation data set and the first fault type data set includes: Obtain a first input format of a preset generative adversarial network model; Determine the operation data corresponding to each fault in the first fault type data set based on the first operation data set to obtain a fault operation association data set; Convert the fault operation association data set according to the first input format to obtain a first conversion data set; Optimize the generative adversarial network model according to a preset first regularization formula to obtain a target generative adversarial network model; Input the first conversion data set into the target generative adversarial network model to obtain the first synthetic data set.

4. The method according to claim 3, wherein The optimizing the generative adversarial network model according to a preset first regularization formula to obtain a target generative adversarial network model includes: Obtain a first device parameter set of the target device; Determine the physical characteristics corresponding to each device parameter in the first device parameter set to obtain n physical characteristics; n is an integer greater than 1; Generate n physical models according to the n physical characteristics; each of the n physical models represents the physical operation characteristics of a certain aspect of the target device; Determine the objective function of the generative adversarial network model according to the first regularization formula and the n physical models; Optimize the generative adversarial network model based on a preset optimization algorithm and the objective function to obtain a target generative adversarial network model.

5. The method according to claim 4, characterized in that, The determining the physical characteristics corresponding to each device parameter in the first device parameter set to obtain n physical characteristics includes: Classify all device parameters in the first device parameter set to obtain m device types; m is an integer greater than 1; the device type is used to represent the functional type of the target device; Determine the physical characteristics corresponding to each device type in the multiple device types according to a preset physical characteristic mapping relationship and the first device parameter set to obtain n physical characteristics; the physical characteristic mapping relationship includes the association relationship between the device parameters under different device types and the corresponding physical characteristics.

6. The method according to any one of claims 1-5, characterized in that, The generating the operation and maintenance repair strategy of the target device according to the fault prediction result includes: Determine the areas to be repaired of the target device according to the fault prediction result to obtain multiple areas to be repaired; Determine multiple repair plans for the target device according to the multiple repair areas; Determine the operation and maintenance repair strategy according to the multiple repair plans; Wherein, the determining the multiple repair plans for the target device according to the multiple repair areas includes: Determine the fault prediction probability corresponding to the target area to be repaired to obtain a target fault prediction probability; the target area to be repaired is any one of the multiple areas to be repaired; When the target fault prediction probability is greater than or equal to a preset threshold, obtain the target repair plan for the target area to be repaired from a preset standard repair procedure library; the standard repair procedure library includes standard repair plans formulated according to the areas of the device.

7. The method according to claim 6, wherein The determining the operation and maintenance repair strategy according to the multiple repair plans includes: Obtain the repair components corresponding to each repair plan in the multiple repair plans to obtain multiple repair components; the area to be repaired includes at least one repair component; Determine the health status scores corresponding to the multiple repair components to obtain multiple health status scores; Sort the multiple health status scores in ascending order to obtain a health status score sequence; each repair component in the health status score sequence corresponds to a health status score; Determine the repair plan corresponding to each repair component in the health status score sequence to obtain a target repair plan sequence; Determine the operation and maintenance repair strategy according to the order of the target repair plan sequence.

8. An apparatus for formulating an intelligent operation and maintenance inspection strategy, characterized in that, Applied to a server, the device includes: An acquisition unit, configured to acquire a first operation data set of a target device in a historical time period, and a second operation data set of the target device in an operation and maintenance time period; the start time of the operation and maintenance time period is later than the end time of the historical time period; A determination unit, configured to determine the fault type corresponding to the target device according to the first operation data set to obtain a first fault type data set; The determination unit is further configured to determine a first synthetic data set according to the first operation data set and the first fault type data set; The determining unit is further configured to determine an operation fault data set of the target device according to the first operation data set, the first fault type data set, and the first synthetic data set; The calculation unit is configured to train a preset fault prediction model through the operation fault data set to obtain a fault prediction model; The calculation unit is configured to input the second operation data set into the fault prediction model to obtain a fault prediction result; The control unit is configured to generate an operation and maintenance repair strategy for the target device according to the fault prediction result, so as to dispatch corresponding operation and maintenance personnel for device operation and maintenance repair through the operation and maintenance repair strategy.

9. A server, characterized in that, including: a processor, a memory, a communication interface, and one or more programs; The one or more programs are stored in the memory and configured to be executed by the processor, and the programs include instructions for performing the steps in the method according to any one of claims 1-7.

10. A system for formulating an intelligent operation and maintenance inspection strategy, characterized in that, The system for formulating the intelligent operation and maintenance repair strategy is configured to execute the method according to any one of claims 1-7.

Images