Operation and maintenance work order management method and system based on artificial intelligence

Abstract

The invention discloses an artificial intelligence-based operation and maintenance work order management method and system, and relates to the technical field of artificial intelligence and operation and maintenance management, and the method comprises the steps: constructing a CNN model, optimizing a strategy network and a value network through a PPO algorithm, optimizing model parameters through multi-task learning and gradient updating, outputting an equipment fault probability and a recommendation action, and generating a JSON work order. Based on a classic GBDT framework, a quantum decision tree is used for performing split gain calculation, a loss function is optimized through a quantum Monte Carlo method, the quantum decision tree is iteratively constructed, and a candidate scheduling scheme is output. According to the method, the CNN model is combined with the PPO algorithm to optimize the generation of the work order, the accuracy and timeliness of fault prediction and action recommendation are enhanced, quantum machine learning is introduced into work order scheduling modeling, and the multi-target adaptive capacity of scheduling is improved through a quantum decision tree and a quantum Monte Carlo optimization mechanism.

Description

Technical Field

[0001] The present invention relates to the technical fields of artificial intelligence and operation and maintenance management, and particularly to an operation and maintenance work order management method and system based on artificial intelligence. Background Art

[0002] With the rapid progress of artificial intelligence, big data, and Internet of Things technologies, the operation and maintenance management of equipment has become increasingly complex. Especially in some high-risk industries, the operating status of valves directly affects the safety and stability of the entire production system. Timely maintenance and fault prediction of valves are particularly important. Traditional equipment operation and maintenance management usually rely on manual inspections and manual recording of work orders, which are inefficient and prone to omissions or errors. Equipment operation and maintenance are gradually moving towards intelligence and automation. Through sensor monitoring, data collection and analysis, combined with machine learning and deep learning technologies, the system can monitor the equipment status in real time, automatically generate fault work orders and make predictions.

[0003] However, there are still some problems in the existing operation and maintenance work order management. The existing models cannot fully extract the deep correlation features between valve image, voice, text, and time series data, resulting in limited fault prediction accuracy. They fail to be optimized by combining quantum computing and do not effectively utilize the advantages of qubit parallel search and quantum state probability distribution, which limits the further improvement of scheduling accuracy and algorithm convergence speed. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides an operation and maintenance work order management method and system based on artificial intelligence, which solves the problem that the existing models cannot fully extract the deep correlation features between valve image, voice, text, and time series data, resulting in limited fault prediction accuracy, fails to be optimized by combining quantum computing, and does not effectively utilize the advantages of qubit parallel search and quantum state probability distribution, which limits the further improvement of scheduling accuracy and algorithm convergence speed.

[0006] To solve the above technical problems, the present invention provides the following technical solutions: In a first aspect, the present invention provides an operation and maintenance work order management method based on artificial intelligence, which includes collecting multimodal data, generating a multimodal work order data packet, constructing a CNN model, using the PPO algorithm to optimize the policy network and value network, optimizing the model parameters through multi-task learning and gradient update, outputting the device failure probability and recommended actions, generating a JSON work order, using the IBM Qiskit framework for quantum state conversion, based on the classical GBDT framework, using a quantum decision tree for splitting gain calculation, and optimizing the loss function through the quantum Monte Carlo method, iteratively constructing a quantum decision tree, outputting a candidate scheduling plan, slightly perturbing the current optimized objective function value through the simulated annealing algorithm, outputting an optimal scheduling plan, generating task instructions through IMS, and executing them, and storing the multimodal data generated by collection and analysis.

[0007] As a preferred solution of the operation and maintenance work order management method based on artificial intelligence according to the present invention, wherein: the collecting multimodal data, generating a multimodal work order data packet, constructing a CNN model, using the PPO algorithm to optimize the policy network and value network, optimizing the model parameters through multi-task learning and gradient update, outputting the device failure probability and recommended actions, and generating a JSON work order include: The multimodal data includes valve pressure values, valve images, and voice data; Construct a CNN model, collect historical multimodal data, use EfficientNet-B0 to extract valve image features, and generate a valve image feature vector through global average pooling; Extract the features of the valve pressure value through time series analysis technology to generate a valve pressure value feature vector, and use DistilBERT to encode the log text to generate a text feature vector; Fuse the valve image feature vector, text feature vector, and valve pressure value feature vector through an attention mechanism to generate a multimodal feature vector, output the number of fault categories through a fully connected layer, and output the fault category probability through a Softmax layer; Define the state as the concatenation of the multimodal feature vector and the fault probability, and define the discrete action space as no operation, repair, and replacement; Define the policy network using a feedforward neural network, input the state, and output the action probability distribution of the current state through two fully connected layers. Set the action selected according to the current state as ; Define the value network using an MLP, input the multimodal feature vector, and output the value estimate of the current state through two fully connected layers; Set the reward function based on the value network, calculate the reward value, optimize the policy network and value network through the PPO algorithm, and calculate the goodness or badness value of the fault category for the current state and the policy probability ratio between the current time t and time t-1; Calculate the policy gradient value; Set each fault category as a task, use the Adam optimizer, update the parameters of the policy network and value network through the policy gradient value and value estimation, update the parameters of the CNN model by calculating the gradient value of the loss function, set the maximum number of iterations using the rule of thumb, and stop the iteration when the maximum iteration is reached; Input the multi-modal data into the optimized CNN model, output the device failure probability through Softmax, output the recommended action through the policy network, package the output results of the CNN model and the priority scores into a JSON message, and integrate them into a JSON format task unit through Schema standardization in combination with the multi-modal work order data packet.

[0008] As a preferred solution of the operation and maintenance work order management method based on artificial intelligence according to the present invention, wherein: the IBM Qiskit framework is used for the conversion of quantum states, based on the classical GBDT framework, the quantum decision tree is used for splitting gain calculation, and the loss function is optimized by the quantum Monte Carlo method, and the quantum decision tree is iteratively constructed to output the candidate scheduling scheme, including: Query the current technician data through the API interface, including the technician skill level and the technician available time; Use timestamp standardization to convert the task unit and technician data into a unified time format, perform time synchronization through the time convolutional network, and generate a time synchronization feature vector through the cross splicing method; Use the IBM Qiskit framework to match the time features of each pair of technician-task units in the time synchronization feature vector, and convert the time matching features into quantum states through amplitude encoding; Allocate a qubit to each pair of technician-task units to generate a quantum state data set; Based on the gradient boosting framework of classical GBDT, initialize the first quantum decision tree, use the average matching degree of all samples as the initial prediction value, and calculate the initial residual; Apply a parameterized quantum gate to the quantum state data set, calculate the splitting gain of each pair of technician-task units, extract the splitting gain using quantum measurement, and set the number of measurements according to the Qiskit default value; Select the pair of technician-task units with the largest splitting gain as the splitting point, and output the splitting point and the corresponding quantum state data set of the splitting point; Define the mean square error loss function, use the quantum Monte Carlo method to encode the initial residual into a residual quantum state, simulate a random walk through the quantum circuit, adjust the expected value of the mean square error loss function, and use the COBYLA optimizer to optimize the parameters of the quantum circuit, and stop the optimization when the expected value reaches the minimum; Based on the gradient boosting idea of classical GBDT, multiple quantum decision trees are iteratively constructed. The predicted values of trees and the residuals are input into the th tree. The outputs of all trees are weighted and accumulated to output the final predicted value, and the final predicted value is converted into a matching probability; The matching probabilities are sorted in descending order, and the top E pairs are selected. Chromosomes are generated through encoding, and each chromosome is a complete scheduling plan; Based on empirical analysis, the weights of the response time score, resource utilization score, and GBDT prediction score are set, and a multi-objective fitness function is defined; The fitness value of each chromosome is calculated. Chromosomes with high probability values are selected through roulette wheel selection, and the next generation of chromosomes is generated through crossover and mutation. The iterative operation is repeated and stopped when the fitness value converges, and the chromosomes of the last iteration are output and formatted into a list of candidate scheduling plans.

[0009] As a preferred solution of the operation and maintenance work order management method based on artificial intelligence according to the present invention, the multi-modal data collected is first preprocessed, and the multi-modal work order data packet is generated, including: A digital twin model is constructed, the preprocessed valve pressure value is input, the current stress value is calculated, and the stress value at future time t+1 is predicted through a linear regression formula. The safety threshold is set by a statistical method, and the predicted stress value at future time t+1 less than the safety threshold is selected and marked as abnormal; The valve image, text, and current stress value are calculated through a Transformer encoder and weighted fusion to generate a comprehensive feature vector; The Q table is initialized, the actions are set as generating a work order and not generating a work order, and the reward value of the current state is calculated; According to the Q-learning algorithm, the Q value is updated. The Q value of generating a work order is selected to be greater than the Q value of not generating a work order. The comprehensive feature vector, location, timestamp, and predicted state are extracted to form a work order data packet, which is serialized in JSON format, and the multi-modal work order data packet is output.

[0010] As a preferred solution of the operation and maintenance work order management method based on artificial intelligence according to the present invention, the current optimization objective function value is slightly perturbed through the simulated annealing algorithm to output the optimal scheduling plan, including: The optimization objective function is defined, and the initial temperature is set based on the analysis of the problem characteristics, which is the starting stage of the simulated annealing process; The optimization objective function value of the current plan is calculated as the starting point of the simulated annealing process. The current solution is slightly perturbed to generate a new solution, and the new objective function value is recalculated; It is judged whether to accept the new solution according to the difference of the optimization objective function, and the accepted new solution is updated as the current solution; After each iteration, the initial temperature decays according to the temperature reduction strategy, and the iteration continues. When the temperature drops to the lowest value, the iteration stops, and the current solution is output to generate the optimal scheduling plan.

[0011] As a preferred solution of the operation and maintenance work order management method based on artificial intelligence according to the present invention, wherein: generating a task instruction through IMS and executing it includes: Generating a task instruction from the optimal scheduling plan through the IMS system, pushing it to the technician terminal, and the technician executes the task instruction.

[0012] As a preferred solution of the operation and maintenance work order management method based on artificial intelligence according to the present invention, wherein: storing the multi-modal data generated by collection and analysis includes: It means storing the collected multi-modal data and the optimal scheduling plan generated by analysis in a database. The database marks the stored data with a timestamp and uploads the stored data to the cloud for backup.

[0013] In a second aspect, the present invention provides an operation and maintenance work order management system based on artificial intelligence, including, A collection and decision-making module for collecting multi-modal data, generating a multi-modal work order data packet, constructing a CNN model, optimizing the policy network and value network using the PPO algorithm, optimizing the model parameters through multi-task learning and gradient update, outputting the device failure probability and recommended actions, and generating a JSON work order; A scheduling optimization module for performing quantum state conversion using the IBM Qiskit framework, calculating the split gain using a quantum decision tree based on the classical GBDT framework, and optimizing the loss function through the quantum Monte Carlo method, iteratively constructing a quantum decision tree, and outputting a candidate scheduling plan; A global adjustment module for slightly perturbing the current optimized objective function value through the simulated annealing algorithm and outputting the optimal scheduling plan; An execution and storage module for generating a task instruction through IMS and executing it, and storing the multi-modal data generated by collection and analysis.

[0014] In a third aspect, the present invention provides a computer device including a memory and a processor, wherein: when the computer program stored in the memory is executed by the processor, any step of the operation and maintenance work order management method based on artificial intelligence as described in the first aspect of the present invention is implemented.

[0015] In a fourth aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored, wherein: when the computer program is executed by the processor, any step of the operation and maintenance work order management method based on artificial intelligence as described in the first aspect of the present invention is implemented.

[0016] The beneficial effects of the present invention are as follows: The present invention collects multi-modal data, generates multi-modal work order data packets, constructs a CNN model, uses the PPO algorithm to optimize the policy network and value network, optimizes the model parameters through multi-task learning and gradient update, outputs the device failure probability and recommended actions, generates JSON work orders, uses the IBM Qiskit framework for quantum state conversion, based on the classical GBDT framework, uses quantum decision trees for split gain calculation, and optimizes the loss function through the quantum Monte Carlo method, iteratively constructs quantum decision trees, and outputs candidate scheduling schemes; enhancing the accuracy and timeliness of fault prediction and action recommendation, and improving the multi-objective adaptation ability of scheduling. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0018] Figure 1 It is a flowchart of the operation and maintenance work order management method based on artificial intelligence in Embodiment 1.

[0019] Figure 2 It is a schematic diagram of the operation and maintenance work order management system based on artificial intelligence in Embodiment 1.

[0020] Figure 3 It is a flowchart of generating multi-modal feature vectors in Embodiment 1.

[0021] Figure 4 It is a flowchart of outputting the optimal scheduling scheme in Embodiment 1. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0022] In order to make the above objects, features, and advantages of the present invention more obvious and understandable, the following will provide a detailed description of the specific embodiments of the present invention in conjunction with the drawings in the specification.

[0023] Many specific details are set forth in the following description in order to provide a thorough understanding of the present invention. However, the present invention may be implemented in other ways different from those described herein. Those skilled in the art can make similar extensions without departing from the spirit of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0024] Second, the "one embodiment" or "embodiment" referred to herein means a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The appearances of "in one embodiment" in different places in this specification do not all refer to the same embodiment, nor are they separate or alternative embodiments that are mutually exclusive of other embodiments.

[0025] Embodiment 1, referring to Figures 1 to 4 , is the first embodiment of the present invention. This embodiment provides an operation and maintenance work order management method based on artificial intelligence, including the following steps: S1. Collect multimodal data, generate a multimodal work order data packet, construct a CNN model, optimize the policy network and value network using the PPO algorithm, optimize the model parameters through multi-task learning and gradient update, output the device failure probability and recommended actions, and generate a JSON work order; Specifically, collecting multimodal data and generating a multimodal work order data packet includes: The intelligent sensors include a pressure sensor, an infrared camera, a voice recognition module, and a microphone; The multimodal data includes valve pressure values, valve images, and voice data; The preprocessing includes converting the voice data into text using the voice recognition module, normalizing the valve pressure values using linear normalization, and performing grayscale conversion and scaling on the valve images using the weighted average method and linear bi-interpolation respectively; Construct a digital twin model, including an input layer, a model layer, a simulation layer, an analysis layer, and an output layer; Collect historical valve pressure values, perform normalization processing, and train the digital twin model; Input the preprocessed valve pressure values, calculate the current stress value, and predict the stress value at future time t + 1 through a linear regression formula. The output formula is: , , Wherein, is the stress value at the current time , is the normalized valve pressure value, is the valve area (provided by the manufacturer), is the time 's predicted value, ΔW is the stress change rate within the historical time t (obtained by collecting historical valve stress values using intelligent sensors and calculating the average change rate); Set a safety threshold through the statistical method, screen the predicted stress values at future time t + 1 that are less than the safety threshold, and mark them as abnormal; Output the digital twin status data, including the current stress value, predicted stress value, and anomaly flag; Calculate the comprehensive feature vector by passing the valve image, text, and current stress value through the Transformer encoder and weighted fusion; Initialize the Q-table, set the actions as generating a work order and not generating a work order, and calculate the reward value for the current state. The formula is: J , where J is the reward value for the current state, is the intensity value of the comprehensive feature vector (calculated through the L2 norm), is the predicted stress severity value (obtained by comparing the current stress value with the maximum stress value), and are the weights of the intensity value of the comprehensive feature vector and the weight of the predicted stress severity value (both weights are obtained by the entropy weight method); Update the Q value according to the Q-learning algorithm, select the work order generation Q value greater than the non-generation Q value, extract the comprehensive feature vector, location, timestamp, and predicted state to form a work order data packet, and serialize it in JSON format to output the multi-modal work order data packet.

[0026] By collecting multi-modal data, improve the comprehensiveness of work order analysis, avoid the limitations of a single data source, calculate the current stress value through the digital twin model, and predict the future stress value to avoid losses caused by over-maintenance or sudden failures. By generating a comprehensive feature vector, provide a better state representation for the follow-up.

[0027] Furthermore, build a CNN model, optimize the policy network and value network using the PPO algorithm, optimize the model parameters through multi-task learning and gradient update, output the device failure probability and recommended actions, and generate a JSON work order, including: Build a CNN model, including an input layer, convolutional layer, fully connected layer, and output layer; Collect historical multi-modal data, input it into the CNN model, extract the valve image features through EfficientNet-B0, and perform global average pooling on the extracted valve image features to generate a valve image feature vector; Extract the features of the valve pressure value through time series analysis technology to generate a valve pressure value feature vector; Encode the log text using DistilBERT to generate a text feature vector; Fuse the valve image feature vector, text feature vector, and valve pressure value feature vector through the attention mechanism to generate a multi-modal feature vector. Pass it through a fully connected layer to output the number of fault categories, and pass it through a Softmax layer to output the fault category probability; Define the state as the concatenation of the multi-modal feature vector and the fault probability, and define the discrete action space as no operation, repair, and replacement; Use a feed-forward neural network to define the policy network. Input the state, pass it through two fully connected layers, and output the action probability distribution of the current state. Set the action selected according to the current state to be ; Use an MLP to define the value network. Input the multi-modal feature vector, pass it through two fully connected layers, and output the value estimate of the current state; Set the reward function based on the value network and calculate the reward value. The formula is: , where, is the reward value at the current time , is the discount factor (historical training and verification), is the current time of the state of the value estimate, is the immediate reward value (set according to the discrete action space); Generate the priority score by normalizing the reward value; Optimize the policy network and the value network through the PPO algorithm, and calculate the goodness value of the fault category for the current state. The formula is: , where, is the goodness value at the current time ; Calculate the policy probability ratio of the current time t and the time -1 , and the formula is: , where, is the probability distribution of the policy at the current time , is the probability distribution of the policy at the time -1; Calculate the policy gradient value. The formula is: , where, is the policy gradient value, is the policy probability ratio of the current time t and the time -1, is a constant; Set each fault category as a task, use the Adam optimizer, update the parameters of the policy network and the value network through the policy gradient value and value estimation, update the parameters of the CNN model by calculating the gradient value of the loss function, set the maximum number of iterations using the rule of thumb, and stop iterating when the maximum iteration is reached; Input the multimodal data into the optimized CNN model, output the device failure probability through Softmax, and output the recommended action through the policy network; Package the output result of the CNN model and the priority score into a JSON message, and integrate them into a JSON format task unit through Schema standardization in combination with the multimodal work order data packet.

[0028] By constructing a CNN model, improve the accuracy of fault detection, extract features based on EfficientNet-B0, time series, and DistilBERT encoded log texts, enhance the robustness of the model, improve the comprehensiveness of fault diagnosis, improve the classification accuracy, and improve the sensitivity of the model to key fault signals based on the attention mechanism-based multimodal feature fusion. The reinforcement learning architecture of the policy network and the value network reduces manual intervention and avoids short-sighted decisions. The PPO algorithm is optimized to improve training stability, realize intelligent sorting of work orders, and optimize resource allocation.

[0029] S2. Use the IBM Qiskit framework for quantum state transformation, based on the classical GBDT framework, use quantum decision trees for split gain calculation, and optimize the loss function through the quantum Monte Carlo method, iteratively construct quantum decision trees, and output candidate scheduling plans, including: Query the current technician data through the API interface, including the technician skill level and the technician's available time; Use timestamp standardization to convert the task unit and technician data into a unified time format, perform time synchronization through a time convolutional network, and generate a time synchronization feature vector through cross-joining; Use the IBM Qiskit framework to convert the time matching features of each pair of technician-task units in the time synchronization feature vector into quantum states through amplitude encoding; Allocate a quantum bit to each pair of technician-task units to generate a quantum state dataset, representing the superposition state of the so-called features; Based on the gradient boosting framework of classical GBDT, initialize the first quantum decision tree, use the average matching degree of all samples as the initial prediction value, and calculate the initial residual, the formula is: , , where, is the initial prediction value, is the true matching degree of the th sample (obtained from historical experimental data), is the total number of samples, and the samples are historical technician-task unit pairs, is the initial residual of the th sample; Apply a parameterized quantum gate to the quantum state dataset, and calculate the splitting gain for each pair of technician-task units. The formula is: , where, is the splitting gain of the th feature, is the current prediction value; Use quantum measurement to extract the splitting gain, and set the number of measurements according to the Qiskit default value; Select the pair of technician-task units with the largest splitting gain as the splitting point, and output the splitting point and the corresponding quantum state dataset; Define the mean squared error loss function, and use the quantum Monte Carlo method to encode the initial residual as a residual quantum state. The formula is: , , where L is the expected value of the mean squared error loss function, is the encoded residual quantum state; Simulate a random walk through a quantum circuit, adjust the expected value of the mean squared error loss function, and use the COBYLA optimizer to optimize the quantum circuit parameters. Stop the optimization when the expected value reaches the minimum; Based on the gradient boosting idea of classical GBDT, iteratively construct multiple quantum decision trees. Input the predicted values of g - 1 trees and the residual into the gth tree, and perform a weighted accumulation of the outputs of all trees to output the final predicted value. The formula is: , where, is the learning rate (set according to cross-validation), is the sum of the predicted values of g - 1 trees, is the fitting value of the gth tree to the residual; Convert the final predicted value to a matching probability. The formula is: , where, is the matching probability of technician and task unit , is the matching probability of technician and task unit The final predicted value; Sort the matching probabilities in descending order, filter the top E pairs (set according to specific requirements), and generate chromosomes through encoding. Each chromosome is a complete scheduling plan; Based on empirical analysis, set the weights of the response time score, resource utilization score, and GBDT prediction score to be , and respectively, and define the multi-objective fitness function. The formula is: , , , , where, is the multi-objective fitness function value of the th chromosome, is the number of task units, is the actual response time (preset value), is the expected start time (judged according to the available time of technicians), is the duration of the task unit, is the matching degree between the task unit and the technician, is the matching probability predicted by GBDT, is the response time score, is the resource utilization score, is the GBDT prediction score, is the th chromosome; Calculate the fitness value of each chromosome. The formula is: , where, is the sum of the multi-objective fitness function values of all chromosomes, is the fitness value of the th chromosome, that is, the probability of being selected; Select chromosomes with high probability values through roulette wheel, generate the next generation of chromosomes through crossover and mutation, repeat the iterative operation and stop when the fitness value converges (if the best fitness value after continuous iteration does not have a new highest record, the algorithm has converged), and output the chromosomes of the last iteration, formatted as a list of candidate scheduling plans.

[0030] By using the IBM Qiskit framework for quantum state transformation, the quantum circuit simulates the random walk to adjust the expected value of the loss function, and the COBYLA is used to optimize the quantum circuit parameters, efficiently explore the optimization space, reduce the calculation time, avoid the local optimum problem of gradient descent - type optimizers, improve the global optimality of the scheduling scheme, reduce the number of iterations required for classical optimization, reduce the calculation cost, construct a quantum decision tree based on the GBDT framework for multi - objective scheduling optimization, efficiently process high - dimensional features, improve the accuracy of matching prediction, enhance the adaptability of the scheduling scheme, encode to generate chromosomes, calculate fitness values, screen, cross - mutate and output candidate scheduling schemes, avoid scheduling imbalance caused by a single optimization objective, enhance the robustness of the scheduling scheme, reduce manual intervention, and improve the operation and maintenance efficiency.

[0031] S3. By using the simulated annealing algorithm to make a small perturbation to the current optimized objective function value and output the optimal scheduling scheme, including: Use the CRITIC method to calculate the weights of the expected repair time, resource utilization rate, and quality stability respectively, and define the optimized objective function. The formula is: , where Z is the optimized objective function, and are the weights of the expected repair time, resource utilization rate, and quality stability, is the expected repair time (obtained from historical repair data), is the resource utilization rate (the ratio of the current number of technicians in use to the total number of technicians), is the quality stability (obtained from historical repair data); Set the initial temperature based on the analysis of the problem characteristics, which is the start stage of the simulated annealing process; Calculate the optimized objective function value of the current scheme as the starting point of the simulated annealing process; Make a small perturbation to the current solution to generate a new solution and recalculate the new objective function value; Judge whether to accept the new solution according to the difference of the optimized objective function. If the difference is less than 0, accept the new solution. If the difference is greater than 0, calculate the acceptance probability. The formula is: , where, is the acceptable probability, is the difference of the optimized objective function, is the initial temperature; Generate a uniform random number. If the acceptance probability is less than the uniform random number, accept the new solution. Otherwise, retain the current solution; Update the accepted new solution as the current solution; After each iteration, the initial temperature decays according to the cooling strategy. The formula is: , Among them, is the attenuated temperature, is the initial temperature, is the temperature reduction coefficient; Based on the current solution and temperature attenuation, continue the iteration. When the temperature drops to the lowest value, stop the iteration, output the current solution, and generate the optimal scheduling plan.

[0032] Through the high-temperature exploration and low-temperature convergence characteristics of the simulated annealing algorithm, avoid the local optimum problem of traditional optimization methods. The small perturbation and exponential cooling strategy reduce the computational overhead, while ensuring the superiority of the solution. Based on the weight setting and probability acceptance mechanism of historical data, improve the practicality and robustness of the plan.

[0033] S4. Generate task instructions through IMS, execute them, and store, collect, and analyze the generated multimodal data; Specifically, generating task instructions through IMS and executing them includes: Generate task instructions from the optimal scheduling plan through the IMS system, push them to the technician terminal, and the technician executes the task instructions.

[0034] Generating task instructions through the IMS system and pushing them to the technician terminal reduces the delay of traditional manual dispatch, improves the execution accuracy, and enhances the transparency of operation and maintenance management.

[0035] Furthermore, storing, collecting, and analyzing the generated multimodal data includes: Store the collected multimodal data and the generated optimal scheduling plan in the database. The database marks the timestamp for the stored data and uploads the stored data to the cloud for backup.

[0036] By uploading the scheduling plan and execution status to the central database in real time, improve data consistency and traceability, and reduce decision-making biases caused by scattered data.

[0037] This embodiment also provides an operation and maintenance work order management system based on artificial intelligence, including: A collection and decision-making module, used to collect multimodal data, generate multimodal work order data packets, construct a CNN model, optimize the policy network and value network using the PPO algorithm, optimize the model parameters through multi-task learning and gradient update, output the device failure probability and recommended actions, and generate JSON work orders; A scheduling optimization module, used to perform quantum state conversion using the IBM Qiskit framework, based on the classical GBDT framework, calculate the split gain using a quantum decision tree, and optimize the loss function through the quantum Monte Carlo method, iteratively construct a quantum decision tree, and output candidate scheduling plans; A global adjustment module, which is used to slightly perturb the current optimized objective function value through the simulated annealing algorithm and output the optimal scheduling plan; An execution storage module, which is used to generate task instructions through IMS, execute them, and store, collect, and analyze the generated multimodal data.

[0038] This embodiment also provides a computer device, which is applicable to the case of the operation and maintenance work order management method based on artificial intelligence, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the operation and maintenance work order management method based on artificial intelligence as proposed in the above embodiment.

[0039] This computer device may be a terminal. The computer device includes a processor, a memory, a communication interface, a display screen, and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner. The wireless manner can be achieved through WIFI, a carrier network, NFC (Near Field Communication), or other technologies. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer covered on the display screen, or a button, a trackball, or a touchpad set on the shell of the computer device, or an external keyboard, touchpad, or mouse, etc.

[0040] This embodiment also provides a storage medium, on which a computer program is stored. When the program is executed by a processor, it implements the operation and maintenance work order management method based on artificial intelligence as proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as a static random access memory (abbreviated as SRAM), an electrically erasable programmable read-only memory (abbreviated as EEPROM), an erasable programmable read-only memory (abbreviated as EPROM), a programmable read-only memory (abbreviated as PROM), a read-only memory (abbreviated as ROM), a magnetic memory, a flash memory, a magnetic disk, or an optical disc.

[0041] In summary, the present invention collects multimodal data, generates a multimodal work order data packet, constructs a CNN model, uses the PPO algorithm to optimize the policy network and the value network, optimizes the model parameters through multi-task learning and gradient update, outputs the device failure probability and recommended actions, generates a JSON work order, uses the IBM Qiskit framework for quantum state conversion, based on the classical GBDT framework, uses quantum decision trees for split gain calculation, and optimizes the loss function through the quantum Monte Carlo method, iteratively constructs quantum decision trees, and outputs candidate scheduling schemes; enhancing the accuracy and timeliness of fault prediction and action recommendation, and improving the multi-objective adaptation ability of scheduling.

[0042] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered within the scope of the claims of the present invention.

Claims

1. An operation and maintenance work order management method based on artificial intelligence, characterized by: include, Collect multimodal data, generate multimodal work order data packets, build CNN models, use PPO algorithm to optimize policy networks and value networks, optimize model parameters through multi-task learning and gradient update, output equipment failure probability and recommended actions, and generate JSON work orders; Use IBM Qiskit framework to convert quantum states. Based on the framework of classic GBDT, use quantum decision tree to calculate split gain, optimize loss function through quantum Monte Carlo method, iteratively construct quantum decision tree, and output candidate scheduling solutions. Use the simulated annealing algorithm to slightly perturb the current optimization objective function value and output the optimal scheduling solution; IMS generates task instructions, executes them, and stores, collects, and analyzes the generated multimodal data.

2. The artificial intelligence-based operation and maintenance work order management method according to claim 1, characterized in that: The method collects multimodal data, generates multimodal work order data packets, builds a CNN model, uses the PPO algorithm to optimize the policy network and value network, optimizes model parameters through multi-task learning and gradient update, outputs device failure probability and recommended actions, and generates a JSON work order, including: The multimodal data includes valve pressure value, valve image and voice data; Build a CNN model, collect historical multimodal data, use EfficientNet-B0 to extract valve image features, and generate valve image feature vectors through global average pooling; The features of valve pressure values ​​are extracted through time series analysis technology to generate valve pressure value feature vectors. DistilBERT is used to encode log text to generate text feature vectors. The valve image feature vector, text feature vector, and valve pressure value feature vector are fused through the attention mechanism to generate a multimodal feature vector. The number of fault categories is output through the fully connected layer, and the probability of the fault category is output through the Softmax layer. Define the state as the concatenation of multimodal feature vectors and failure probability, and define the discrete action space as no operation, repair, and replacement; Use a feedforward neural network to define the policy network, input the state, pass through two fully connected layers, output the action probability distribution of the current state, and set the action selected according to the current state as ; Use MLP to define the value network, input the multimodal feature vector, pass through two fully connected layers, and output the value estimate of the current state; Set the reward function based on the value network, calculate the reward value, optimize the strategy network and value network through the PPO algorithm, calculate the good or bad value of the fault category to the current state and the strategy probability ratio between the current time t and time t-1; Calculate the policy gradient value; Set each fault category as a task, use the Adam optimizer, update the parameters of the policy network and value network through the policy gradient value and value estimation, update the parameters of the CNN model by calculating the gradient value of the loss function, use the empirical rule to set the maximum number of iterations, and stop the iteration when the maximum number of iterations is reached; The multimodal data is input into the optimized CNN model, the equipment failure probability is output through Softmax, the recommended action is output through the policy network, the output results and priority scores of the CNN model are packaged into JSON messages, and combined with the multimodal work order data packets, they are integrated into JSON format task units through Schema standardization.

3. The artificial intelligence-based operation and maintenance work order management method according to claim 2, characterized in that: The method uses the IBM Qiskit framework to convert quantum states, uses a quantum decision tree to calculate split gain based on the framework of the classical GBDT, optimizes the loss function through the quantum Monte Carlo method, iteratively constructs a quantum decision tree, and outputs candidate scheduling solutions, including: Query the current technician data through the API interface, including technician skill level and technician availability; The task unit and technician data are converted into a unified time format using timestamp standardization, time synchronization is performed using a temporal convolutional network, and a time synchronization feature vector is generated using a cross-joining method. The IBM Qiskit framework is used to convert the time matching features of each pair of technicians and task units in the time synchronization feature vector into quantum states through amplitude encoding; Assign a quantum bit to each technician-task unit pair to generate a quantum state data set; Based on the gradient boosting framework of the classic GBDT, the first quantum decision tree is initialized, the average matching degree of all samples is used as the initial prediction value, and the initial residual is calculated; Apply parameterized quantum gates to the quantum state dataset, calculate the splitting gain for each technician-task unit pair, extract the splitting gain using quantum measurements, and set the number of measurements according to the Qiskit defaults; Select the technician-task unit pair with the largest splitting gain as the splitting point, and output the splitting point and the quantum state data set corresponding to the splitting point; Define the mean square error loss function, use the quantum Monte Carlo method to encode the initial residual into the residual quantum state, simulate the random walk through the quantum circuit, adjust the expected value of the mean square error loss function, use the COBYLA optimizer to optimize the quantum circuit parameters, and stop the optimization when the expected value reaches the minimum; Based on the gradient boosting idea of ​​classic GBDT, multiple quantum decision trees are iteratively constructed, and the prediction values ​​and residuals of g-1 trees are input into the first Trees, weighted accumulation of all tree outputs, output of the final prediction value, and conversion of the final prediction value into matching probability; Sort the matching probabilities in descending order, select the top E pairs, and generate chromosomes through encoding. Each chromosome is a complete scheduling solution. Based on empirical analysis, the weights of response time score, resource utilization score and GBDT prediction score are set, and a multi-objective fitness function is defined; Calculate the fitness value of each chromosome, select chromosomes with high probability values ​​through roulette, generate the next generation of chromosomes through crossover mutation, repeat the iterative operation and stop when the fitness value converges, output the chromosome of the last iteration, and format it as a list of candidate scheduling solutions.

4. The artificial intelligence-based operation and maintenance work order management method according to claim 1, characterized in that: The collected multimodal data is preprocessed first, and the generation of a multimodal work order data packet includes: Build a digital twin model, input the preprocessed valve pressure value, calculate the current stress value, and predict the stress value at the future time t+1 through the linear regression formula. Set the safety threshold through the statistical method, screen the predicted stress value at the future time t+1 that is less than the safety threshold, and mark it as abnormal; The valve image, text and current stress value are calculated through Transformer encoder and weighted fusion to generate a comprehensive feature vector; Initialize the Q table, set the action to generate a work order or not generate a work order, and calculate the reward value of the current state; The Q value is updated according to the Q-learning algorithm, and the Q value of the generated work order is greater than the Q value of the non-generated work order. The comprehensive feature vector, location, timestamp and predicted status are extracted to form a work order data packet, which is serialized in JSON format and output as a multimodal work order data packet.

5. The artificial intelligence-based operation and maintenance work order management method according to claim 3, characterized in that: The method of performing a small perturbation on the current optimization objective function value by using a simulated annealing algorithm and outputting an optimal scheduling solution includes: Define the optimization objective function, set the initial temperature based on the analysis of the problem characteristics, and simulate the beginning stage of the annealing process; Calculate the optimization objective function value of the current solution as the starting point of the simulated annealing process, make a small perturbation to the current solution, generate a new solution, and recalculate the new objective function value; Determine whether to accept the new solution based on the difference in the optimization objective function, and update the accepted new solution as the current solution; After each iteration, the initial temperature decays according to the cooling strategy, and the iteration continues. When the temperature drops to the minimum value, the iteration stops, the current solution is output, and the optimal scheduling plan is generated.

6. The artificial intelligence-based operation and maintenance work order management method according to claim 4, characterized in that: The generating of task instructions through IMS and executing the same include: The optimal scheduling plan generates task instructions through the IMS system and pushes them to the technician terminal, who then executes the task instructions.

7. The artificial intelligence-based operation and maintenance work order management method according to claim 4, characterized in that: The multimodal data generated by the storage, collection and analysis includes: It means storing the collected multimodal data and the optimal scheduling plan generated by analysis in a database. The database marks a timestamp for the stored data and uploads the stored data to the cloud for backup.

8. An operation and maintenance work order management system based on artificial intelligence, based on the operation and maintenance work order management method based on artificial intelligence according to any one of claims 1 to 7, characterized in that: include, The collection decision module is used to collect multimodal data, generate multimodal work order data packets, build a CNN model, use the PPO algorithm to optimize the policy network and value network, optimize the model parameters through multi-task learning and gradient update, output the equipment failure probability and recommended actions, and generate a JSON work order; The scheduling optimization module is used to convert quantum states using the IBM Qiskit framework. Based on the framework of the classic GBDT, it uses a quantum decision tree to calculate split gains, optimizes the loss function through the quantum Monte Carlo method, iteratively constructs a quantum decision tree, and outputs candidate scheduling solutions. The global adjustment module is used to make small perturbations to the current optimization objective function value through the simulated annealing algorithm and output the optimal scheduling solution; The execution storage module is used to generate task instructions through IMS, execute them, store, collect and analyze the generated multimodal data.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the artificial intelligence-based operation and maintenance work order management method described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the artificial intelligence-based operation and maintenance work order management method described in any one of claims 1 to 7 are implemented.

Images