Maintenance task distribution method and system based on GBRT algorithm
The maintenance task distribution model is constructed through the GBRT algorithm, which solves the problem that traditional maintenance task distribution depends on manual experience, realizes data-driven intelligent decision-making, improves the scientificity and accuracy of maintenance task allocation, dynamically adjusts resource allocation, and improves operation and maintenance Efficiency and equipment reliability.
Patent Information
- Application Number
- CN202411043110.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-31
- Publication Date
- 2024-11-08
AI Technical Summary
Traditional maintenance tasks distribution mainly relies on manual experience and are easily affected by subjective factors. It is difficult to fully consider historical data, team efficiency and current workloads, resulting in uneven resource allocation and slow response speed.
Using the maintenance task distribution method based on GBRT algorithm, by collecting and preprocessing historical maintenance plan data, equipment failure historical data and current work arrangement data, establishing an efficiency model for team execution work orders, evaluating the long-term operation effect of equipment, and building a reflective personnel ability and Vectors of equipment complexity features, train an iterative decision tree combination model, optimize maintenance task allocation, and generate reasonable task distribution suggestions based on prediction results.
Realize data-driven intelligent decision-making, avoids the subjectivity of manual decision-making, improves the scientificity and accuracy of maintenance tasks allocation, can dynamically adapt to the current work arrangements and team status, ensures the optimal allocation of resources, and improves overall operation and maintenance Efficiency and equipment reliability.
Smart Images

Figure CN118916677A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of GBRT algorithms, and more specifically, to a maintenance task distribution method and system based on the GBRT algorithm. Background Art
[0002] The present invention relates to the technical field of data analysis, artificial intelligence and equipment maintenance management, and specifically to an intelligent decision-making method and its application system for efficiently and accurately distributing maintenance tasks based on historical data of maintenance plans, historical data of equipment and facility failures and current work arrangements, using a Gradient Boosted Regression Trees (GBRT) algorithm.
[0003] In the production and operation of hydropower stations, regular maintenance and fault repair of equipment and facilities are key links to ensure normal operation. Traditional maintenance task distribution mainly relies on manual experience, which is easily affected by subjective factors and is difficult to fully consider various factors such as historical data, team efficiency and current workload, resulting in uneven resource allocation and slow response speed. Therefore, there is an urgent need for a maintenance task distribution solution based on data-driven and intelligent decision-making. Summary of the invention
[0004] The present invention proposes a maintenance task distribution method and system based on the GBRT algorithm to solve the above-mentioned problems: traditional maintenance task distribution mainly relies on manual experience, is easily affected by subjective factors, and is difficult to fully consider historical data, team efficiency, current workload and other factors, resulting in uneven resource allocation, slow response speed and other problems.
[0005] Technical solution: A maintenance task distribution method based on the GBRT algorithm, comprising the following steps:
[0006] S1. Collect and pre-process historical maintenance plan data, equipment failure history data and current work schedule data;
[0007] S2. Using the historical maintenance plan data, establish an efficiency model for the team to execute work orders, and add performance indicators including work order completion time, consumables utilization rate, and fault recurrence rate;
[0008] S3. Evaluate the long-term operation effect of the equipment after the maintenance by the team through the equipment failure history data, including the failure interval time and equipment performance recovery level;
[0009] S4. Using the GBRT algorithm, construct a vector that reflects the characteristics of personnel capabilities, project suitability, and equipment complexity through feature engineering, train and iterate the model of multiple decision tree combinations, optimize the efficiency and effect indicators after maintenance task allocation, and perform cross-validation and hyperparameter adjustment;
[0010] S5. Based on the list of pending maintenance tasks, the real-time status of the equipment and the current work schedule, the trained GBRT model is used to predict the expected efficiency and effect of each team in performing the tasks;
[0011] S6. Based on the prediction results and business rules, the system generates reasonable maintenance task distribution suggestions, including the team to which the task is assigned, the estimated completion time, and the required resources.
[0012] Preferably, the S1 collects historical records including maintenance items, execution personnel, start and end time, and maintenance result information;
[0013] Organize basic equipment information, historical operating status, fault type, occurrence and repair time, maintenance team and personnel information;
[0014] Get real-time status information on the current team's workload, personnel availability, and skill matching.
[0015] Preferably, the step S2 of establishing an efficiency model for a team to execute a work order comprises the following steps:
[0016] S2-1. Select the key factors that affect the efficiency of work order execution in the historical maintenance plan data, including task type, number of executors, task complexity, tools and consumables used;
[0017] S2-2. Generate new features based on the original data, including work order completion time, end time to start time, task complexity score, and team average experience;
[0018] S2-3. Standardize the numerical features to balance the impact of features of different dimensions on the model;
[0019] S2-4. The dataset is divided into a training set and a test set at a ratio of 80% and 20%; the training set is used to train the model, and the test set is used to evaluate the model performance.
[0020] Preferably, the model performs statistics on frequently occurring identical faults and analyzes the causes.
[0021] Preferably, the model performs fault cause analysis including obtaining historical maintenance data containing fault information, wherein the historical maintenance data includes the fault type, occurrence time, repair time, the team and personnel involved in the maintenance, and detailed records of the fault cause; processing missing values and abnormal values, and standardizing the data format;
[0022] Classify fault data by type and establish fault category labels; extract features that may cause faults from historical data, including equipment usage time, maintenance frequency, and operating environment parameters; use statistical analysis and machine learning methods, including association rule mining, decision trees, and random forests, to analyze the relationship between each feature and the occurrence of the fault, and identify the main cause of the fault.
[0023] Preferably, the model frequent fault statistics include calculating the frequency of occurrence of each type of fault and identifying frequent faults; analyzing the time of fault occurrence and identifying the high-incidence time periods and periodic patterns of faults.
[0024] Preferably, the step S5 using the trained GBRT model to predict the expected efficiency and effect of each team in performing tasks includes the following steps:
[0025] S5-1. Use the trained GBRT model, which has learned how to predict the efficiency and effectiveness of task execution based on feature data; construct an input feature vector based on real-time data, and the input features include but are not limited to personnel capability index experience, skill level, task and personnel fit, equipment complexity, task type and complexity, and current workload and availability.
[0026] S5-2. Input the constructed feature vector into the trained GBRT model to predict the expected efficiency and effectiveness indicators of each team in performing tasks, including task completion time, resource utilization, and fault recurrence rate.
[0027] Preferably, S6 analyzes the prediction results of the GBRT model to evaluate the expected efficiency and effect of each team in performing specific tasks; according to business rules and model prediction results, high-priority tasks are preferentially assigned to teams with good prediction results; tasks are evenly distributed to avoid a team being too busy or idle; urgent tasks and key equipment are considered;
[0028] Generate a specific task allocation plan based on the prediction results and decision rules, including the specific team to which the task is assigned, the estimated completion time, and the required resources;
[0029] Push the task allocation plan to each team, and the team will perform maintenance tasks according to the allocation plan;
[0030] Record the actual results of task execution, including actual completion time, resource consumption, and fault recurrence.
[0031] Preferably, the maintenance task distribution system based on the GBRT algorithm includes a data collection module, a data preprocessing module, a model training module, a task distribution module and a result feedback module.
[0032] Preferably, the maintenance task distribution system based on the GBRT algorithm includes an interface, which displays maintenance information, including current maintenance tasks, progress, responsible persons, and historical maintenance records; the interface includes a discussion area and message board functions, allowing maintenance personnel to communicate in real time and share experiences and solutions.
[0033] The present invention realizes accurate and efficient distribution of maintenance tasks by introducing the GBRT algorithm, which has the following significant advantages:
[0034] (a) Data-driven decision-making: Making full use of historical data and real-time work information avoids the subjectivity and limitations of manual decision-making and improves the scientificity and accuracy of maintenance task allocation.
[0035] (b) Dynamic adaptability: It can adjust the task distribution strategy in real time according to the current work schedule and team status to ensure the optimal allocation of resources.
[0036] (c) Predictability and continuous optimization: Through model prediction, we can not only predict the effects of different distribution schemes, but also continuously optimize the task allocation mechanism based on model feedback to improve overall operation and maintenance efficiency and equipment reliability.
[0037] In summary, the present invention provides an innovative intelligent distribution solution for maintenance tasks, which effectively solves the resource allocation problem in traditional maintenance management, helps enterprises to achieve refined and intelligent management of equipment maintenance work, reduce operation and maintenance costs, and improve equipment operation efficiency. BRIEF DESCRIPTION OF THE DRAWINGS
[0038] Figure 1 This is a flow chart of model training of the present invention;
[0039] Figure 2 A feature field table for training the model of the present invention;
[0040] Figure 3 This is the target field table for model training of the present invention. DETAILED DESCRIPTION
[0041] For examples, please refer to Table 1 - Table 2 - Figure 1 The present invention provides a maintenance task distribution method based on the GBRT algorithm, comprising the following steps:
[0042] S1. Collect and pre-process historical maintenance plan data, equipment failure history data and current work schedule data;
[0043] S2. Use historical maintenance plan data to establish an efficiency model for the team to execute work orders, and add performance indicators including work order completion time, consumables utilization rate, and fault recurrence rate;
[0044] S3. Evaluate the long-term operation effect of the equipment after maintenance by the team through the equipment failure history data, including the failure interval time and equipment performance recovery level;
[0045] S4. Using the GBRT algorithm, construct a vector that reflects the characteristics of personnel capabilities, project suitability, and equipment complexity through feature engineering, train and iterate the model of multiple decision tree combinations, optimize the efficiency and effect indicators after maintenance task allocation, and perform cross-validation and hyperparameter adjustment;
[0046] Specifically, data including personnel capability assessment, project suitability, equipment complexity and other characteristics are collected.
[0047] Perform data preprocessing, feature selection, and feature scaling to ensure the quality and applicability of the data.
[0048] Construct vector representations that reflect characteristics such as personnel capabilities, project fit, and equipment complexity.
[0049] The GBRT algorithm is used to iteratively construct an integrated model of multiple decision trees.
[0050] GBRT is an integrated learning method. Each decision tree is constructed based on the residuals of all previous trees to gradually reduce the bias of the model and improve the predictive ability of the overall model.
[0051] When training the GBRT model, you can adjust hyperparameters such as tree depth and learning rate to optimize the performance and generalization ability of the model.
[0052] Through methods such as cross-validation, the performance of the model on the training set and validation set is evaluated, and the model parameters are adjusted to improve the prediction accuracy.
[0053] Analyze the importance of each feature in the GBRT model and understand the contribution of each feature to the prediction of fault types.
[0054] Based on the model's prediction results, identify and understand the main factors and key features that affect the fault type.
[0055] Specifically, cross-validation and hyperparameter adjustment were performed to ensure that the dataset included features reflecting personnel capabilities, project suitability, equipment complexity, etc., and appropriate data cleaning and feature processing were performed.
[0056] Select evaluation metrics: Define metrics for evaluating model performance, such as mean squared error (MSE), mean absolute error (MAE), etc., and choose the most appropriate metric based on actual conditions.
[0057] Divide the dataset: Divide the dataset into a training set and a validation set, usually using a cross-validation approach such as K-fold Cross Validation or Hold-out Validation.
[0058] Define the GBRT model: Use the GBRT library in Python (such as XGBoost, LightGBM, etc.), define the model structure, and set the initial hyperparameter values, including learning rate, tree depth, number of leaf nodes, etc.
[0059] Cross-validation training model: For each set of hyperparameter settings, the model is trained using the cross-validation method:
[0060] Divide the training set into K folds.
[0061] For each fold, it is used as a validation set in turn, and the remaining folds are used as training sets. The model is trained multiple times and the average performance index is calculated.
[0062] Evaluate model performance: For each set of hyperparameter settings, calculate the mean and standard deviation of the model's performance indicators (such as MSE, MAE, etc.) on the validation set.
[0063] Select the optimal hyperparameter combination based on the performance on the validation set.
[0064] Hyperparameter adjustment: Adjust the hyperparameters based on the results of cross-validation and repeat steps 4 and 5 until the optimal hyperparameter combination is found.
[0065] Model selection and application: Select the model and hyperparameter combination with the best performance and retrain the model on the entire training set.
[0066] Use the final optimized model for prediction and application.
[0067] S5. Based on the list of pending maintenance tasks, the real-time status of the equipment and the current work schedule, the trained GBRT model is used to predict the expected efficiency and effect of each team in performing the tasks;
[0068] S6. Based on the prediction results and business rules, the system generates reasonable maintenance task distribution suggestions, including the team to which the task is assigned, the estimated completion time, and the required resources.
[0069] S1 collects historical records including maintenance items, executors, start and end times, and maintenance results;
[0070] Organize basic equipment information, historical operating status, fault type, occurrence and repair time, maintenance team and personnel information;
[0071] Get real-time status information on the current team's workload, personnel availability, and skill matching.
[0072] S2 establishes the efficiency model of the team executing the work order, which includes the following steps:
[0073] S2-1. Select the key factors that affect the efficiency of work order execution in the historical maintenance plan data, including task type, number of executors, task complexity, tools and consumables used;
[0074] S2-2. Generate new features based on the original data, including work order completion time, end time to start time, task complexity score, and team average experience;
[0075] S2-3. Standardize the numerical features to balance the impact of features of different dimensions on the model;
[0076] S2-4. The dataset is divided into a training set and a test set at a ratio of 80% and 20%; the training set is used to train the model, and the test set is used to evaluate the model performance.
[0077] The model counts frequently occurring identical faults and analyzes the causes.
[0078] The model performs fault cause analysis by obtaining historical maintenance data containing fault information. The historical maintenance data includes the fault type, occurrence time, repair time, the team and personnel involved in the maintenance, and detailed records of the fault cause; processing missing values and abnormal values, and standardizing data formats;
[0079] Classify fault data by type and establish fault category labels; extract features that may cause faults from historical data, including equipment usage time, maintenance frequency, and operating environment parameters; use statistical analysis and machine learning methods, including association rule mining, decision trees, and random forests, to analyze the relationship between each feature and the occurrence of the fault, and identify the main cause of the fault.
[0080] Specifically, collect fault data in historical data: including fault type, occurrence time, repair time, fault description, etc.; equipment information: equipment usage time, maintenance frequency, operating environment parameters (such as temperature, humidity, etc.); maintenance records: detailed information of each maintenance, including time, maintenance content, participants, etc.
[0081] Handle missing values and outliers to ensure data integrity.
[0082] Standardize various types of data to ensure uniform dimensions.
[0083] Integrate fault data, equipment information and maintenance records into a unified data set.
[0084] Specifically, establish fault category labels: According to the fault description and type, classify and establish fault category labels including mechanical fault, electrical fault, and control system fault.
[0085] Feature extraction: Extract features from the dataset that may cause failures, including but not limited to:
[0086] Equipment usage time (accumulated running time)
[0087] Maintenance frequency (number of maintenance per year)
[0088] Operating environment parameters (temperature, humidity, vibration, etc.)
[0089] The association rule mining method is used to discover the association relationship between fault features.
[0090] Specifically, the Apriori algorithm is used to identify the combination of fault features that often occur simultaneously.
[0091] Generate association rules based on frequent item sets, including the combination of antecedents (conditions) and consequents (conclusions), and evaluate the strength and frequency of the relationship between them.
[0092] Analyze the support (frequency), confidence (the probability of the conclusion occurring when the condition occurs) and lift (the degree of influence of the condition on the conclusion) of association rules.
[0093] Determine the main feature combinations and conditions that lead to specific failure types, identify possible failure causes and related factors, and adjust the equipment maintenance plan and preventive maintenance strategy based on the analysis results to reduce the occurrence and impact of common failures.
[0094] Decision tree: Build a decision tree model to analyze the impact of each feature on the fault type and identify the main causes of the fault.
[0095] Specifically, a decision tree model is constructed based on the selected features.
[0096] The decision tree model splits the features according to certain rules (such as information gain, Gini coefficient, etc.) to form a series of decision nodes and leaf nodes.
[0097] Analyze the splitting rules and feature importance of decision tree nodes.
[0098] According to the structure of the decision tree and the importance of features, the main features and key factors that affect the fault type are identified. According to the interpretability of the decision tree model, it is determined which features have the greatest impact on the classification of the fault type. According to the classification results of the leaf nodes and the splitting of the features, the main causes and related factors that may lead to the fault are identified.
[0099] A random forest model was constructed to improve the robustness and accuracy of the analysis through multiple sampling and modeling.
[0100] Specifically, based on the selected features, a random forest model is constructed.
[0101] A random forest is an ensemble model consisting of multiple decision trees, each trained with a different random sample and a random subset of features.
[0102] The random forest model improves the robustness and accuracy of the analysis through multiple sampling and modeling.
[0103] Each tree is trained on a different random sample and feature subset. This ensemble approach reduces the risk of overfitting and improves the generalization ability of the model.
[0104] Analyze the contribution and importance of individual decision trees in a random forest model.
[0105] Through ensemble learning, random forests can comprehensively consider the contribution of each feature to the classification of fault types, helping to identify and understand the main influencing factors and key features of equipment failures.
[0106] The model's frequent fault statistics include calculating the frequency of each type of fault and identifying frequent faults; analyzing the time of fault occurrence and identifying high-incidence periods and periodic patterns.
[0107] S5 uses the trained GBRT model to predict the expected efficiency and effect of each team's task execution, including the following steps:
[0108] S5-1. Use the trained GBRT model, which has learned how to predict the efficiency and effectiveness of task execution based on feature data; construct input feature vectors based on real-time data. Input features include but are not limited to personnel capability indicators, experience, skill level, task-personnel fit, equipment complexity, task type and complexity, and current workload and availability.
[0109] S5-2. Input the constructed feature vector into the trained GBRT model to predict the expected efficiency and effectiveness indicators of each team in performing tasks, including task completion time, resource utilization, and fault recurrence rate.
[0110] S6 analyzes the prediction results of the GBRT model and evaluates the expected efficiency and effect of each team in performing specific tasks; according to business rules and model prediction results, high-priority tasks are assigned to teams with good prediction results first; tasks are evenly distributed to avoid a team being too busy or idle; urgent tasks and key equipment are taken into consideration;
[0111] Generate a specific task allocation plan based on the prediction results and decision rules, including the specific team to which the task is assigned, the estimated completion time, and the required resources;
[0112] Push the task allocation plan to each team, and the team will perform maintenance tasks according to the allocation plan;
[0113] Record the actual results of task execution, including actual completion time, resource consumption, and fault recurrence.
[0114] A maintenance task distribution system based on GBRT algorithm includes a data collection module, a data preprocessing module, a model training module, a task distribution module and a result feedback module.
[0115] A maintenance task distribution system based on GBRT algorithm includes an interface that displays maintenance information, including current maintenance tasks, progress, responsible persons, and historical maintenance records; the interface includes discussion area and message board functions, allowing maintenance personnel to communicate in real time and share experiences and solutions;
[0116] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. The above embodiments and descriptions are only preferred examples of the present invention and are not intended to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention. The scope of protection of the present invention is defined by the attached claims and their equivalents.
Claims
1. A maintenance task distribution method based on GBRT algorithm, characterized in that: The maintenance task distribution method based on the GBRT algorithm comprises the following steps: S1. Collect and pre-process historical maintenance plan data, equipment failure history data and current work schedule data; S2. Using the historical maintenance plan data, establish an efficiency model for the team to execute work orders, and add performance indicators including work order completion time, consumables utilization rate, and fault recurrence rate; S3. Evaluate the long-term operation effect of the equipment after the maintenance by the team through the equipment failure history data, including the failure interval time and equipment performance recovery level; S4. Using the GBRT algorithm, construct a vector that reflects the characteristics of personnel capabilities, project suitability, and equipment complexity through feature engineering, train and iterate the model of multiple decision tree combinations, optimize the efficiency and effect indicators after maintenance task allocation, and perform cross-validation and hyperparameter adjustment; S5. Based on the list of pending maintenance tasks, the real-time status of the equipment and the current work schedule, the trained GBRT model is used to predict the expected efficiency and effect of each team in performing the tasks; S6. Based on the prediction results and business rules, the system generates reasonable maintenance task distribution suggestions, including the team to which the task is assigned, the estimated completion time, and the required resources.
2. A maintenance task distribution method based on GBRT algorithm according to claim 1, characterized in that: The S1 collects historical records including maintenance items, execution personnel, start and end time, and maintenance result information; Organize basic equipment information, historical operating status, fault type, occurrence and repair time, maintenance team and personnel information; Get real-time status information on the current team's workload, personnel availability, and skill matching.
3. The maintenance task distribution method based on the GBRT algorithm according to claim 1 is characterized in that: The S2 establishes the efficiency model of the team executing the work order, including the following steps: S2-1. Select the key factors that affect the efficiency of work order execution in the historical maintenance plan data, including task type, number of executors, task complexity, tools and consumables used; S2-2. Generate new features based on the original data, including work order completion time, end time to start time, task complexity score, and team average experience; S2-3. Standardize the numerical features to balance the impact of features of different dimensions on the model; S2-4. The dataset is divided into a training set and a test set at a ratio of 80% and 20%; the training set is used to train the model, and the test set is used to evaluate the model performance.
4. A maintenance task distribution method based on GBRT algorithm according to claim 3, characterized in that: The model counts the frequent occurrence of the same faults and analyzes the causes.
5. A maintenance task distribution method based on GBRT algorithm according to claim 4, characterized in that: The model performs fault cause analysis by obtaining historical maintenance data containing fault information, including fault type, occurrence time, repair time, teams and personnel involved in maintenance, and detailed records of fault causes; processing missing values and abnormal values, and standardizing data formats; Classify fault data by type and establish fault category labels; extract features that may cause faults from historical data, including equipment usage time, maintenance frequency, and operating environment parameters; use statistical analysis and machine learning methods, including association rule mining, decision trees, and random forests, to analyze the relationship between each feature and the occurrence of the fault, and identify the main cause of the fault.
6. A maintenance task distribution method based on GBRT algorithm according to claim 4, characterized in that: Model frequent fault statistics include calculating the frequency of each type of fault and identifying frequent faults; Analyze the time when faults occur to identify peak periods and periodic patterns.
7. The maintenance task distribution method based on the GBRT algorithm according to claim 1 is characterized in that: The S5 uses the trained GBRT model to predict the expected efficiency and effect of each team in performing tasks, including the following steps: S5-1. Using the trained GBRT model, the trained GBRT model has learned how to predict the efficiency and effect of task execution based on feature data; constructing an input feature vector based on real-time data, the input features include but are not limited to personnel capability index experience, skill level, task and personnel fit, equipment complexity, task type and complexity, and current workload and availability; S5-2. Input the constructed feature vector into the trained GBRT model to predict the expected efficiency and effectiveness indicators of each team in performing tasks, including task completion time, resource utilization, and fault recurrence rate.
8. The maintenance task distribution method based on the GBRT algorithm according to claim 1 is characterized in that: The S6 analyzes the prediction results of the GBRT model and evaluates the expected efficiency and effect of each team in performing specific tasks; high-priority tasks are preferentially assigned to teams with good prediction results according to business rules and model prediction results; Balance the distribution of tasks to avoid a team being too busy or idle; consider urgent tasks and key equipment; Generate a specific task allocation plan based on the prediction results and decision rules, including the specific team to which the task is assigned, the estimated completion time, and the required resources; Push the task allocation plan to each team, and the team will perform maintenance tasks according to the allocation plan; Record the actual results of task execution, including actual completion time, resource consumption, and fault recurrence.
9. A maintenance task distribution system based on GBRT algorithm, characterized in that: The maintenance task distribution system based on the GBRT algorithm includes a data collection module, a data preprocessing module, a model training module, a task distribution module and a result feedback module.
10. A maintenance task distribution system based on GBRT algorithm according to claim 9, characterized in that: The maintenance task distribution system based on the GBRT algorithm includes an interface, which displays maintenance information, including current maintenance tasks, progress, responsible persons, and historical maintenance records; the interface includes a discussion area and a message board function, allowing maintenance personnel to communicate in real time and share experiences and solutions.
Citation Information
Patent Citations
Repair task scheduling method based on predictive maintenance
CN116644898A
Offshore wind power multi-dimensional pre-maintenance method based on XGBoost algorithm
CN116976850A
Work order generation and matching method and device
CN117520099A
Cited By
Power equipment maintenance plan intelligent arrangement data model based on optimization solver
CN119359290A
Efficient task allocation method and system for OA software and medium
CN120069424A
Team management enabling training system
CN120197891A
