Method and system for managing equipment information for post-repair evaluation

Abstract

The application provides an equipment information management method and system for post-overhaul evaluation, relates to the technical field of equipment overhaul management, and comprises the following steps: constructing an overhaul directed graph model, dividing an overhaul whole process into multiple logical links; establishing a multi-dimensional evaluation index system, and determining the weights of indexes by using an analytic hierarchy process; collecting multi-source overhaul data, including manually entered qualitative and quantitative data, and system-integrated cost and working hour data; performing standardization processing on the data, converting qualitative data into quantitative data, and normalizing all quantitative data; calculating the scores of each link efficiency and the overall overhaul efficiency index according to a weighted fusion algorithm, identifying low-efficiency links, and automatically triggering an improvement process to form a closed-loop optimization mechanism. The application realizes the system integration of overhaul data and multi-dimensional efficiency evaluation, and effectively improves the quantitative analysis and continuous optimization capability of overhaul management.

Description

Technical Field

[0001] This invention relates to the field of equipment maintenance management technology, and specifically to a method and system for managing equipment information for post-maintenance evaluation. Background Technology

[0002] In industrial environments such as power plants and factories, regular maintenance of various equipment is a crucial step in ensuring production safety and stable operation. With the development of industrial informatization, equipment maintenance management has transitioned from traditional paper records to digital systems. Various enterprise resource planning (ERP) systems, manufacturing execution systems (MES), and equipment management systems are widely used in maintenance operations to record maintenance results, costs, and time information.

[0003] However, existing post-maintenance evaluation technologies still have significant limitations, mainly in that: current evaluation systems focus primarily on outcome data after maintenance, such as whether equipment has returned to normal, whether the total maintenance cost has exceeded the budget, and whether the overall project schedule has been delayed, while neglecting the execution quality and efficiency of each stage during the maintenance process. For example, although the equipment may ultimately be successfully repaired, there may be multiple misjudgments in the fault diagnosis stage, leading to unnecessary component replacements; or improper process execution during the repair stage may create hidden dangers for subsequent operation. This "results-oriented, process-neglecting" evaluation model fails to truly reflect the overall picture of maintenance and cannot identify specific bottlenecks and risk points in the process. Therefore, there is an urgent need for a method and system that can overcome the limitations of existing "results-oriented" evaluations, and achieve refined monitoring and management of the execution efficiency of each stage by constructing a dynamic value chain model covering the entire maintenance process. Summary of the Invention

[0004] The main objective of this invention is to provide a method and system for managing equipment information for post-maintenance evaluation, thereby solving the problems mentioned in the background art.

[0005] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is: a method for managing equipment information for post-maintenance evaluation, comprising the following steps: S1. Divide the entire maintenance process of the equipment into several logical steps and construct a directed graph model for maintenance. S2. Obtain historical maintenance data, extract multi-dimensional evaluation indicators from the historical maintenance data, and calculate the weight of each evaluation indicator. S3. Collect multi-source maintenance data; S4. Standardize the collected multi-source maintenance data, map qualitative data to quantitative data through preset transformation rules, normalize all quantitative data according to indicator type to obtain a standardized dataset, and analyze the standardized dataset to identify abnormal data. S5. Based on the weighted fusion algorithm, combined with the standardized dataset and the weights of each evaluation indicator, calculate the performance score of the logical link corresponding to the abnormal data and the overall maintenance performance index; identify the inefficient links based on the performance score and the overall maintenance performance index, and automatically trigger the improvement process to form a closed-loop optimization mechanism.

[0006] Furthermore, the detailed process of step S1 is as follows: S101. Obtain the entire maintenance process of the equipment under maintenance. The entire maintenance process includes the business process and the maintenance process. The business process and the maintenance process are divided into several steps according to the execution steps. S102. Configure a unique step ID for each step in the maintenance process, and establish a business description document for each step in the business process; associate the step ID of each step in the entire maintenance process with the corresponding business description document to form a logical step. S103. Set each logical link as a node, and connect each node with directed edges according to the flow relationship of each logical link to form a maintenance directed graph model. The node attributes include the stage ID, input parameters, and output parameters; the edge attributes of a directed edge include the flow trigger condition and the standard flow time. After each maintenance process is completed, the directed graph model collects the actual processing time of each logical step in the entire maintenance process and updates the standard processing time using the moving average method. The updated expression is as follows: (1); in, The new standard takes time to process. This refers to the total number of maintenance processes that were statistically analyzed before updating the directed graph model. The original standard processing time, This represents the actual processing time of each logical step.

[0007] Furthermore, the detailed process of step S2 is as follows: S201. Obtain historical maintenance data, divide the historical maintenance data into three dimensions, and extract the corresponding evaluation indicators from each dimension. The dimensions include: technological efficiency, economic efficiency, and time efficiency. Technical performance dimensions include: fault detection indicators, quality compliance indicators, and process conformity indicators; The economic efficiency dimension includes: unit resource consumption indicators, resource utilization rate indicators, and cost deviation rate indicators; Time efficiency dimensions include: time-based target indicators, time-consuming percentage of each process, and maintenance delay rate. S202. The weights of each evaluation indicator are calculated using the analytic hierarchy process, as follows: For any dimension, the evaluation indicators within that dimension are numbered sequentially, and pairwise importance comparisons are performed on each evaluation indicator within that dimension to obtain importance scale values ​​between the evaluation indicators, thus constructing a judgment matrix for that dimension; the element values ​​of the judgment matrix are determined by the scale values ​​between the evaluation indicators corresponding to their respective numbers. The importance scale value between assessment indicators is determined by experts based on the importance relationship between the assessment indicators, and the importance scale value is determined by the importance relationship. For any dimension of the judgment matrix, we have: Calculate the largest eigenvalue and the corresponding eigenvector of the judgment matrix, normalize the eigenvector to obtain the weight of each evaluation index, and denote the normalized eigenvector as the weight vector of the evaluation index for that dimension. The consistency score for this dimension is calculated using the following expression: (2); in, For consistency score, The largest eigenvalue, This represents the number of evaluation indicators for this dimension. Query Hierarchical Analysis Standard The table will show the number of evaluation metrics for that dimension. Corresponding average random consistency index Calculate the consistency ratio The expression is as follows: (3); when If the value is less than 0.1, it indicates that the weight allocation of each evaluation indicator is reasonable; otherwise, step S202 is repeated until the condition is met. Less than 0.1.

[0008] Furthermore, the multi-source maintenance data includes two categories: manually entered data D1 and industrial business system integrated data D2; The data types for manually entered data D1 include: qualitative data D11 and quantitative data D12; The data types of the industrial business system integration data D2 include: cost data D21 and working hour data D22.

[0009] Furthermore, the detailed process of step S4 is as follows: S401. Based on the preset qualitative and quantitative conversion rule library, convert the qualitative data in the multi-source maintenance data into quantitative data, and map D11 to quantitative data S1. The transformation method is as follows: assign a unique score to each type of qualitative data; the resulting quantitative dataset D3 is obtained after the transformation. S402. Merge the quantitative dataset D3 with the original quantitative data D12 to obtain the total quantitative dataset D4; select a normalization algorithm according to the indicator type to convert the D4 data into normalized values ​​to obtain the standardized dataset D5; the indicator types include: positive indicators and negative indicators; S403. Calculate the mean and standard deviation of each evaluation indicator in D5; if the data exceeds the range threshold, the data is determined to be abnormal data. S404. Use the isolated forest algorithm to perform secondary verification on the abnormal data and output the probability of the abnormal data. At the same time, combine the maintenance directed graph model to trace the related nodes of the abnormal data and output the cause of the abnormality. The input to the Isolation Forest algorithm model is the normalized feature values ​​of nine evaluation indicators for outlier data, and the output is the outlier probability corresponding to each outlier data point in the input.

[0010] Furthermore, the detailed process of step S5 is as follows: S501. From the logical steps corresponding to the abnormal data, extract the standardized data corresponding to the logical steps, and combine them with the weights of the corresponding evaluation indicators. Calculate the performance score of the logical steps using a weighted summation method, as shown in the following expression: (4); in, For the first The performance score of the logical steps corresponding to each abnormal data point. for The first logical step corresponding to the first abnormal data The normalized value of each evaluation indicator. For the first The weight of each evaluation indicator; S502. Based on the performance score of each logical step and its weight, the overall maintenance performance index is calculated using a weighted summation method, as shown in the following expression: (5); in, The overall maintenance efficiency index. For the first The weight of the logical step corresponding to each abnormal data point; S503. Preset overall index threshold and process score threshold; when the overall maintenance efficiency index is less than the overall index threshold, the overall efficiency is judged to be low; when the logic process efficiency score is less than the process score threshold, the logic process is judged to be a low efficiency process. S504. Build a neural network prediction model to predict the future. The performance score of a logical step in the entire maintenance process at a given time point is calculated. If, in two consecutive maintenance processes, a certain logical step... If the predicted values ​​are all lower than the score threshold of the process, the logic process will be judged as an inefficient process in advance. The model's input consists of normalized time-series feature values ​​of nine evaluation metrics for outlier data, with a 9-dimensional dimension. The output is the future... The efficiency score of each logical step in the entire maintenance process at any given moment; S505. Automatically generate improvement task sheets for inefficient processes; the improvement task sheet includes root cause analysis, responsible department, rectification deadline, rectification objectives, and rectification priority; the priority calculation formula is: (6); in, Priority The threshold for the stage score; S506. The system allocates resources based on priority. High-priority tasks are automatically matched with senior technicians as rectification personnel and granted priority scheduling permissions. Low-priority tasks send standardized operation manuals and self-rectification guidelines to the rectification personnel in charge. The system recalculates after rectification is completed. and The system will also provide similar cases for rectification personnel, who can refer to these cases to carry out rectification. The process for selecting similar cases is as follows: A knowledge base of improvement solutions is established in advance, and similar cases are matched using the cosine similarity algorithm; S507. Recalculate the inefficient processes after rectification. and Let the overall index threshold be... When the rectified inefficient processes meet the requirements and When the condition is met, it indicates that the closed-loop optimization of the inefficient link is complete; when it is not met, rectification is carried out again until the closed-loop optimization of the inefficient link is complete.

[0011] The present invention also provides an equipment information management system for post-maintenance evaluation, for implementing the steps in the above method, including: The maintenance directed graph model building module is used to divide the entire maintenance process into several logical steps and build a maintenance directed graph model. The multi-dimensional evaluation indicator management module is used to establish an evaluation indicator system based on technical, economic, and time dimensions, and to determine the weight of each evaluation indicator through the analytic hierarchy process. The multi-source maintenance data acquisition module is used to collect manually entered qualitative and quantitative data, as well as cost and working hour data integrated from industrial business systems, and to perform data verification and completion. The data standardization processing module is used to standardize the collected multi-source maintenance data, including converting qualitative data into quantitative data and normalizing all quantitative data. The efficiency calculation and closed-loop optimization module is used to calculate the efficiency score and overall maintenance efficiency index of the logical links corresponding to abnormal data based on standardized data and evaluation index weights, identify inefficient links and automatically trigger improvement processes to form a closed-loop optimization mechanism. The evaluation indicator management module is used to configure, maintain, and manage multi-dimensional evaluation indicators.

[0012] Furthermore, the multi-source maintenance data acquisition module includes: a manual data entry unit, a system data integration unit, and a data verification and completion unit; The manual data entry unit is used to collect qualitative and quantitative data of the maintenance process through a structured form interface, and to perform mandatory field verification and modification log recording. The system data integration unit is used to automatically collect cost data and working hour data from external industrial business systems through standardized API interfaces; The data verification and completion unit is used to verify the range, logic and integrity of the collected data, and to complete and mark abnormal data using historical average values.

[0013] Furthermore, the performance calculation and closed-loop optimization module includes: a process performance calculation unit, an inefficiency identification unit, and an improvement task generation and tracking unit; The process efficiency calculation unit is used to calculate the efficiency score of a logical process corresponding to abnormal data by combining its standardized data and the weights of the corresponding evaluation indicators using a weighted summation method. The low-efficiency identification unit is used to determine whether the overall efficiency is too low and to identify specific low-efficiency links based on preset overall efficiency thresholds and process efficiency thresholds. An improved task generation and tracking unit is used to automatically generate improvement task sheets for identified inefficient processes and track the rectification results for re-evaluation of efficiency.

[0014] Furthermore, the evaluation indicator management module includes: an indicator configuration unit, a weight calculation unit, and a weight storage and retrieval unit; The indicator configuration unit is used to define and maintain various evaluation indicators under the dimensions of technology, economy, and time efficiency. The weight calculation unit is used to construct a judgment matrix using the analytic hierarchy process (AHP) and to calculate and verify the weights of each evaluation index. The weight storage and retrieval unit is used to store the final indicator weight vector and provide weight data support for performance calculation.

[0015] Beneficial effects: (1) By constructing a directed graph model for maintenance, the structured modeling and segment division of the entire maintenance process were realized, which improved the controllability and transparency of the maintenance process; (2) Extract multi-dimensional evaluation indicators, integrate technical, economic and time factors, and comprehensively and objectively evaluate maintenance efficiency to avoid one-sided evaluation; (3) It realizes the standardized processing and fusion analysis of multi-source heterogeneous data, supports the automatic conversion of qualitative data to quantitative data, and improves the efficiency of data utilization and the accuracy of evaluation; (4) Adopt a closed-loop optimization mechanism to automatically identify inefficient links and trigger improvement tasks, forming a continuous optimization cycle of "evaluation-improvement-re-evaluation" to improve the overall maintenance management level. Attached Figure Description

[0016] The present invention will be further described below with reference to the accompanying drawings and embodiments: Figure 1 This is a schematic diagram of the method flow of the present invention; Figure 2 This is a schematic diagram of the system structure of the present invention. Detailed Implementation

[0017] Example 1 like Figure 1 The equipment information management method shown, used for post-maintenance evaluation, includes the following steps: S1. Divide the entire maintenance process of the equipment into several logical steps and construct a directed graph model for maintenance. The detailed process of this step is as follows: S101. Obtain the entire maintenance process of the equipment under maintenance. The entire maintenance process includes the business process and the maintenance process. The business process and the maintenance process are divided into several steps according to the execution steps. S102. Configure a unique step ID for each step in the maintenance process, and establish a business description document for each step in the business process; associate the step ID of each step in the entire maintenance process with the corresponding business description document to form a logical step. S103. Set each logical link as a node, and connect each node with directed edges according to the flow relationship of each logical link to form a maintenance directed graph model. The node attributes include the stage ID, input parameters, and output parameters; the edge attributes of a directed edge include the flow trigger condition and the standard flow time. After each maintenance process is completed, the directed graph model collects the actual processing time of each logical step in the entire maintenance process and updates the standard processing time using the moving average method. The updated expression is as follows: (1); in, The new standard takes time to process. This refers to the total number of maintenance processes that were statistically analyzed before updating the directed graph model. The original standard processing time, The actual processing time of each logical step. The number of maintenance processes that were counted before updating the directed graph model for maintenance.

[0018] S2. Obtain historical maintenance data, extract multi-dimensional evaluation indicators from the historical maintenance data, and calculate the weight of each core evaluation indicator. The detailed process of this step is as follows: S201. Obtain historical maintenance data, divide the historical maintenance data into three dimensions, and extract the corresponding evaluation indicators from each dimension. The dimensions include: technological efficiency, economic efficiency, and time efficiency. Technical performance dimensions include: fault detection indicators, quality compliance indicators, and process conformity indicators; The economic efficiency dimension includes: unit resource consumption indicators, resource utilization rate indicators, and cost deviation rate indicators; Time efficiency dimensions include: time-based target indicators, time-consuming percentage of each process, and maintenance delay rate. S202. The weights of each evaluation indicator are calculated using the analytic hierarchy process, as follows: For any dimension, the evaluation indicators within that dimension are numbered sequentially, and pairwise importance comparisons are performed on each indicator to obtain importance scale values ​​between them, thus constructing a judgment matrix for that dimension. The element values ​​of the judgment matrix are determined by the scale values ​​between the evaluation indicators corresponding to their respective numbers, specifically: the judgment matrix... Line number The element of the column is the first element in that dimension. The evaluation indicator is relative to the first The importance scale value of each evaluation indicator; for example, for the technical effectiveness dimension, if the fault detection indicators, quality compliance indicators, and process compliance indicators are labeled as 1, 2, and 3 respectively, then for the judgment matrix of the technical effectiveness dimension, the element in the first row and second column is the importance scale value of the fault detection indicators relative to the quality compliance indicators. The importance scale values ​​between evaluation indicators are determined by experts based on the importance relationship between the indicators, and the importance scale values ​​are determined based on the importance relationship, specifically as follows: When both are equally important, the importance scale value is 1; The former is slightly more important than the latter, and the importance scale value is 3; The former is significantly more important than the latter, so the importance scale value is 5; The former is significantly more important than the latter, and its importance scale value is set to 7; The former is extremely more important than the latter, and its importance scale value is set to 9; When experts consider the importance relationship between two factors to be between two adjacent levels, the importance scale is taken as the average of the two importance scale values. For example, when experts consider the importance relationship between two factors to be between "equally important" and "the former is slightly more important than the latter," the importance scale value is 2. The importance scale value reflects the relative relationship between the two factors. Therefore, the product of the importance scale value of the former relative to the latter and the importance scale value of the latter relative to the former is 1. For example, in a certain dimension, the importance scale value of evaluation indicator A relative to B is 3. Conversely, the importance scale value of B relative to A is 1 / 3. For any dimension of the judgment matrix, we have: Calculate the largest eigenvalue and the corresponding eigenvector of the judgment matrix. Since the eigenvector corresponding to the largest eigenvalue retains the relative importance relationship dominated by the matrix, the eigenvector is normalized to obtain the weight of each evaluation index. At this time, the normalized eigenvector can be recorded as the weight vector of the evaluation index of that dimension. In the feature vector, the first element corresponds to the evaluation index with index 1 in the same dimension, the second element corresponds to the evaluation index with index 2 in the same dimension, and the third element corresponds to the evaluation index with index 3 in the same dimension. The consistency score for this dimension is calculated using the following expression: (2); in, For consistency score, The largest eigenvalue, The number of evaluation indicators for this dimension is 3 in this embodiment; Query Hierarchical Analysis Standard The table will show the number of evaluation metrics for that dimension. Corresponding average random consistency index Calculate the consistency ratio The expression is as follows: (3).

[0019] when If the value is less than 0.1, it indicates that the weight allocation of each evaluation indicator is reasonable; otherwise, step S202 is repeated until the condition is met. Less than 0.1.

[0020] S3. Collect multi-source maintenance data, which includes qualitative and quantitative data from manually entered maintenance processes, as well as cost and time data integrated from industrial business systems. Multi-source maintenance data includes two categories: manually entered data D1 and integrated data from industrial business systems D2; The data types for manually entered data D1 include: qualitative data D11 and quantitative data D12; The manual data entry method for D1 is through a structured form interface; the form has required and optional fields, and the form cannot be submitted if the required fields are not entered, and the log is automatically saved after the data is modified; The data validation method for manually entered data D1 is to perform range validation on D12 and integrity validation on D11. The data types of the industrial business system integration data D2 include: cost data D21 and working hour data D22; The D2 data acquisition method for industrial business system integration involves connecting with the industrial business system through a standardized API interface. The data verification method for industrial business system integration data D2 is to perform non-negativity verification on D21 and logical verification on D22; abnormal data is completed by using the arithmetic mean of the most recent historical normal data of the corresponding evaluation index, and the completed data is marked as completed data.

[0021] S4. Standardize the collected multi-source maintenance data, mapping qualitative data to quantitative data using preset transformation rules. Normalize all quantitative data according to indicator type to obtain a standardized dataset. Analyze the standardized dataset to identify outliers. The detailed process of this step is as follows: The detailed process is as follows: S401. Based on the preset qualitative and quantitative conversion rule library, convert the qualitative data in the multi-source maintenance data into quantitative data, and map D11 to quantitative data S1. The transformation method is as follows: assign a unique score to each type of qualitative data; the resulting quantitative dataset D3 is obtained after the transformation. Qualitative data here refers to the level of data. For example, quality data will be qualitatively classified as excellent, good, average, poor, etc. At this time, the qualitative level is converted into a specific score, which is the unique score mentioned above, and the qualitative data can be converted into quantitative data. S402. Merge the quantitative dataset D3 with the original quantitative data D12 to obtain the total quantitative dataset D4; select a normalization algorithm according to the indicator type to convert the D4 data into normalized values ​​to obtain the standardized dataset D5. Indicator types include positive indicators and negative indicators; Positive indicators are those whose values ​​are positively correlated with performance levels, while negative indicators are those whose values ​​are negatively correlated with performance levels. Positive indicators are normalized using a maximum value normalization algorithm, while negative indicators are normalized using a minimum value normalization algorithm. S403. Calculate the mean and standard deviation of each evaluation indicator in D5; if the data exceeds the range threshold, the data is determined to be abnormal data; the range threshold here is constructed by the mean and standard deviation of the evaluation indicator, specifically: the upper limit of the range threshold is the sum of the mean and the standard deviation, and the lower limit of the range threshold is the difference between the mean and the standard deviation. S404. Use the isolated forest algorithm to perform secondary verification on the abnormal data and output the abnormal probability of the abnormal data. The input to the Isolation Forest model is the normalized feature values ​​of the nine evaluation indicators of the abnormal data obtained in step S403, with a dimension of 9. The output is the anomaly probability corresponding to each abnormal data point in the input; here, the anomaly probability refers to the likelihood that the abnormal data will become abnormal. The model building process is as follows: Extract normal data from history after removing outliers as input for training; determine the initial values ​​of the model's hyperparameters, including: 100 isolated trees, 256 subsamples, and a maximum depth of 8 per tree; 256 samples are randomly drawn with replacement from the normal data sample set as the construction samples for a single isolated tree; a feature dimension is randomly selected from the current sample set, and a splitting threshold within the range of the feature value is randomly selected to perform a binary split on the sample set; The left and right child nodes after the split recursively repeat the above splitting process until any of the following stopping conditions are met: (1) Reach the preset maximum depth of a single tree; (2) The current node contains less than or equal to 1 sample and cannot be split further; Complete the construction of the topology of a single isolated tree; based on the above tree construction process, generate 100 isolated trees to complete the construction of the isolated forest model; The anomaly probability is obtained by normalizing the anomaly score, and the expression for the anomaly score is as follows: (4); in, For abnormal data, anomaly scoring, This is the average length of all tree paths corresponding to this abnormal data. Let Euler's constant be 1. This represents the number of samples contained in a single isolated leaf node. Set a secondary verification threshold to reclassify abnormal data with an anomaly probability less than the secondary verification threshold as normal data, and retain abnormal data with an anomaly probability greater than or equal to the secondary verification threshold as abnormal data. After the secondary verification is completed, the associated nodes of the abnormal data are traced by combining the directed graph model of maintenance, and the cause of the abnormality is output. Specifically, the process involves using each maintenance logic step as a directed graph network node, and the maintenance process flow and data transmission relationships as directed edges. Each node is pre-bound with a unique step ID, its business type, associated evaluation indicator number, and standard operating parameters. Each abnormal data point, after secondary verification through an isolated forest, carries: its associated indicator dimension number, collection time, maintenance section, and step code attribute. Based on the step ID carried by the abnormal data, the corresponding node in the maintenance directed graph is precisely matched, and the corresponding node is taken as the root cause node of the abnormality. If the step ID match is ambiguous, all candidate nodes bound to the same indicator in the directed graph are retrieved in reverse through the 9-dimensional evaluation indicator dimension type corresponding to the abnormal data. Based on the forward and backward flow relationships of the directed edges, the root cause node and the propagation associated node of the abnormality are locked. There can be more than one root cause node of the abnormality. For the root cause node of the anomaly, identify the evaluation indicators that exceed the range threshold, and determine the cause of the failure based on the type of evaluation indicator; the correspondence between evaluation indicators and failure causes is as follows: The causes of abnormal fault detection indicators are: incomplete fault feature collection, insufficient detection and identification sensitivity, and missed detection of hidden faults. The causes of failures corresponding to abnormal quality standards are: non-standard maintenance operations, unqualified parts assembly, and failure to strictly implement acceptance standards. The causes of failures corresponding to abnormal process compliance indicators are: failure to follow standardized operating procedures, process parameters deviating from the rated range, and missing key operating steps. The causes of abnormal unit resource consumption indicators are: excessive consumption of consumables, excessive energy consumption of equipment under no-load, and redundant waste of resources. The causes of abnormal resource utilization indicators are: idle manpower and equipment, imbalance in resource scheduling and matching, and inefficient use of workstation resources. The causes of the abnormal cost deviation rate index are: increased material purchase unit price, additional costs due to rework and repair, and deviations in budget control execution. The causes of abnormal time-related indicators are: single-stage operation timeout, waiting delays between processes, and delays in the connection between preceding and following processes; The abnormal time consumption percentage of the process is due to the following reasons: the time consumption of this process is abnormally high as a proportion of the total process time, or the cycle time of the upstream and downstream processes is mismatched. The abnormal maintenance delay rate indicates the following causes of failure: accumulated time exceeding the deadline in multiple stages, repeated repairs causing work stoppages, and delays in the supply of supporting materials. S5. Based on the weighted fusion algorithm, and combining the standardized dataset and the weights of each evaluation indicator, calculate the performance score of the logical link corresponding to the abnormal data and the overall maintenance performance index; identify inefficient links based on the performance score and the overall maintenance performance index, and automatically trigger the improvement process to form a closed-loop optimization mechanism. The detailed process of this step is as follows: S501. From the logical steps corresponding to the abnormal data, extract the standardized data corresponding to the logical steps, and combine them with the weights of the corresponding evaluation indicators. Calculate the performance score of the logical steps using a weighted summation method, as shown in the following expression: (5); in, For the first The performance score of the logical steps corresponding to each abnormal data point. for The first logical step corresponding to the first abnormal data The normalized value of each evaluation indicator. For the first The weight of each evaluation indicator; S502. Based on the performance score of each logical step and its weight, the overall maintenance performance index is calculated using a weighted summation method, as shown in the following expression: (6); in, The overall maintenance efficiency index. For the first The weight of the logical step corresponding to each abnormal data point; S503. Preset overall index threshold and process score threshold; when the overall maintenance efficiency index is less than the overall index threshold, the overall efficiency is judged to be low; when the logic process efficiency score is less than the process score threshold, the logic process is judged to be a low efficiency process. S504. Build a neural network prediction model to predict the future. The efficiency score of a logical step in the entire maintenance process at a given time point is calculated. If, in two consecutive maintenance processes, a certain logical step's performance is below a certain threshold... If the predicted values ​​are all lower than the score threshold of the process, the logic process will be judged as an inefficient process in advance. This embodiment uses an LSTM neural network to build the prediction model. The model's input is the normalized time-series feature values ​​of nine evaluation indicators for abnormal data, with a dimension of 9. The output is the future... The efficiency score of each logical step in the entire maintenance process at any given moment; The LSTM prediction model consists of an input layer, an LSTM layer, and a fully connected output layer. The LSTM layer has two layers. The first layer is used for shallow temporal feature extraction, specifically capturing short-term local temporal correlations and instantaneous change patterns between various indicators, outputting a complete temporal feature sequence. The second layer is used for deep temporal feature fusion and long-term dependency modeling. Specifically, based on the shallow features output from the first layer, it further explores long-term temporal evolution trends, periodic fluctuations in maintenance efficiency, and the linkage patterns between preceding and following processes. The mean squared error loss function and the Adam optimizer are used during training. The input data is divided into training and validation sets in an 8:2 ratio. The training process is as follows: Initialize the weight parameters of the prediction model, input the time-series features of the training set in batches, calculate the prediction performance score through forward propagation, calculate the loss value according to the loss function, optimize and update the model weights and bias parameters using the Adam optimizer during the training process, evaluate the prediction accuracy using the validation set, set the early stopping condition and the maximum number of iterations, and perform iterative training. When the early stopping condition is met or the maximum number of iterations is reached, the iteration stops, and the training ends. S505. Automatically generate improvement task sheets for inefficient processes; the improvement task sheet includes root cause analysis, responsible department, rectification deadline, rectification objectives, and rectification priority; the priority calculation formula is: (7); in, Priority The threshold for the stage score; S506. The system allocates resources according to priority. High-priority tasks are automatically matched with senior technical personnel as rectification personnel and given priority scheduling permissions. Low-priority tasks push standardized operation manuals and self-rectification guidelines to the rectification personnel in charge. The system will also provide similar cases for rectification personnel to refer to for rectification. The process for selecting similar cases is as follows: A knowledge base of improvement solutions is established in advance, and similar cases are matched using the cosine similarity algorithm; S507. Recalculate the inefficient processes after rectification. and Let the overall index threshold be... When the rectified inefficient processes meet the requirements and If the condition is met, it indicates that the closed-loop optimization of the inefficient link is complete; if it is not met, it indicates that the logic link is still an inefficient link, and rectification should be carried out again until the closed-loop optimization of the inefficient link is completed.

[0022] Example 2 like Figure 2 As shown in the illustration, and further explained in conjunction with Embodiment 1, this embodiment provides an equipment information management system for post-maintenance evaluation, used to implement the steps of the method in Embodiment 1, including: The maintenance directed graph model building module is used to divide the entire maintenance process into several logical steps and build a maintenance directed graph model.

[0023] The multi-dimensional evaluation indicator management module is used to establish an evaluation indicator system based on technical, economic, and time dimensions, and to determine the weight of each evaluation indicator through the analytic hierarchy process.

[0024] The multi-source maintenance data acquisition module is used to collect manually entered qualitative and quantitative data, as well as cost and working hour data integrated from industrial business systems, and to perform data verification and completion. The multi-source maintenance data acquisition module includes: a manual data entry unit, a system data integration unit, and a data verification and completion unit. The manual data entry unit is used to collect qualitative and quantitative data of the maintenance process through a structured form interface, and to perform mandatory field verification and modification log recording. The system data integration unit is used to automatically collect cost data and working hour data from external industrial business systems through standardized API interfaces; The data verification and completion unit is used to verify the range, logic and integrity of the collected data, and to complete and mark abnormal data using historical average values.

[0025] The data standardization processing module is used to standardize the collected multi-source maintenance data, including converting qualitative data into quantitative data and normalizing all quantitative data.

[0026] The efficiency calculation and closed-loop optimization module is used to calculate the efficiency score and overall maintenance efficiency index of the logical links corresponding to abnormal data based on standardized data and evaluation index weights, identify inefficient links and automatically trigger improvement processes to form a closed-loop optimization mechanism; the efficiency calculation and closed-loop optimization module includes: link efficiency calculation unit, inefficiency identification unit and improvement task generation and tracking unit; The process efficiency calculation unit is used to calculate the efficiency score of a logical process corresponding to abnormal data by combining its standardized data and the weights of the corresponding evaluation indicators using a weighted summation method. The low-efficiency identification unit is used to determine whether the overall efficiency is too low and to identify specific low-efficiency links based on preset overall efficiency thresholds and process efficiency thresholds. An improved task generation and tracking unit is used to automatically generate improvement task sheets for identified inefficient processes and track the rectification results for re-evaluation of efficiency.

[0027] The evaluation indicator management module is used to configure, maintain, and manage multi-dimensional evaluation indicators. The evaluation indicator management module includes: an indicator configuration unit, a weight calculation unit, and a weight storage and retrieval unit. The indicator configuration unit is used to define and maintain various evaluation indicators under the dimensions of technology, economy, and time efficiency. The weight calculation unit is used to construct a judgment matrix using the analytic hierarchy process (AHP) and to calculate and verify the weights of each evaluation index. The weight storage and retrieval unit is used to store the final indicator weight vector and provide weight data support for performance calculation.

[0028] The above embodiments are merely preferred technical solutions of the present invention and should not be considered as limitations on the present invention. The scope of protection of the present invention should be limited to the technical solutions described in the claims, including equivalent substitutions of the technical features described in the claims. That is, equivalent substitutions and improvements within this scope are also within the scope of protection of the present invention.

Claims

1. A method for managing equipment information for post-maintenance evaluation, characterized in that, Includes the following steps: S1. Divide the entire maintenance process of the equipment into several logical steps and construct a directed graph model for maintenance. S2. Obtain historical maintenance data, extract multi-dimensional evaluation indicators from the historical maintenance data, and calculate the weight of each evaluation indicator. S3. Collect multi-source maintenance data; S4. Standardize the collected multi-source maintenance data, map qualitative data to quantitative data through preset transformation rules, normalize all quantitative data according to indicator type to obtain a standardized dataset, and analyze the standardized dataset to identify abnormal data. S5. Based on the weighted fusion algorithm, combined with the standardized dataset and the weights of each evaluation indicator, calculate the performance score of the logical link corresponding to the abnormal data and the overall maintenance performance index; identify the inefficient links based on the performance score and the overall maintenance performance index, and automatically trigger the improvement process to form a closed-loop optimization mechanism.

2. The equipment information management method for post-maintenance evaluation according to claim 1, characterized in that, The detailed process of step S1 is as follows: S101. Obtain the entire maintenance process of the equipment under maintenance. The entire maintenance process includes the business process and the maintenance process. The business process and the maintenance process are divided into several steps according to the execution steps. S102. Configure a unique step ID for each step in the maintenance process, and establish a business description document for each step in the business process; associate the step ID of each step in the entire maintenance process with the corresponding business description document to form a logical step. S103. Set each logical link as a node, and connect each node with directed edges according to the flow relationship of each logical link to form a maintenance directed graph model. The node attributes include the stage ID, input parameters, and output parameters; the edge attributes of a directed edge include the flow trigger condition and the standard flow time. After each maintenance process is completed, the directed graph model collects the actual processing time of each logical step in the entire maintenance process and updates the standard processing time using the moving average method. The updated expression is as follows: （1）； in, The new standard takes time to process. This refers to the total number of maintenance processes that were statistically analyzed before updating the directed graph model. The original standard processing time, This represents the actual processing time of each logical step.

3. The equipment information management method for post-maintenance evaluation according to claim 2, characterized in that, The detailed process of step S2 is as follows: S201. Obtain historical maintenance data, divide the historical maintenance data into three dimensions, and extract the corresponding evaluation indicators from each dimension. The dimensions include: technological efficiency, economic efficiency, and time efficiency. Technical performance dimensions include: fault detection indicators, quality compliance indicators, and process conformity indicators; The economic efficiency dimension includes: unit resource consumption indicators, resource utilization rate indicators, and cost deviation rate indicators; Time efficiency dimensions include: time-based target indicators, time-consuming percentage of each process, and maintenance delay rate indicators; S202. The weights of each evaluation indicator are calculated using the analytic hierarchy process, as follows: For any dimension, the evaluation indicators within that dimension are numbered sequentially, and pairwise importance comparisons are performed on each evaluation indicator within that dimension to obtain importance scale values ​​between the evaluation indicators, thus constructing a judgment matrix for that dimension; the element values ​​of the judgment matrix are determined by the scale values ​​between the evaluation indicators corresponding to their respective numbers. The importance scale value between assessment indicators is determined by experts based on the importance relationship between the assessment indicators, and the importance scale value is determined by the importance relationship. For any dimension of the judgment matrix, we have: Calculate the largest eigenvalue and the corresponding eigenvector of the judgment matrix, normalize the eigenvector to obtain the weight of each evaluation index, and denote the normalized eigenvector as the weight vector of the evaluation index for that dimension. The consistency score for this dimension is calculated using the following expression: （2）； in, For consistency score, The largest eigenvalue, This represents the number of evaluation indicators for this dimension. Query Hierarchical Analysis Standard The table will show the number of evaluation metrics for that dimension. Corresponding average random consistency index Calculate the consistency ratio The expression is as follows: （3）； when If the value is less than 0.1, it indicates that the weight allocation of each evaluation indicator is reasonable; otherwise, step S202 is repeated until the condition is met. Less than 0.

1.

4. The equipment information management method for post-maintenance evaluation according to claim 1, characterized in that, Multi-source maintenance data includes two categories: manually entered data D1 and integrated data from industrial business systems D2; The data types for manually entered data D1 include: qualitative data D11 and quantitative data D12; The data types of the industrial business system integration data D2 include: cost data D21 and working hour data D22.

5. The equipment information management method for post-maintenance evaluation according to claim 4, characterized in that, The detailed process of step S4 is as follows: S401. Based on the preset qualitative and quantitative conversion rule library, convert the qualitative data in the multi-source maintenance data into quantitative data, and map D11 to quantitative data S1. The conversion method is as follows: assign a unique score to each type of qualitative data; The converted dataset is D3. S402. Merge the quantitative dataset D3 with the original quantitative data D12 to obtain the total quantitative dataset D4; Select the normalization algorithm according to the indicator type to convert the D4 data into normalized values, resulting in the standardized dataset D5; the indicator types include: positive indicators and negative indicators; S403. Calculate the mean and standard deviation of each evaluation indicator in D5; if the data exceeds the range threshold, the data is determined to be abnormal data. S404. Use the isolated forest algorithm to perform secondary verification on the abnormal data and output the probability of the abnormal data. At the same time, combine the maintenance directed graph model to trace the related nodes of the abnormal data and output the cause of the abnormality. The input to the Isolation Forest algorithm model is the normalized feature values ​​of nine evaluation indicators for outlier data, and the output is the outlier probability corresponding to each outlier data point in the input.

6. The equipment information management method for post-maintenance evaluation according to claim 3 or 5, characterized in that, The detailed process of step S5 is as follows: S501. From the logical steps corresponding to the abnormal data, extract the standardized data corresponding to the logical steps, and combine them with the weights of the corresponding evaluation indicators. Calculate the performance score of the logical steps using a weighted summation method, as shown in the following expression: （4）； in, For the first The performance score of the logical steps corresponding to each abnormal data point. for The first logical step corresponding to the first abnormal data The normalized value of each evaluation indicator. For the first The weight of each evaluation indicator; S502. Based on the performance score of each logical step and its weight, the overall maintenance performance index is calculated using a weighted summation method, as shown in the following expression: （5）； in, This refers to the overall maintenance efficiency index. For the first The weight of the logical step corresponding to each abnormal data point; S503. Preset overall index threshold and process score threshold; when the overall maintenance efficiency index is less than the overall index threshold, the overall efficiency is judged to be low; when the logic process efficiency score is less than the process score threshold, the logic process is judged to be a low efficiency process. S504. Build a neural network prediction model to predict the future. The efficiency score of a logical step in the entire maintenance process at a given time point is calculated. If, in two consecutive maintenance processes, a certain logical step... If the predicted values ​​are all lower than the score threshold of the process, the logic process will be judged as an inefficient process in advance. The model's input consists of normalized time-series feature values ​​of nine evaluation metrics for outlier data, with a 9-dimensional dimension. The output is the future... The efficiency score of each logical step in the entire maintenance process at any given moment; S505. Automatically generate improvement task sheets for inefficient processes; the improvement task sheet includes root cause analysis, responsible department, rectification deadline, rectification objectives, and rectification priority; the priority calculation formula is: （6）； in, Priority The threshold for the stage score; S506. The system allocates resources based on priority. High-priority tasks are automatically matched with senior technicians as rectification personnel and granted priority scheduling permissions. Low-priority tasks send standardized operation manuals and self-rectification guidelines to the rectification personnel in charge. The system recalculates after rectification is completed. and The system will also provide similar cases for rectification personnel, who can refer to these cases to carry out rectification. The process for selecting similar cases is as follows: A knowledge base of improvement solutions is established in advance, and similar cases are matched using the cosine similarity algorithm; S507. Recalculate the inefficient processes after rectification. and Let the overall index threshold be... When the rectified inefficient processes meet the requirements and When the condition is met, it indicates that the closed-loop optimization of the inefficient link is complete; when it is not met, rectification is carried out again until the closed-loop optimization of the inefficient link is complete.

7. An equipment information management system for post-maintenance evaluation, used to implement the steps of the method according to any one of claims 1 to 6, characterized in that, include: The maintenance directed graph model building module is used to divide the entire maintenance process into several logical steps and build a maintenance directed graph model. The multi-dimensional evaluation indicator management module is used to establish an evaluation indicator system based on technical, economic, and time dimensions, and to determine the weight of each evaluation indicator through the analytic hierarchy process. The multi-source maintenance data acquisition module is used to collect manually entered qualitative and quantitative data, as well as cost and working hour data integrated from industrial business systems, and to perform data verification and completion. The data standardization processing module is used to standardize the collected multi-source maintenance data, including converting qualitative data into quantitative data and normalizing all quantitative data. The efficiency calculation and closed-loop optimization module is used to calculate the efficiency score and overall maintenance efficiency index of the logical links corresponding to abnormal data based on standardized data and evaluation index weights, identify inefficient links and automatically trigger improvement processes to form a closed-loop optimization mechanism. The evaluation indicator management module is used to configure, maintain, and manage multi-dimensional evaluation indicators.

8. The equipment information management system for post-maintenance evaluation according to claim 7, characterized in that, The multi-source maintenance data acquisition module includes: a manual data entry unit, a system data integration unit, and a data verification and completion unit; The manual data entry unit is used to collect qualitative and quantitative data of the maintenance process through a structured form interface, and to perform mandatory field verification and modification log recording. The system data integration unit is used to automatically collect cost data and working hour data from external industrial business systems through standardized API interfaces; The data verification and completion unit is used to verify the range, logic and integrity of the collected data, and to complete and mark abnormal data using historical average values.

9. The equipment information management system for post-maintenance evaluation according to claim 7, characterized in that, The performance calculation and closed-loop optimization module includes: a process performance calculation unit, an inefficiency identification unit, and an improvement task generation and tracking unit; The process efficiency calculation unit is used to calculate the efficiency score of a logical process corresponding to abnormal data by combining its standardized data and the weights of the corresponding evaluation indicators using a weighted summation method. The low-efficiency identification unit is used to determine whether the overall efficiency is too low and to identify specific low-efficiency links based on preset overall efficiency thresholds and process efficiency thresholds. An improved task generation and tracking unit is used to automatically generate improvement task sheets for identified inefficient processes and track the rectification results for re-evaluation of efficiency.

10. The equipment information management system for post-maintenance evaluation according to claim 7, characterized in that, The evaluation indicator management module includes: an indicator configuration unit, a weight calculation unit, and a weight storage and retrieval unit; The indicator configuration unit is used to define and maintain various evaluation indicators under the dimensions of technology, economy, and time efficiency. The weight calculation unit is used to construct a judgment matrix using the analytic hierarchy process (AHP) and to calculate and verify the weights of each evaluation index. The weight storage and retrieval unit is used to store the final indicator weight vector and provide weight data support for performance calculation.

Images