A method and system for improving service efficiency of operation and maintenance personnel under the constraints of cognitive science

Abstract

The present application relates to the technical field of highway electromechanical operation and maintenance, and particularly relates to a method and system for improving service efficiency of operation and maintenance personnel under the constraint of cognitive science; individual characteristics of electromechanical operation and maintenance personnel are described in multiple dimensions through efficiency and label mapping; in view of the fact that the existing training system is designed homogeneously and ignores the individual knowledge blind spots or skill short boards of operation and maintenance personnel, it is difficult to improve the service efficiency of operation and maintenance personnel, and the individual knowledge gap quantification model is proposed; the individual knowledge gap or skill short board is mined based on the multi-dimensional portrait label, and a personalized training scheme for operation and maintenance personnel conforming to the laws of cognitive science is generated in combination with the knowledge graph and large model technology. The present application aims to optimize the whole-chain technical system closed loop from efficiency evaluation, knowledge gap diagnosis to operation and maintenance service efficiency improvement, and ensure the long-term solidification of knowledge.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of expressway electromechanical operation and maintenance technology, and in particular to an operation and maintenance personnel service efficiency improvement method and system under the constraint of cognitive science. BACKGROUND

[0002] As the core carrier of intelligent transportation, the expressway electromechanical system is the core support to ensure the safe, convenient and efficient operation of the expressway. The electromechanical operation and maintenance work covers many key links such as the toll collection system, the communication system, the monitoring system and the power supply and distribution system. With the exponential increase in operation and maintenance complexity, the traditional manual inspection combined with experience-based judgment mode has been difficult to meet the industry demand for minute-level fault response. Under this background, the service efficiency upgrade of operation and maintenance personnel has become a key bottleneck to ensure the reliability of the road network.

[0003] For a long time, training needs research has been the focus of researchers in the field of human resource development at home and abroad. Overall, training needs analysis is mainly divided into three kinds: subjective evaluation method, OTP model (Organization Task Person Model) and performance analysis model (Performance Analysis Model). However, in the actual work of expressway electromechanical operation and maintenance, the traditional operation and maintenance personnel service efficiency evaluation and optimization method exposes many problems to be solved, which seriously restricts the further improvement of the quality and efficiency of electromechanical operation and maintenance. First, the current expressway electromechanical operation and maintenance field generally adopts a unified training mode. The traditional training method does not evaluate personnel work efficiency based on historical work orders, operation records and other multi-dimensional business data to build a dynamic ability portrait, lacks a precise diagnosis mechanism for the dynamic knowledge gap of individuals in the monitoring, communication, toll collection and other multi-device systems, resulting in problems such as mismatch between training content and actual ability of operation and maintenance personnel, lack of targeting, etc. Second, the existing technology ignores the human memory and cognitive rules and fails to establish a scientific review and reinforcement mechanism, lacks consideration of individualization and cognitive science, making it difficult to maintain training effectiveness and resulting in low knowledge retention rate and repeated occurrence of similar device faults. These defects together cause the industry dilemma of rising waste rate of training resources and high recurrence rate of operation and maintenance faults, and there is an urgent need for a breakthrough solution that integrates cognitive science theory and intelligent diagnosis technology. SUMMARY

[0004] According to the first aspect of the present application, the present application claims a cognitive science constrained operation and maintenance personnel service efficiency improvement method, comprising the following steps:

[0005] Step S1: Construct a multi-level operation and maintenance service efficiency evaluation index system, calculate the index weight based on the analytic hierarchy process, and use the fuzzy comprehensive evaluation method to quantitatively evaluate the efficiency grade of the target operation and maintenance personnel to obtain the comprehensive efficiency score and grade;

[0006] Step S2: Based on the comprehensive performance score and the level and multi-source operation and maintenance behavior data, a hierarchical operation and maintenance personnel portrait label system is constructed, and a multi-dimensional portrait label and weight representing the ability characteristics of the target operation and maintenance personnel are generated and dynamically updated through fusion calculation of a topic model and a sequence neural network.

[0007] Step S3: Based on the preset post ability model and the multi-dimensional portrait label and weight, the weighted distance of the personal ability vector and the standard ability vector in multiple dimensions is calculated, and the severity of individual knowledge gap and skill short board is diagnosed and quantified.

[0008] Step S4: Based on the severity of the individual knowledge gap and skill short board, the field knowledge graph is retrieved to obtain the associated knowledge unit, and the large language model adapted to the field is driven to generate a structured personalized training course directory and content, and the learning and review nodes are planned according to the cognitive forgetting law, and a complete personalized training scheme is output.

[0009] Further, the step S1 of constructing the multi-level operation and maintenance service performance evaluation index system comprises:

[0010] The first level index covers four dimensions of knowledge and skill level, daily work performance, routine operation and maintenance inspection, and fault problem handling.

[0011] It is subdivided into ten second-level index items, and further decomposed into forty-two third-level specific evaluation index items, forming a three-level evaluation index system.

[0012] Further, the step S1 of quantitatively evaluating the performance level of the target operation and maintenance personnel by using the fuzzy comprehensive evaluation method comprises:

[0013] Based on the three-level evaluation index system, an evaluation factor set is constructed, and an evaluation set containing five performance levels is defined;

[0014] A semi-trapezoidal distribution function is used to map the measured values of each third-level index item to the membership degrees of each level in the evaluation set, and a fuzzy membership degree matrix is constructed;

[0015] The index weight set calculated based on the analytic hierarchy process is weighted and combined with the fuzzy membership degree matrix layer by layer to obtain a comprehensive membership degree vector;

[0016] The comprehensive membership degree vector and the preset level score vector are weighted and calculated to obtain a comprehensive performance score, and the final performance level is determined according to the score interval.

[0017] Further, the step S2 of constructing the hierarchical operation and maintenance personnel portrait label system comprises:

[0018] A label framework including knowledge dimension, behavior dimension and experience dimension is designed;

[0019] The label of the knowledge dimension is obtained by weighted calculation of the theoretical examination scores of various types of equipment and the corresponding equipment failure frequency;

[0020] The label of the behavior dimension is determined by analyzing historical work order records and calculating the proportion of operation processes that comply with standard operating procedures;

[0021] The label of the experience dimension is calculated by aggregating the complexity indicators of historical work orders, which are obtained by weighting the number of processing steps and the number of types of equipment involved.

[0022] Further, the multi-dimensional portrait label and weight representing the ability characteristics of the target operation and maintenance personnel generated and dynamically updated in step S2 include:

[0023] Based on the operation and maintenance behavior data of all operation and maintenance personnel, a global benchmark label weight is extracted using an unsupervised topic model;

[0024] For the behavior data of the target operation and maintenance personnel at a specific time point, an initial label weight at that time point is extracted using a supervised topic model;

[0025] The initial label weight and the global benchmark label weight are first weighted and fused, and the fusion result is used as a training target;

[0026] The feature vector of the behavior data at the specific time point is used as input, and a long short-term memory network is used for training to correct the initial label weight and obtain a corrected weight;

[0027] The corrected weight and the global benchmark label weight are second weighted and fused to obtain a comprehensive portrait weight;

[0028] The long short-term memory network is used again to predict the weight trend at a future time point based on a sequence of historical comprehensive portrait weights, and the labels are filtered according to a preset threshold to achieve adaptive updating of the label weight.

[0029] Further, the construction of the post ability model in step S3 includes:

[0030] Quantitative post ability benchmark values are set for the knowledge dimension, the behavior dimension, and the experience dimension;

[0031] The benchmark value of the knowledge dimension is set according to the level of equipment principle knowledge required by the post;

[0032] The benchmark value of the behavior dimension is set according to the threshold of the standard operating procedure compliance rate required by the post;

[0033] The benchmark value of the experience dimension is set according to the typical complex fault categories and quantities required by the post level.

[0034] Further, the step S3 of calculating the weighted distance specifically comprises:

[0035] Quantifying the actual performance value of the operation and maintenance personnel in the knowledge dimension, behavior dimension and experience dimension into a personal ability vector;

[0036] Calculating the square of the difference between each dimension of the personal ability vector and the standard ability vector;

[0037] Divide the square difference value of each dimension by the variance of the dimension to eliminate the dimension effect;

[0038] Weighted sum and square root of the variance normalized results of each dimension to obtain the final weighted Euclidean distance to quantify the ability gap.

[0039] Further, the step S4 of driving the large language model adapted to the field comprises:

[0040] Constructing a triple training data composed of scene description, knowledge gap and expert solution from historical work order, equipment manual field corpus;

[0041] Convert the triple into structured instruction data through a template, use low-rank adaptive technology, and only inject trainable parameters into part of the linear projection layer of the basic large language model for incremental fine-tuning;

[0042] In the fine-tuning process, introduce regularization constraints based on knowledge graph consistency to ensure the field accuracy of the generated content.

[0043] Further, the step S4 of generating a structured personalized training course directory and content comprises:

[0044] Arrange the training priority in descending order according to the severity of the knowledge gap, and search for associated knowledge points with logical association in the field knowledge graph from the knowledge gap with the highest priority as the starting point;

[0045] Organize the associated knowledge points according to the structure of basic concepts, operation processes and fault cases to form a course directory tree;

[0046] The step S4 of planning learning and review nodes in combination with the cognitive forgetting law comprises:

[0047] Take the knowledge point unit in the course directory tree as a basic learning unit;

[0048] According to the cognitive forgetting curve model, automatically plan and generate the subsequent review time points after each learning unit is completed;

[0049] Integrating the learning units and review nodes into the training timeline forms a complete training plan including initial learning and multiple review nodes.

[0050] According to the second aspect of the present application, the present application claims a service performance improvement system for operation and maintenance personnel under the constraint of cognitive science, comprising an efficiency evaluation unit, a user portrait unit, a knowledge gap diagnosis unit, a service performance improvement training unit, and a training effect evaluation and feedback unit.

[0051] The efficiency evaluation unit comprises operation and maintenance service performance evaluation index system management and target operation and maintenance personnel service capability evaluation. The index system management is used for the addition, deletion, modification, and query of the index and visual display.

[0052] The service capability evaluation is used for real-time collection of operation and maintenance work order response time, fault repair time, and maintenance record text. Based on the analytic hierarchy process and fuzzy comprehensive evaluation method, the operation and maintenance personnel's knowledge and skill level, daily work performance, routine operation and maintenance inspection quality and efficiency, and fault problem handling quality and efficiency are quantitatively evaluated and the results are visually displayed.

[0053] The user portrait unit comprises portrait tag system management and target operation and maintenance personnel single user portrait display. The portrait tag system management is used for the addition, deletion, modification, and query of the portrait tag and visual display.

[0054] The operation and maintenance personnel portrait is used to construct a single user multi-dimensional portrait containing a scenario-based identity label using a four-layer tag architecture based on the operation and maintenance service performance evaluation results, and to perform multi-dimensional visual display.

[0055] The knowledge gap diagnosis unit compares the difference between the target post capability matrix and the personnel skill level through an individual knowledge gap diagnosis model to identify the knowledge gap and skill short board of the operation and maintenance personnel.

[0056] The service performance improvement training unit comprises personalized training scheme generation and display of time node planning covering initial learning and multiple reviews, display of training course recommendation list for basic concepts, operation processes, or fault cases, and training content selection.

[0057] The training effect evaluation and feedback unit is used to collect the feedback opinions of the operation and maintenance personnel on the training, and to optimize and adjust the personalized training scheme in combination with the training effect evaluation results.

[0058] The service performance improvement system for operation and maintenance personnel under the constraint of cognitive science is used to execute the service performance improvement method for operation and maintenance personnel under the constraint of cognitive science.

[0059] The present application relates to the field of highway electromechanical operation and maintenance technology, and particularly relates to a method and system for improving service efficiency of operation and maintenance personnel under the constraint of cognitive science; individual characteristics of electromechanical operation and maintenance personnel are described in multiple dimensions through efficiency and label mapping; in view of the fact that the existing training system is designed homogeneously and ignores the individual knowledge blind spots or skill short boards of operation and maintenance personnel, it is difficult to improve the service efficiency of operation and maintenance personnel, an individual knowledge gap quantification model is proposed; individual knowledge gaps or skill short boards are mined based on multi-dimensional portrait labels, and a personalized training scheme of operation and maintenance personnel conforming to the laws of cognitive science is generated in combination with knowledge graph and large model technology. The present application aims to optimize the whole-chain technical system closed loop from efficiency evaluation, knowledge gap diagnosis to operation and maintenance service efficiency improvement, and ensure the long-term solidification of knowledge. BRIEF DESCRIPTION OF DRAWINGS

[0060] Figure 1 A workflow diagram of a method for improving service efficiency of operation and maintenance personnel under the constraint of cognitive science is requested to be protected by the embodiments of the present application.

[0061] Figure 2 A key indicator diagram of operation and maintenance service efficiency evaluation of a method for improving service efficiency of operation and maintenance personnel under the constraint of cognitive science is requested to be protected by the embodiments of the present application.

[0062] Figure 3 A whole process diagram of a portrait construction method of a method for improving service efficiency of operation and maintenance personnel under the constraint of cognitive science is requested to be protected by the embodiments of the present application.

[0063] Figure 4 A system architecture diagram of a system for improving service efficiency of operation and maintenance personnel under the constraint of cognitive science is requested to be protected by the embodiments of the present application. DETAILED DESCRIPTION

[0064] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of protection of the present application.

[0065] In this document, reference to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the application. The appearances of the phrase in various places in the specification are not necessarily all referring to the same embodiment, nor are they necessarily mutually exclusive of one another. Those skilled in the art will appreciate that the embodiments described herein are merely examples of the application and should not be construed as limiting the scope of the application.

[0066] According to the first embodiment of the present application, the present application claims a method for improving the service efficiency of operation and maintenance personnel under the constraints of cognitive science, referring to Figure 1 , comprising the following steps:

[0067] Step S1: Construct a multi-level operation and maintenance service efficiency evaluation index system, calculate the index weight based on the analytic hierarchy process, and use the fuzzy comprehensive evaluation method to quantitatively evaluate the efficiency level of the target operation and maintenance personnel, obtain the comprehensive efficiency score and grade;

[0068] Step S2: Based on the comprehensive efficiency score and grade and multi-source operation and maintenance behavior data, a hierarchical operation and maintenance personnel portrait label system is constructed, and a multi-dimensional portrait label and weight representing the ability characteristics of the target operation and maintenance personnel are generated and dynamically updated through the fusion calculation of topic model and sequence neural network;

[0069] Step S3: Based on the preset post ability model and the multi-dimensional portrait label and weight, the weighted distance of the personal ability vector and the standard ability vector in multiple dimensions is calculated, and the severity of individual knowledge gap and skill short board is diagnosed and quantified;

[0070] Step S4: Based on the severity of the individual knowledge gap and skill short board, retrieve the domain knowledge graph to obtain the associated knowledge units, and drive the large language model adapted to the domain to generate a structured personalized training course directory and content, plan learning and review nodes combined with the cognitive forgetting law, and output a complete personalized training scheme.

[0071] Further, the step S1 of constructing a multi-level operation and maintenance service efficiency evaluation index system comprises:

[0072] The first-level index covers four dimensions of knowledge and skill level, daily work performance, routine operation and maintenance inspection, and fault problem handling;

[0073] Subdivided into ten second-level index items, and further decomposed into forty-two third-level specific evaluation index items, forming a three-level evaluation index system.

[0074] Further, the step S1 of using the fuzzy comprehensive evaluation method to quantitatively evaluate the efficiency level of the target operation and maintenance personnel comprises:

[0075] Based on the three-level evaluation index system, an evaluation factor set is constructed, and an evaluation set containing five efficiency levels is defined;

[0076] A semi-trapezoidal distribution function is used to map the measured values of each third-level index item to the membership degree of each level in the evaluation set, and a fuzzy membership degree matrix is constructed;

[0077] The index weight set calculated based on the analytic hierarchy process is weighted and synthesized layer by layer with the fuzzy membership degree matrix to obtain a comprehensive membership degree vector;

[0078] The comprehensive membership vector is weighted and calculated with a preset level score vector to obtain a comprehensive performance score, and the final performance level is determined based on the score range.

[0079] In this embodiment, an evaluation index system for operation and maintenance service efficiency is constructed. Based on expert opinions and a review of relevant literature, this invention uses four aspects—knowledge and skills level, daily work performance, routine operation and maintenance inspections, and troubleshooting—comprising 10 major items and 42 sub-items, as key indicators for evaluating the operation and maintenance service efficiency of highway electromechanical maintenance personnel. Figure 2 A three-level indicator system for evaluating the operational efficiency of highway electromechanical maintenance personnel was constructed using a multi-dimensional assessment and step-by-step convergence approach.

[0080] Based on the actual needs of evaluating the operational efficiency of highway electromechanical maintenance personnel, each indicator in the indicator system is used as a factor set for fuzzy comprehensive evaluation. Let the evaluation factor set be... ,but The first representing the impact assessment object These factors ultimately correspond to 42 indicators in the three-tiered indicator layer. Then, based on the current performance evaluation, this invention divides the performance level of operation and maintenance services into five levels of performance evaluation standards, with each indicator's evaluation set set as follows:

[0081] ;

[0082] For the first, second, and third level indicator layers, experts in highway electromechanical maintenance were invited to conduct pairwise comparisons and scores of the importance of each indicator item using the 1-9 scaling method, and corresponding judgment matrices were constructed. Next, the judgment matrices were normalized to obtain the weight vector of each level indicator relative to its upper-level indicators. Finally, the weights at each level were synthesized layer by layer to obtain the final weights of each fourth-level indicator relative to the target layer P, and the comprehensive weight vector was set as follows: ;

[0083] The membership degree of each indicator to the evaluation set is determined by using a semi-trapezoidal distribution function, and a single-factor evaluation matrix is ​​constructed. Suppose that for a certain operations and maintenance personnel, there are 42 known tertiary indicators, and the measured values ​​of each tertiary indicator are mapped to a fuzzy membership degree matrix through a membership function. :

[0084]

[0085] By progressively weighting and synthesizing the weight sets and membership matrices of each level of indicators, a weighted average operator is used to obtain the comprehensive membership vector of the corresponding indicators. The results are as follows:

[0086] ;

[0087] The synthetic operation is adopted by using different fuzzy operators, and the actual situation and operation effect are determined.

[0088] In order to perform fine result output, a weighted score calculation is adopted, assuming that the service performance evaluation grade of the application corresponds to a score vector The operation and maintenance service performance evaluation comprehensive calculation score of a certain person is:

[0089]

[0090] Among them, The performance level is determined as excellent, The performance level is determined as good, The performance level is determined as qualified, The performance level is determined as needing improvement, The performance level is determined as unqualified.

[0091] Further, the step S2 of constructing the hierarchical operation and maintenance personnel portrait label system comprises:

[0092] A label framework including a knowledge dimension, a behavior dimension and an experience dimension is designed;

[0093] The label of the knowledge dimension is obtained by weighted calculation of the theoretical examination scores of various devices and the corresponding device fault frequencies;

[0094] The label of the behavior dimension is determined by analyzing historical work order records and calculating the proportion of operation processes that meet standard operation procedures;

[0095] The label of the experience dimension is calculated by aggregating the complexity indicators of historical work orders, and the complexity indicators are weighted by the number of processing steps and the number of involved device types.

[0096] Further, the step S2 of generating and dynamically updating the multi-dimensional portrait label and weight representing the ability characteristics of the target operation and maintenance personnel comprises:

[0097] Based on the operation and maintenance behavior data of all operation and maintenance personnel, a global benchmark label weight is extracted by using an unsupervised topic model;

[0098] For the behavior data of the target operation and maintenance personnel at a specific time point, an initial label weight at the time point is extracted by using a supervised topic model;

[0099] The initial label weight and the global benchmark label weight are first weighted and fused, and the fusion result is taken as a training target;

[0100] The feature vector of the specific time point behavior data is taken as input, and a long short-term memory network is used for training to correct the initial label weight to obtain a corrected weight;

[0101] The corrected weight and the global reference label weight are subjected to a second round of weighted fusion to obtain a comprehensive portrait weight;

[0102] The long short-term memory network is used again to predict the weight trend of the future time point based on the historical comprehensive portrait weight sequence, and the labels are filtered according to a preset threshold to realize adaptive updating of the label weight.

[0103] In this embodiment, a hierarchical label architecture is adopted to convert multi-dimensional data such as work records, training experiences, performance evaluations, and fault handling work orders of the expressway electromechanical maintenance personnel into an interpretable and operable portrait label system covering multiple dimensions such as knowledge level, behavior specification, and experience ability, and the label system is referred to Figure 3 to provide a basis for subsequent optimization of the maintenance service capability, and the portrait label system is designed as shown in Table 1.

[0104] Table 1: Portrait label system design table

[0105]

[0106] In order to extract the global information of the maintenance personnel's maintenance behavior, the maintenance behavior data of all maintenance personnel is abstracted into a document, and an unsupervised LDA topic model is used to extract the distribution of the implicit topic of the document as the global label information of all maintenance personnel, that is, the global reference label weight. Then, L-LDA is used to analyze the maintenance behavior data of each time point of the maintenance personnel to obtain the distribution of the maintenance personnel portrait label at each time point, and the topic probability is extracted as the portrait label weight;

[0107] The portrait label weight of each time point of the maintenance personnel obtained and the global reference label weight are subjected to weighted fusion, and this weight vector is taken as a learning target. The maintenance behavior data word vector of each time point of the maintenance personnel is taken as input, and the long short-term memory network is used to correct the weight obtained by L-LDA by virtue of the excellent characteristics of high abstraction to learn the portrait label weight considering the local behavior data of the maintenance personnel. The comprehensive portrait weight is obtained by fusing the portrait label weight with the global reference label weight. Finally, the long short-term memory network is used to predict the subsequent portrait weight by taking the comprehensive portrait weight as input, and the obtained labels are sorted, and finally the labels are filtered according to the set weight threshold, so that the portrait label and the label weight of the maintenance personnel can be adaptively updated.

[0108] Further, the construction of the post capability model in step S3 includes:

[0109] Set quantitative job capability benchmark values for the knowledge dimension, behavior dimension, and experience dimension;

[0110] The benchmark value of the knowledge dimension is set according to the level of equipment principle knowledge required to be mastered by the post;

[0111] The benchmark value of the behavior dimension is set according to the standard operating procedure compliance rate threshold required to be reached by the post;

[0112] The benchmark value of the experience dimension is set according to the number of typical complex faults required to be handled by the post level.

[0113] Further, the step S3 of calculating the weighted distance specifically includes:

[0114] Quantify the actual performance values of the operation and maintenance personnel in the knowledge dimension, behavior dimension, and experience dimension into a personal capability vector;

[0115] Calculate the square of the difference value between each dimension of the personal capability vector and the standard capability vector;

[0116] Divide the square difference value of each dimension by the variance of the dimension to eliminate the dimension effect;

[0117] Weighted sum and square root of the variance normalized results of each dimension to obtain the final weighted Euclidean distance, which is used to quantify the capability gap.

[0118] In this embodiment, according to the designed portrait label system, the job capability benchmark is modeled, and the standard capability vector is :

[0119]

[0120] The post standard of equipment principle mastery ; the post standard of behavior specification represents the SOP compliance rate threshold; represents the number of complex fault handling corresponding to the post level.

[0121] Each operation and maintenance personnel constructs a personal operation and maintenance service capability vector :

[0122]

[0123] The actual mastery of equipment principle ; the behavior specification represents the actual SOP compliance rate; represents the ratio of the actual complex fault handling amount to the complex fault handling amount corresponding to the post level.

[0124] The difference between the individual and the operation and maintenance service capability of the post is quantified by a weighted Euclidean distance, so that:

[0125]

[0126] wherein, is the standard deviation of the dimension, used for normalizing the dimension. When , it represents that the capability meets the standard, , it represents a slight defect, , it represents a serious defect.

[0127] By comparing the individual portrait label of the operation and maintenance personnel with the post capability model, the gap between the individual in each dimension label and the post requirement is found to establish the training benchmark, that is, the knowledge gap and skill short board of the individual.

[0128] Further, the step S4 of driving the large language model adapted to the field includes:

[0129] Constructing a triple training data composed of scene description, knowledge gap and expert solution from historical work orders and device manual field corpus;

[0130] Convert the triple into structured instruction data through a template, and use low-rank adaptive technology to inject trainable parameters into part of the linear projection layer of the basic large language model for incremental fine-tuning;

[0131] In the fine-tuning process, a regularization constraint based on knowledge graph consistency is introduced to ensure the field accuracy of the generated content.

[0132] Further, the step S4 of generating a structured individualized training course directory and content includes:

[0133] According to the descending order of the severity of the knowledge gap, arrange the training priority, and take the knowledge gap with the highest priority as the starting point to search for associated knowledge points with logical association in the field knowledge graph;

[0134] Organize the associated knowledge points according to the structure of basic concepts, operation processes and fault cases to form a course directory tree;

[0135] The step S4 of planning learning and review nodes in combination with the cognitive forgetting law includes:

[0136] Take the knowledge point unit in the course directory tree as a basic learning unit;

[0137] According to the cognitive forgetting curve model, automatically plan and generate the review time points after each learning unit is completed;

[0138] Integrate the learning units and review nodes into the training timeline to form a complete training plan that includes initial learning and multiple review nodes.

[0139] Wherein, in this embodiment, the low-rank adaptation (LoRA) technology is adopted, and trainable parameters are only injected into the query and value projection matrices of the Transformer layer. Specifically, two trainable matrices and are introduced to form the parameter update expression . Wherein, the rank is significantly smaller than the original dimension. The training data construction adopts the domain adaptation instruction fine-tuning paradigm: <scene, knowledge gap, expert answer> triples are extracted from historical work orders, device manuals and other raw corpus, and are converted into structured instruction data through a template engine. For example, fault handling records are converted into instruction formats such as "generate training content containing [knowledge points] for [user level] [device type] appearing [fault phenomenon]". A knowledge consistency regularization term is introduced during training to calculate the cosine similarity between the extracted entity relationships in the generated text and the knowledge graph, forming a hybrid loss function: ;

[0140] The expressway mechanical and electrical maintenance knowledge graph is introduced as an external knowledge base of the GraphRAG module in the large model applied to expressway mechanical and electrical maintenance, to establish a logical mapping between fault modes, solutions and device types; a GraphRAG engine based on graph neural networks is designed to generate subgraph retrieval paths using fault association chains between work orders; historical work order cases are then encoded as (fault feature-solution) knowledge triples for domain pre-training, and then the large model is incrementally fine-tuned through new work order features;

[0141] The educational cognitive theory and domain practice experience are encoded into computable generation constraints, such as inputting a knowledge gap list and arranging priorities in descending order of gap values; then a content generation framework is developed for work order feature perception, to retrieve associated knowledge units from the knowledge graph and generate a course directory tree; the Ebbinghaus forgetting curve is combined with the large model to plan learning-review nodes; and finally, a service efficiency improvement plan for maintenance personnel that conforms to the laws of cognitive science is generated.

[0142] According to the second embodiment of the present application, the present application claims a service efficiency improvement system for maintenance personnel under the constraints of cognitive science, comprising an efficiency evaluation unit, a user portrait unit, a knowledge gap diagnosis unit, a service efficiency improvement training unit, and a training effect evaluation and feedback unit.

[0143] The performance evaluation unit includes operation and maintenance service performance evaluation index system management and target operation and maintenance personnel operation and maintenance service capability evaluation, and the index system management is used for adding, deleting, modifying and inquiring the index and visual display.

[0144] The service capability evaluation is used for collecting operation and maintenance work order response time, fault repair time and maintenance record text in real time, and realizing quantitative evaluation and result visual display of operation and maintenance personnel knowledge and skill level, daily work performance, conventional operation and maintenance inspection quality and efficiency and fault problem processing quality based on analytic hierarchy process and fuzzy comprehensive evaluation method.

[0145] The user portrait unit includes portrait label system management and single user portrait display of the target operation and maintenance personnel, and the portrait label system management is used for adding, deleting, modifying and inquiring the portrait label and visual display.

[0146] The portrait of the operation and maintenance personnel is used for constructing a single user multi-dimensional portrait containing a scenario-based identity label by using a four-layer label architecture according to the operation and maintenance service performance evaluation result, and performing multi-dimensional visual display.

[0147] The knowledge gap diagnosis unit is used for identifying the knowledge gap and skill short board of the operation and maintenance personnel by comparing the difference between the target post capability matrix and the personnel skill level through an individual knowledge gap diagnosis model.

[0148] The service performance improvement training unit includes personalized training scheme generation and display covering time node planning of initial learning and multiple reviews, display of a training course recommendation list for basic concepts, operation processes or fault cases, and training content selection.

[0149] The training effect evaluation and feedback unit is used for collecting feedback opinions of the operation and maintenance personnel on the training, and optimizing and adjusting the personalized training scheme in combination with the training effect evaluation result.

[0150] The operation and maintenance personnel service performance improvement system under the constraint of cognitive science is used for executing the operation and maintenance personnel service performance improvement method under the constraint of cognitive science.

[0151] In several embodiments provided in the present application, it should be understood that the disclosed system, device and method can be implemented in other ways.

[0152] In addition, the various functional units in the embodiments of the present application can be integrated in one processing unit, or each can exist as an independent physical unit, or two or more than two of them can be integrated in one physical unit. The above-mentioned integrated unit can be implemented in the form of hardware, or in the form of a software functional unit. The above is only an embodiment of the present application, and does not limit the patent scope of the present application, and any equivalent structure or equivalent process transformation using the content of the specification and drawings of the present application, or direct or indirect application in other related technical fields, are also included in the patent protection scope of the present application.

[0153] The specific embodiments of the application are described above, but it is only as an example, and the present application is not limited to the specific embodiments described above. Any equivalent modification or replacement of the present application for those skilled in the art is also within the scope of the present application, therefore, any equivalent transformation and modification, improvement, etc. made without departing from the spirit and principle range of the present application should be covered in the scope of the present application.

Claims

1. A method for improving the service efficiency of operations and maintenance personnel under the constraints of cognitive science, characterized in that, Includes the following steps: Step S1: Construct a multi-level operation and maintenance service performance evaluation index system, calculate the index weights based on the analytic hierarchy process, and use the fuzzy comprehensive evaluation method to quantitatively evaluate the performance level of the target operation and maintenance personnel to obtain the comprehensive performance score and level. Step S2: Based on the comprehensive performance score and level and multi-source operation and maintenance behavior data, construct a hierarchical operation and maintenance personnel profile label system. Through the fusion calculation of topic model and sequence neural network, generate and dynamically update multi-dimensional profile labels and weights representing the capability characteristics of the target operation and maintenance personnel. Step S3: Based on the preset job competency model and the multi-dimensional profile labels and weights, calculate the weighted distance between the individual competency vector and the standard competency vector in multiple dimensions, and diagnose and quantify the severity of individual knowledge gaps and skill deficiencies. Step S4: Based on the severity of the individual's knowledge gaps and skill deficiencies, retrieve the domain knowledge graph to obtain related knowledge units, and drive the domain-adaptive optimized large language model to generate a structured personalized training course catalog and content. Combine the cognitive forgetting law to plan learning and review nodes, and output a complete personalized training program. The step S4, which drives the domain-adaptive large language model, includes: Construct training data consisting of triplets, including scenario descriptions, knowledge gaps, and expert solutions, from domain corpora of historical work orders and equipment manuals; The triples are transformed into structured instruction data through templates. Low-rank adaptive technology is used to inject trainable parameters into only a portion of the linear projection layers of the basic large language model for incremental fine-tuning. The incremental fine-tuning is achieved by introducing a highway electromechanical maintenance knowledge graph as an external knowledge base for the GraphRAG module in the large model of highway electromechanical maintenance, and establishing a logical mapping between fault modes, solutions and equipment types. Design a GraphRAG engine based on graph neural networks to generate subgraph retrieval paths using fault association chains between work orders; Then, historical work order cases are encoded into fault feature-solution knowledge triples for domain pre-training, and then the large model is incrementally fine-tuned by new work order features. During the fine-tuning process, regularization constraints based on knowledge graph consistency are introduced to ensure the domain accuracy of the generated content. Step S4 generates a structured, personalized training course catalog and content, including: Training priorities are sorted in descending order of the severity of the knowledge gap. Starting with the knowledge gap with the highest priority, a related subgraph search is performed in the domain knowledge graph to find related knowledge points with logical connections. The related knowledge points are organized according to the structure of basic concepts, operation procedures, and failure cases to form a course catalog tree; Step S4, which plans learning and review nodes based on cognitive forgetting patterns, includes: Use the knowledge point units in the course catalog tree as basic learning units; Based on the cognitive forgetting curve model, after each learning unit is completed, subsequent review time points are automatically planned and generated; The learning units and review points are integrated into the training timeline to form a complete training plan that includes initial learning and multiple review points.

2. The method for improving the service efficiency of operation and maintenance personnel under the constraints of cognitive science as described in claim 1, characterized in that, Step S1 involves constructing a multi-level operation and maintenance service performance evaluation index system, including: The first-level indicators cover four dimensions: knowledge and skills level, daily work performance, routine operation and maintenance inspection, and troubleshooting. It is further subdivided into ten secondary indicators, and then into forty-two tertiary specific evaluation indicators, forming a three-level evaluation indicator system.

3. The method for improving the service efficiency of operation and maintenance personnel under the constraints of cognitive science as described in claim 2, characterized in that, Step S1 employs fuzzy comprehensive evaluation to quantitatively assess the performance level of the target maintenance personnel, including: Based on the aforementioned three-level evaluation index system, an evaluation factor set is constructed, and an evaluation set containing five effectiveness levels is defined. A fuzzy membership matrix is ​​constructed by mapping the measured values ​​of each third-level indicator to the membership degree of each level in the evaluation set using a semi-trapezoidal distribution function. The index weight set calculated based on the analytic hierarchy process is combined with the fuzzy membership matrix through a layer-by-layer weighted synthesis operation to obtain the comprehensive membership vector. The comprehensive membership vector is weighted and calculated with a preset level score vector to obtain a comprehensive performance score, and the final performance level is determined based on the score range.

4. The method for improving the service efficiency of operation and maintenance personnel under the constraints of cognitive science as described in claim 1, characterized in that, Step S2 involves constructing a hierarchical profile and tagging system for operations and maintenance personnel, including: Design a tag framework that includes knowledge, behavior, and experience dimensions; The labels for the knowledge dimension are obtained by weighting the theoretical test scores of various types of equipment with their corresponding equipment failure frequencies. The labels for the behavioral dimensions are determined by analyzing historical work order records and calculating the proportion of operation processes that conform to standard operating procedures. The labels for the experience dimension are calculated by aggregating the complexity index of historical work orders. The complexity index is derived by weighting the number of processing steps and the number of equipment types involved.

5. The method for improving the service efficiency of operation and maintenance personnel under the constraints of cognitive science according to claim 1, characterized in that, Step S2 generates and dynamically updates multi-dimensional profile labels and weights representing the capability characteristics of the target maintenance personnel, including: Based on the operational behavior data of all operations and maintenance personnel, an unsupervised topic model is used to extract global baseline label weights; For the behavioral data of the target maintenance personnel at a specific time point, a supervised topic model is used to extract the initial label weights for that time point; The initial label weights and the global baseline label weights are weighted and fused in the first round, and the fusion result is used as the training target. Using the feature vector of the behavioral data at the specific time point as input, a long short-term memory network is used for training to correct the initial label weights and obtain the corrected weights. The corrected weights are then combined with the global baseline label weights in a second round of weighted fusion to obtain the comprehensive profile weights. By leveraging the Long Short-Term Memory network again, the weight trend of future time points is predicted using the historical comprehensive profile weight sequence, and the labels are filtered according to a preset threshold to achieve adaptive updating of label weights.

6. The method for improving the service efficiency of operation and maintenance personnel under the constraints of cognitive science as described in claim 1, characterized in that, The construction of the job competency model in step S3 includes: Quantitative benchmark values ​​for job competencies are set for the knowledge, behavior, and experience dimensions. The baseline value for the knowledge dimension is set based on the level of knowledge of equipment principles required for the job position. The baseline value for the behavioral dimension is set based on the standard operating procedure compliance rate threshold required for the job position. The baseline value for the experience dimension is set based on the typical and numerous complex fault types required to be handled at the job level.

7. The method for improving the service efficiency of operation and maintenance personnel under the constraints of cognitive science as described in claim 1, characterized in that, The calculation of the weighted distance in step S3 specifically includes: The actual performance values ​​of operations and maintenance personnel in the knowledge, behavior and experience dimensions are quantified into individual ability vectors. Calculate the square of the difference in each dimension between the individual ability vector and the standard ability vector; Divide the squared difference of each dimension by the variance of that dimension to eliminate the influence of dimensions; The results of each dimension after variance normalization are weighted, summed, and squared to obtain the final weighted Euclidean distance, which is used to quantify the capability gap.

8. A system for improving the service efficiency of operation and maintenance personnel under the constraints of cognitive science, characterized in that, It includes a performance evaluation unit, a user profiling unit, a knowledge gap diagnosis unit, a service performance improvement training unit, and a training effectiveness evaluation and feedback unit; The performance evaluation unit includes the management of the operation and maintenance service performance evaluation indicator system and the evaluation of the operation and maintenance service capabilities of the target operation and maintenance personnel. The indicator system management is used for adding, deleting, modifying, querying and visualizing the indicators. Service capability assessment is used to collect real-time operation and maintenance work order response time, fault repair time and maintenance record text. Based on the analytic hierarchy process and fuzzy comprehensive evaluation method, it realizes the quantitative evaluation and result visualization of the knowledge and skills level of operation and maintenance personnel, daily work performance, routine operation and maintenance inspection quality and efficiency, and fault problem handling quality and efficiency. The user profiling unit includes the management of the profile tag system and the display of single-user profiles of target operation and maintenance personnel. The profile tag system management is used for adding, deleting, modifying, querying and visualizing profile tags. The profiles of operations and maintenance personnel are used to construct multi-dimensional profiles of single users with contextualized identity identifiers based on the performance evaluation results of operations and maintenance services, using a four-layer tag architecture, and then to display them in a multi-dimensional visualization. The knowledge gap diagnosis unit uses an individual knowledge gap diagnosis model to compare the difference between the target job competency matrix and the personnel skill level to identify the knowledge gaps and skill deficiencies of operation and maintenance personnel. The service efficiency improvement training unit includes the generation and presentation of personalized training programs covering the time nodes for initial learning and multiple reviews, the presentation of a recommended list of training courses for basic concepts, operating procedures or fault cases, and the selection of training content. The training effectiveness evaluation and feedback unit is used to collect feedback from operation and maintenance personnel on the training, and to optimize and adjust the personalized training plan based on the training effectiveness evaluation results. The aforementioned system for improving the service efficiency of operations and maintenance personnel under cognitive science constraints is used to execute the method for improving the service efficiency of operations and maintenance personnel under cognitive science constraints as described in any one of claims 1-7.

Images