A multi-terminal integrated project and maintenance information management platform and method

Abstract

This invention discloses a multi-terminal integrated project and maintenance information management platform and method, relating to the field of maintenance information management technology. The multi-terminal integrated project and maintenance information management platform and method includes: S1, acquiring raw data of maintenance approval behavior and preprocessing the raw data; S2, quantifying the consistency level of decision-making in approval behavior expression, generating expression type labels and expression anomaly template matching markers for approval behavior; S3, constructing a set of expression behavior profile indicators, evaluating the expression bias characteristics of approvers, classifying approvers into expression behavior bias levels, and matching maintenance process response strategies; S4, constructing a work order risk analysis input set using work orders as analysis units, evaluating the abnormal risk value of each work order, classifying work orders into risk levels, and matching maintenance review intervention mechanisms. This solves the problem of low management transparency and weak risk identification capabilities caused by unstructured expression of maintenance approval opinions.

Description

Technical Field

[0001] This invention relates to the field of maintenance information management technology, specifically to a multi-terminal integrated project and maintenance information management platform and method. Background Technology

[0002] Currently, maintenance and repair work plays a crucial role in ensuring operational safety and efficiency in fields such as rail transit, energy equipment, and large industrial systems. With the increasing sophistication of enterprise operation and maintenance management, the generation, approval, and execution of maintenance work orders have gradually shifted from manual, offline processes to digital processing based on information platforms. Existing technologies have included research on monitoring equipment operating status and fault prediction modeling.

[0003] For example, the invention with publication number CN118643414A provides a maintenance and management method and system for gravure printing machines, relating to the field of printing machine technology. The method includes: collecting operating parameters of key components of the gravure printing machine through IoT sensors; preprocessing the operating parameters to determine the corresponding operating characteristics; inputting the operating characteristics into a pre-built fault diagnosis model to determine the diagnostic confidence level of the operating characteristics; labeling operating characteristics with diagnostic confidence levels lower than a preset diagnostic threshold through active learning; acquiring the operating parameters of key components of the gravure printing machine in real time; updating the fault diagnosis model online through incremental learning; and combining an adaptive weight adjustment strategy to balance the influence of real-time input operating parameters and historical parameters on the fault diagnosis model, iteratively optimizing the performance of the fault diagnosis model until a preset number of iterations is reached.

[0004] For example, the invention with announcement number CN116166194A discloses a method and system for equipment support, maintenance and repair management. The method includes: generating a maintenance work order A for equipment A by a server, wherein the server does not update the state of equipment A in the main memory when generating maintenance work order A; reading the state of equipment A from the main memory by the server; updating the state of equipment A by the server so that the state of equipment A changes from state A to state B, and storing the state B of equipment A in a first low-cost memory by the server; sending maintenance work order A to maintenance personnel by the server; and updating the state B of equipment A in the first low-cost memory to state A if one of the following conditions is met: the server receives feedback from maintenance personnel indicating that maintenance is not required; or the server receives feedback from maintenance personnel indicating that maintenance is completed within a predetermined time.

[0005] However, the above solutions mainly focus on the equipment status level, lacking semantic modeling and quantitative control mechanisms for the approval behavior itself. In addition, most current maintenance platforms have not established structural diagnostic pathways for approval language expression, making it impossible to achieve process supervision of the expression quality at approval nodes, nor have they formed a systematic modeling mechanism for the bias of the approver's expression behavior, let alone quantitatively assess and intervene in the risks caused by abnormal expression throughout the entire work order process.

[0006] Therefore, in order to address the above issues, there is an urgent need for a multi-terminal integrated project and maintenance information management platform and method. Summary of the Invention

[0007] Technical problems to be solved

[0008] To address the shortcomings of existing technologies, this invention provides a multi-terminal integrated project and maintenance information management platform and method, which solves the problems of low management transparency and weak risk identification capabilities caused by unstructured expression of maintenance approval opinions.

[0009] Technical solution

[0010] To achieve the above objectives, the present invention provides the following technical solution: a multi-terminal integrated project and maintenance information management platform and method, comprising: S1, acquiring raw data of maintenance approval behavior, and preprocessing the raw data of maintenance approval behavior to obtain preprocessed raw data of maintenance approval behavior; S2, based on the preprocessed maintenance approval behavior data, extracting approval semantic elements and establishing structural relationships, quantifying the consistency level of expression decision of approval behavior, and generating expression type labels and expression anomaly template matching marks for approval behavior; S3, constructing an expression behavior profile indicator set with approvers as the analysis unit, evaluating the expression bias characteristics of approvers, classifying the expression behavior bias level of approvers based on the evaluation results, and matching maintenance process response strategies; S4, constructing a work order risk analysis input set with work orders as the analysis unit, evaluating the abnormal risk value of each work order, classifying the risk level of work orders based on the evaluation results, and matching maintenance review intervention mechanisms.

[0011] Furthermore, the original data of maintenance and repair approval behavior is obtained and preprocessed. The specific steps for obtaining the preprocessed original data are as follows: The maintenance and repair approval behavior data includes work order number, approval response time, approval processing time, approval conclusion, approval opinion, and the maximum length limit of the approval opinion field; the time sequence of the maintenance and repair approval behavior data is corrected using a timestamp alignment reconstruction algorithm to eliminate response record delays and offsets; anomaly cleaning is performed on the maintenance and repair approval behavior data using duplicate submission identification and rejection frequency clustering algorithms to remove irregular flow records and extreme behavior samples; the validity of the maintenance and repair approval behavior data is judged using a semantic consistency judgment algorithm between approval opinions and approval conclusions to identify samples with no expressive opinions and semantic conflicts; and the maintenance and repair approval behavior data is standardized and scaled by performing a unified dimensional conversion using a normalization compression algorithm.

[0012] Furthermore, based on the preprocessed maintenance approval behavior data, the specific steps for extracting approval semantic elements and establishing structural relationships to quantify the consistency level of the expression decision-making in approval behavior are as follows: Based on the preprocessed maintenance approval behavior data, the approval opinion and approval conclusion fields are extracted, and a one-to-one correspondence based on the work order number is established as the basic input for structured parsing; the word count of the approval opinion field is counted, and the length of the approval opinion field is extracted; a judgment on whether the approval opinion and the approval are consistent is generated through text comparison, with consistency marked as 1 and inconsistency marked as 0, generating a consistency marker between opinion and conclusion; the ratio of the length of the approval opinion field to the upper limit of the length of the approval opinion field is calculated, and the ratio is subtracted from 1 to obtain the approval text refinement item; the approval response time is divided by the sum of the approval response time and the approval processing time to obtain the approval response efficiency item; the approval text refinement item and the approval response efficiency item are added together, and then the consistency marker between the approval opinion and the approval conclusion is added to obtain the expression decision-making consistency evaluation value.

[0013] Furthermore, the specific steps for generating expression type labels and expression anomaly template matching tags for approval behaviors are as follows: Real-time comparison of the expression decision consistency assessment value and the decision consistency threshold; dividing the approval behavior into intervals; and generating expression type labels: If the expression decision consistency assessment value is greater than or equal to the first-level decision consistency threshold, the approval behavior is labeled as having a clear expression tendency, and the expression tendency is marked as 2; if the expression decision consistency assessment value is less than the first-level decision consistency threshold but greater than the second-level decision consistency threshold, the approval behavior is labeled as having an ambiguous expression, and the expression tendency is marked as 1; if the expression decision consistency assessment value is less than or equal to the second-level decision consistency threshold... If the value is not specified, the approval behavior is marked as expression deviation type, and the expression tendency is marked as 0. For the approval behavior marked as expression deviation type, the corresponding approval opinions are extracted to construct an expression deviation type corpus set. Fixed-length sliding window segmentation and high-frequency phrase clustering are performed in the corpus set to extract the main expression fragments that appear more than the corpus threshold consecutively, generating an expression abnormal phrase set. The expression templates in the expression abnormal phrase set are matched with historical approval opinions. If the match is successful and the expression decision consistency evaluation value is consistently lower than the secondary decision consistency threshold, the template is marked as an expression abnormal template and the expression abnormal template matching is marked as 1; otherwise, it is marked as 0.

[0014] Furthermore, the specific steps for constructing an expression behavior profile indicator set using approvers as the analysis unit to assess the expression bias characteristics of approvers are as follows: Using approvers as the unit, summarize all approval records corresponding to each approver to construct an expression behavior profile indicator set: extract the number of records with clear expression tendencies, ambiguous expression tendencies, and deviating expression tendencies; calculate the expression deviation rate, expression ambiguity rate, and expression clarity rate; based on the expression decision consistency evaluation value of the approval records, calculate the average consistency value between opinions and approval conclusions, the mean of the expression decision consistency evaluation value, and the variance of the expression consistency evaluation value; count the number of approval records that hit the expression anomaly template to obtain the hit expression... The total historical frequency of abnormal templates; based on the expression behavior profile indicator set, assess the approver's expression bias characteristics: calculate the sum of the expression clarity rate and the average consistency between opinion and conclusion, divide by the sum of the expression deviation rate and the expression ambiguity rate, and add one to obtain the expression clarity ratio; calculate the mean of the expression decision consistency assessment value, divide by the square root of the variance of the expression consistency assessment value, and add one to obtain the expression stability adjustment term; calculate one and add the reciprocal of the sum of the historical frequency of hitting the expression abnormal template to obtain the expression abnormality compression factor; multiply the expression clarity ratio, expression stability adjustment term, and expression abnormality compression factor to obtain the behavioral expression bias assessment value.

[0015] Furthermore, the specific steps for classifying approvers' expression behavior bias levels based on the assessment results and matching them with maintenance process response strategies are as follows: Based on the expression behavior bias assessment value, approvers are classified into expression behavior bias levels and maintenance process response strategies are matched accordingly: If the expression behavior bias assessment value is greater than or equal to the Level 1 expression behavior threshold, the approver is marked as having clear and stable expression, and is granted normal access in the maintenance process without triggering any intervention strategies; if the expression behavior bias assessment value is between the Level 1 and Level 2 expression behavior thresholds, the approver is marked as having moderate expression deviation, and a decision opinion structure verification prompt is enabled for the associated maintenance work orders; if the expression behavior bias assessment value is less than or equal to the Level 2 expression behavior threshold, the approver is marked as having high-risk expression deviation, the approver's independent approval authority is restricted, and a high-risk expression reminder and multi-level review process linkage mechanism is activated for the associated maintenance work orders, automatically transferring them to the targeted review path.

[0016] Furthermore, the specific steps for constructing the work order risk analysis input set using work orders as the analysis unit are as follows: Taking work orders as the unit, summarize all approval node information in the current work order to construct the work order risk analysis input set: extract the total number of approval nodes and the number of valid approval nodes in the current work order; count the number of approval nodes with expression deviation in the current work order, and record them as the number of expression deviation nodes; count the number of approval nodes that hit the expression anomaly template, and record them as the number of anomaly template hit nodes; obtain the consistency mark between opinion and conclusion, the clear expression tendency mark, and the real-time behavioral expression bias assessment value of the corresponding approver for each valid approval node.

[0017] Furthermore, the specific steps for assessing the abnormal risk value of each work order are as follows: Based on the work order expression risk analysis input set, assess the abnormal risk value of each work order: calculate the sum of the number of expression deviation nodes and the number of nodes hit by the abnormal template, and divide by the total number of approval nodes plus one to obtain the structural abnormality amplification item; for all valid approval nodes in the current work order, calculate the product of the inverse value of the opinion and conclusion consistency marker and the inverse value of the expression tendency marker for each node, and then multiply it by the reciprocal of the behavior expression bias assessment value of the approver corresponding to the node plus one, sum and divide by the total number of nodes to obtain the node expression risk mean item; multiply the structural abnormality amplification item and the node expression risk mean item to obtain the work order abnormal risk value.

[0018] Furthermore, the specific steps for classifying work orders by risk level based on the assessment results and matching them with the maintenance audit intervention mechanism are as follows: Based on the abnormal risk value of the work order, the work order risk level is classified: if the abnormal risk value is greater than or equal to the Level 1 risk threshold, the work order is marked as a high-risk work order; if the abnormal risk value is between the Level 1 and Level 2 risk thresholds, the work order is marked as a medium-risk work order; if the abnormal risk value is less than or equal to the Level 2 risk threshold, the work order is marked as a low-risk work order; the risk level marking is written into the work unit data; based on the work order… The risk level results are automatically matched with risk warning strategies and maintenance review linkage mechanisms: For work orders with high risk of expression, a high risk of expression warning is triggered, and the structural integrity check of the approval opinion and the semantic verification mechanism of key nodes are executed simultaneously; if the current approval process of the work order does not have a secondary review node, an expression risk review node is automatically inserted, and the reviewer is prompted to pay special attention to the expression abnormality pattern; for work orders with medium risk of expression, an expression standardization suggestion prompt is triggered, and the expression suggestion template is dynamically displayed in the opinion input area; for work orders with low risk of expression, no expression intervention strategy is executed, and the original process flow is maintained.

[0019] The second aspect of this invention provides a multi-terminal integrated project and maintenance information management platform, comprising: a maintenance approval behavior data acquisition and preprocessing module, a maintenance approval behavior data structured parsing module, a maintenance approver's behavioral bias assessment module, and a maintenance project risk analysis and intelligent diagnosis module, wherein: the maintenance approval behavior data acquisition and preprocessing module is used to acquire raw maintenance approval behavior data and preprocess the raw maintenance approval behavior data to obtain preprocessed raw maintenance approval behavior data; the maintenance approval behavior data structured parsing module is used to extract approval semantic elements and establish a structured parsing module based on the preprocessed maintenance approval behavior data. The system quantifies the consistency level of decision-making in approval behavior and generates expression type labels and expression anomaly template matching marks for approval behavior; the maintenance approver expression behavior bias assessment module is used to construct an expression behavior profile indicator set with approvers as the analysis unit, assess the expression bias characteristics of approvers, classify the expression behavior bias level of approvers based on the assessment results, and match maintenance process response strategies; the maintenance project risk analysis and intelligent diagnosis module is used to construct a work order risk analysis input set with work orders as the analysis unit, assess the abnormal risk value of each work order, classify the risk level of work orders based on the assessment results, and match maintenance review intervention mechanisms.

[0020] Beneficial effects

[0021] The present invention has the following beneficial effects:

[0022] (1) This multi-terminal integrated project and maintenance information management platform and method, by constructing an evaluation value for consistency of expression decision, integrates multiple dimensions such as the conciseness of approval text, approval response efficiency and consistency of opinion conclusions, and forms a unified numerical expression, which significantly enhances the ability to automatically judge the degree of standardization of approval semantic structure, provides a quantifiable basis for subsequent expression type labeling, and provides a unified scoring scale for batch analysis of approval behavior, thereby improving the accuracy and interpretability of structured semantic extraction.

[0023] (2) This multi-terminal integrated project and maintenance information management platform and method solves the problem of difficult classification of approval semantics by classifying approval behaviors into explicit, ambiguous, and deviating expression types based on the consistency evaluation value of expression decisions and generating expression tendency tags. This tagging system not only facilitates the clustering and analysis of approval behavior characteristics, but also serves as a key input for subsequent risk identification, behavior modeling, and expression template extraction, improving the flexibility and adaptability of expression behavior identification and attribution in multiple scenarios.

[0024] (3) This multi-terminal integrated project and maintenance information management platform and method, by proposing a calculation formula for the behavioral expression bias assessment value of approvers, comprehensively measures the clarity, stability, and abnormal tendencies of their approval behavior. It solves the problem of the difficulty in measuring the behavioral patterns of approvers, and can be used to identify approval roles with a high degree of expression deviation. In the actual maintenance process, it provides strong support for access control, node setting, and risk intervention, and is conducive to building a database of approver expression behavior profiles to realize the monitoring of the evolution of approval group behavior and risk warning.

[0025] (4) This multi-terminal integrated project and maintenance information management platform and method solves the problem of fixed and insensitive intervention mechanisms in traditional approval systems by matching different levels of maintenance review intervention strategies based on the approver's behavioral bias level and the work order's risk level. This linkage mechanism enables dynamic strategy adaptation, which not only improves the response speed to high-risk approval behaviors but also ensures the approval efficiency of normal processes. It is suitable for risk-driven approval process design in various maintenance task scenarios. Attached Figure Description

[0026] Figure 1 A flowchart for a multi-terminal integrated project and maintenance information management method;

[0027] Figure 2 This is a structural diagram of a multi-terminal integrated project and maintenance information management platform;

[0028] Figure 3 A hierarchy of behavioral biases expressed by maintenance approvers;

[0029] Figure 4This is a risk level distribution chart for abnormal work orders. Detailed Implementation

[0030] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0031] Please see Figures 1-4 This invention provides a technical solution: a multi-terminal integrated project and maintenance information management platform and method, comprising: S1, acquiring raw data of maintenance approval behavior, and preprocessing the raw data of maintenance approval behavior to obtain preprocessed raw data of maintenance approval behavior; S2, based on the preprocessed maintenance approval behavior data, extracting approval semantic elements and establishing structural relationships, quantifying the consistency level of expression decision of approval behavior, and generating expression type labels and expression anomaly template matching marks for approval behavior; S3, constructing an expression behavior profile indicator set with approvers as the analysis unit, evaluating the expression bias characteristics of approvers, classifying the expression behavior bias level of approvers based on the evaluation results, and matching maintenance process response strategies; S4, constructing a work order risk analysis input set with work orders as the analysis unit, evaluating the abnormal risk value of each work order, classifying the risk level of work orders based on the evaluation results, and matching maintenance review intervention mechanisms.

[0032] Specifically, maintenance approval behavior data is acquired, including work order number, approval response time, approval processing time, approval conclusion, approval opinion, and the maximum length limit of the approval opinion field. The work order number is obtained by extracting the unique identifier field from the task record; the approval response time is extracted by identifying the activation time of the approval node; the approval processing time is calculated by identifying the difference between the submission time and response time of the approval node; the approval conclusion is obtained by analyzing the decision label field carried by the approval action; the approval opinion is obtained by extracting the text input content during the approval process; and the maximum length limit of the approval opinion field is obtained by reading the maximum limit value in the approval field configuration parameters. A timestamp alignment reconstruction algorithm is used to perform time-series correction on the maintenance approval behavior data. The approval response time and submission time are reconstructed using a node time sequence comparison method. Abnormal nodes with response delays or time reversals are identified, and interpolation corrections are performed based on the node sequence position and the time interval between context nodes, thereby eliminating response record delay offsets and ensuring the accuracy and rationality of the approval behavior time sequence. Anomaly cleaning of maintenance approval data is performed using algorithms for identifying duplicate submissions and clustering based on rejection frequency. Duplicate submissions are identified using record clustering based on work order numbers, and frequency features are constructed by statistically analyzing the number of rejections. Clustering algorithms are then used to divide different frequency intervals, identifying work order records with abnormally concentrated processing times. This further removes raw maintenance approval data with irregular characteristics, eliminating non-continuous processing records and extreme behavior samples, thus improving data cleanliness and representativeness. The validity of maintenance approval data is assessed using a semantic consistency judgment algorithm between approval opinions and conclusions. Keyword extraction and semantic matching methods are used to identify core expressions in approval opinions, and consistency judgments are made based on structural tags in approval conclusions. Record types with missing, vague, or invalid opinions, or whose content does not match the decision, are identified, thus eliminating samples with unexpressive opinions and semantic conflicts, ensuring that the retained data has clear, coherent, and truthful meaning. The maintenance and repair approval behavior data is transformed into a unified dimension by a normalization compression algorithm. The approval response time and approval processing time fields are subjected to minimum-maximum standardization based on interval boundaries to normalize them to a uniform scale range. The upper limit of the length of the approval opinion field is normalized by standard deviation to make its mean zero and variance one, thereby eliminating the scale difference between different indicators, realizing the standardized expression and comparability of the original data of maintenance and repair approval behavior, and providing a consistent data foundation for subsequent expression behavior structure analysis and model evaluation.

[0033] This implementation plan constructs a rigorous data preprocessing workflow for maintenance and repair approval behavior. Without altering the original data structure, it extracts and parses standard data such as work order number, approval response time, approval processing time, approval conclusion, approval opinion, and the upper limit of the approval opinion field length. It then sequentially performs time-series correction, anomaly cleaning, expression validity determination, and scale normalization. This ensures that the maintenance and repair approval behavior data possesses characteristics of standardized structure, accurate timing, valid content, and consistent dimensions, effectively improving the accuracy, stability, and interpretability of subsequent expression behavior analysis and model evaluation.

[0034] Specifically, based on the preprocessed maintenance approval behavior data, the following steps are taken to extract approval semantic elements, establish structural relationships, and quantify the consistency level of the expression decisions in the approval behavior: Based on the preprocessed maintenance approval behavior data, the approval opinion field and approval conclusion field are extracted. A one-to-one correspondence based on the work order number is established, and the approval opinion field and approval conclusion field are matched line by line as the basic input for structured parsing. A full-text traversal operation is performed on the approval opinion field to count its total word count, extract the length of the approval opinion field, and calculate the text usage ratio based on the upper limit of the approval opinion field length. The text comparison algorithm module is called to semantically match the approval opinion field and approval conclusion field to identify whether the semantic logic is consistent. Consistent records are marked as... Recording inconsistent records as 1 and inconsistent records as 0, a corresponding opinion and conclusion consistency flag is generated; then, the length of the approval opinion field is compared with the upper limit of the approval opinion field length, and the ratio is subtracted from 1 to extract the approval text refinement item, which is used to measure the compactness of the expressed content; at the same time, the approval response time is used as the numerator, and the sum of the approval response time and the approval processing time is used as the denominator to calculate the approval response efficiency item, which is used to reflect the approver's responsiveness to the task; finally, the approval text refinement item and the approval response efficiency item are summed and accumulated with the opinion and conclusion consistency flag to calculate the expression decision consistency evaluation value, which is used to comprehensively reflect the approver's overall level in terms of semantic expression accuracy, processing response speed and decision logic consistency.

[0035] The specific formula for calculating the consistency evaluation value of the expressed decision is as follows:

[0036]

[0037] In the formula, S represents the evaluation value of decision consistency, L represents the length of the approval opinion field, and L m T represents the maximum length of the approval comments field. r Indicates the approval response time, T d The number indicates the processing time for approval, and "D" indicates a consistency marker between the opinion and the conclusion.

[0038] In this implementation plan, by extracting the approval opinion field and the approval conclusion field, a one-to-one correspondence based on the work order number is established. Combined with dimensions such as the length of the approval opinion field, the upper limit of the approval opinion field length, the approval response time, the approval processing time, and the consistency marker between the opinion and the conclusion, the conciseness of the expression text and the efficiency of the approval response are calculated. Finally, an expression decision consistency evaluation value is formed, which can accurately measure the approver's comprehensive level in terms of the conciseness of the expression content, the timeliness of the response behavior, and the consistency of semantic judgment. This effectively improves the accuracy of the approval semantic structure modeling and the quantitative ability of approval behavior data analysis, and provides a solid foundation for subsequent expression behavior bias identification and risk level assessment.

[0039] Specifically, the steps for generating expression type labels and expression anomaly template matching tags for approval behaviors are as follows: Real-time comparison of the expression decision consistency assessment value and the decision consistency threshold is performed to divide the approval behavior into intervals and generate expression type labels: If the expression decision consistency assessment value is greater than or equal to the first-level decision consistency threshold, the approval behavior is marked as having a clear expression tendency, and the expression tendency is marked as 2; if the expression decision consistency assessment value is less than the first-level decision consistency threshold but greater than the second-level decision consistency threshold, the approval behavior is marked as having an ambiguous expression tendency, and the expression tendency is marked as 1; if the expression decision consistency assessment value is less than or equal to the second-level decision consistency threshold, the approval behavior is marked as having a deviation expression tendency, and the expression tendency is marked as 0. For approval behaviors marked as having a deviation expression tendency, the corresponding approval opinion field is extracted one by one to construct a corpus set of deviation expression behaviors, ensuring that each corpus has a valid correspondence with the approval conclusion field and satisfies the continuity of timestamp logic and the integrity of expression semantics. In the corpus set, a sliding window segmentation operation based on a fixed number of characters is performed to form several sets of segment fragments, and the frequency of occurrence in the segment fragment sets is statistically analyzed using a high-frequency phrase clustering algorithm. The high-frequency phrase clustering algorithm includes: first, performing a sliding window traversal on the speech fragment set based on an n-gram model to extract continuous phrase fragments of a specified length; then, calculating the frequency of each phrase fragment in the corpus, and performing similarity clustering on phrases with frequencies higher than the statistical lower limit based on word form reconstruction and semantic similarity; next, constructing a candidate set of main expressions based on the most frequent representative phrase in each category, and finally selecting main expression fragments with consecutive occurrences exceeding the corpus threshold to generate a set of expression anomalies. Further, semantic matching is performed on various expression templates in the set of expression anomalies with the historical approval opinion field one by one. If a match is successful and the expression decision consistency evaluation value is consistently lower than the secondary decision consistency threshold in multiple consecutive approval records, then the expression template is marked as an expression anomaly template, and the expression anomaly template matching flag for the current approval behavior is assigned a value of 1; if the above conditions are not met, the expression anomaly template matching flag is assigned a value of 0.

[0040] This implementation scheme introduces a comparison mechanism between the expression decision consistency assessment value and the decision consistency threshold to achieve refined hierarchical classification of the expression level of approval behavior. This effectively identifies approval behavior types with clear expression tendencies, ambiguous expressions, and deviations. Simultaneously, by constructing a corpus of deviation-type expressions and extracting core expression fragments based on a sliding window segmentation strategy and a high-frequency phrase clustering algorithm, a set of expression anomaly phrases is further formed. Combined with semantic matching of historical approval opinion fields and the continuity judgment of expression decision consistency assessment values, expression anomaly templates are accurately identified, thereby generating expression anomaly template matching tags. This method improves the systematicness of approval behavior expression feature identification, the accuracy of expression anomaly pattern extraction, and the credibility of expression risk modeling, providing a quantitative foundation and structured support for expression intervention mechanisms and risk linkage responses in subsequent approval processes.

[0041] Specifically, the steps for constructing an expression behavior profile indicator set using approvers as the analysis unit to assess the approver's expression bias characteristics are as follows: First, aggregate all approval records corresponding to the approver to construct the expression behavior profile indicator set. Second, by calling the work order number field, link the approval opinion field and approval conclusion field one by one to extract the number of records with clear expression bias, the number of records with ambiguous expression bias, and the number of records with deviating expression bias. Based on these three types of expression type labels, calculate the expression deviation rate, expression ambiguity rate, and expression clarity rate. Third, based on the expression decision consistency evaluation value field in the approval records, calculate the arithmetic mean of the consistency markers between the opinions and approval conclusions for all records of the approver, generating the average consistency value between opinions and approval conclusions. Calculate the average value of the expression decision consistency evaluation value field to obtain the mean of the expression decision consistency evaluation value. Simultaneously, calculate the variance of the expression decision consistency evaluation value field to obtain the variance of the expression consistency evaluation value. Fourth, further count the number of records where the approver matched the expression anomaly template with a match mark of 1, summarizing the total historical number of times the expression anomaly template was matched. Finally, based on the above expression behavior profile indicator set, assess the approver's expression bias characteristics. The expression clarity rate and the average consistency rate between opinions and approval conclusions are summed. This sum is then divided by the sum of the expression deviation rate and the expression ambiguity rate plus one to obtain the expression clarity ratio. The mean of the expression decision consistency assessment value is divided by the square root of the variance of the expression consistency assessment value, and then one is added to generate the expression stability adjustment term. The reciprocal of the total number of historical hits of the expression anomaly template is added to one to obtain the expression anomaly compression factor. Finally, the expression clarity ratio, expression stability adjustment term, and expression anomaly compression factor are multiplied to generate the final behavioral expression bias assessment value, which provides a quantitative basis for subsequent expression behavior level classification and permission response strategy matching.

[0042] The specific formula for calculating the behavioral expression bias assessment score is as follows:

[0043]

[0044] In the formula, B represents the behavioral expression bias assessment value, and R... m Indicates the clarity of expression. The average value of the consistency between the opinion and the approval conclusion, R f R represents the ambiguity rate. d Indicates the rate of deviation in expression. S represents the mean of the evaluation values ​​for consistency in expressing decisions. f represents the variance of the expression consistency assessment value, and M represents the total number of historical hits on the expression aberration template.

[0045] In this embodiment, Table 1 is a data table of behavioral expression bias assessment values, which systematically displays the quantitative results of the expression behavior characteristics exhibited by five approvers during the maintenance work order processing. The table contains eight key variables, namely, the expression clarity rate, the average consistency between opinions and approval conclusions, the expression ambiguity rate, the expression deviation rate, the mean of expression decision consistency assessment, the variance of the consistency assessment value, the total number of historical hits of the expression abnormality template, and the finally calculated behavioral expression bias assessment value. Among them, approver A1's expression clarity rate was 0.65, the average consistency rate between opinions and approval conclusions was 0.62, the expression ambiguity rate was 0.20, the expression deviation rate was 0.10, the mean expression-decision consistency assessment value was 1.9038, the variance of the consistency assessment value was 0.05, the total number of times the expression abnormality template was hit was 1, and the final calculated behavioral expression bias assessment value was 0.76; approver A2's expression clarity rate was 0.58, the average consistency rate between opinions and approval conclusions was 0.50, the expression ambiguity rate was 0.22, the expression deviation rate was 0.15, the mean expression-decision consistency assessment value was 3.2724, the variance of the consistency assessment value was 0.07, the total number of times the expression abnormality template was hit was 2, and the behavioral expression bias assessment value was 0.68; approver A3's expression clarity rate was 0.78, the average consistency rate between opinions and approval conclusions was 0.72, the expression ambiguity rate was 0.10, and the expression deviation rate was 0.05. The average expression consistency assessment score for approver A4 was 1.7296, with a variance of 0.04. The total number of times the expression aberration template was hit was 1, and the behavioral expression bias assessment score was 0.94. For approver A4, the expression clarity rate was 0.70%, the average consistency between opinion and approval conclusion was 0.68%, the expression ambiguity rate was 0.15%, and the expression deviation rate was 0.10. The average expression consistency assessment score for approver A5 was 3.3341, with a variance of 0.06. The total number of times the expression aberration template was hit was 3, and the behavioral expression bias assessment score was 0.85. For approver A5, the expression clarity rate was 0.63%, the average consistency between opinion and approval conclusion was 0.60%, the expression ambiguity rate was 0.18%, and the expression deviation rate was 0.09. The average expression consistency assessment score for approver A5 was 3.0262, with a variance of 0.06. The total number of times the expression aberration template was hit was 2, and the behavioral expression bias assessment score was 0.82.

[0046] Table 1. Data Table of Behavioral Expression Bias Assessment Values

[0047]

[0048] like Figure 3The chart shown is a hierarchy of behavioral biases among maintenance approvers. It displays the behavioral bias assessment values ​​of five approvers and categorizes them into corresponding behavioral bias levels based on behavioral thresholds. The bar colors reflect different level types: green indicates clear and stable expression, orange indicates moderate deviation, and red indicates high-risk deviation. The horizontal dashed lines indicate the first-level and second-level behavioral bias thresholds. As can be seen from the chart: approvers A3 and A4 have behavioral bias assessment values ​​exceeding the first-level threshold and are marked as clear and stable; approvers A1 and A5 have behavioral bias assessment values ​​between the two thresholds and are marked as moderate deviation; approver A2 has a behavioral bias assessment value below the second-level threshold and is judged as high-risk deviation. Figure 3 It clearly demonstrates the stability of approvers' expression behavior and potential abnormal risks, and can provide a quantitative reference for the authority control and intervention mechanism in the maintenance process.

[0049] This implementation plan constructs a set of expression behavior profile indicators to systematically evaluate the expression bias characteristics of approvers in different approval records, effectively enhancing the ability to quantitatively judge the stability of approvers' expression tendencies and the risk of abnormal deviations. By simultaneously extracting the number of records with clear expression tendencies, the number of records with ambiguous expression tendencies, and the number of records with deviations in expression tendencies, an indicator system of expression deviation rate, expression ambiguity rate, and expression clarity rate is constructed. Based on the average consistency value between opinions and approval conclusions, the mean of expression decision consistency evaluation value, and the variance of expression consistency evaluation value generated from the approval opinion and approval conclusion fields, fine-grained modeling of the consistency of approval behavior expression is further realized. Combining the historical total number of times the expression anomaly template is hit, an expression anomaly compression factor is generated and the expression clarity ratio item and expression stability adjustment item are integrated. The final behavioral expression bias evaluation value can support the accurate classification and configuration of approver access control strategies, improving the level of intelligence and transparency in the maintenance process approval stage.

[0050] Specifically, the steps for classifying approvers' expression behavior bias levels based on the assessment results and matching them with maintenance process response strategies are as follows: Based on the expression bias assessment value, approvers are classified into expression behavior bias levels and maintenance process response strategies are matched accordingly: If the expression bias assessment value is greater than or equal to the Level 1 expression behavior threshold, and the approver's historical approval records show a consistently high level of clarity in expression and consistency between opinions and approval conclusions, then the approver is marked as having clear and stable expression, and is granted normal access in the maintenance process. No additional structural validation mechanisms or semantic intervention prompts are needed through the approval path, and no intervention strategies are triggered. If the expression bias assessment value is between the Level 1 and Level 2 expression behavior thresholds, and the approver's total historical hit count of the expression abnormality template is in the middle range... If the approver is marked as having moderate deviation in expression, and a decision opinion structure verification prompt is enabled for the associated maintenance work order, the prompt module provides real-time prompts on the completeness of the approval opinion field structure and the standardization of expression. If the behavioral expression bias assessment value is less than or equal to the secondary expression behavior threshold, and the approver's corresponding expression deviation rate is higher than the expression clarity rate, and the behavioral expression bias assessment value shows a continuous downward trend, then the approver is marked as having high-risk deviation in expression, the approver's independent approval authority is restricted, the approval authority control module removes the authority allocation for the approver to complete work order approval alone, and a high-risk expression reminder and multi-level review process linkage mechanism is activated for the associated maintenance work order, automatically switching to the targeted review path, where multiple responsible persons jointly complete semantic review and expression standardization review to ensure the quality of maintenance process decision-making and safety compliance level.

[0051] This implementation plan categorizes approvers based on behavioral expression bias assessment values ​​to determine their level of expression bias and matches this with maintenance process response strategies. This enables accurate identification of approvers' expression capabilities and stability, improving the rationality of authority allocation within the approval process. Specifically, the behavioral expression bias assessment values ​​categorize approvers into three types: clear and stable expression, moderately biased expression, and high-risk biased expression. This effectively avoids potential interference risks to approval results caused by inconsistencies in expression. A structure verification prompt mechanism is triggered for approvers with moderately biased expression to enhance their awareness of standardized expression. Simultaneously, approval authority is restricted for approvers with high-risk biased expression, and a multi-level review linkage mechanism is implemented to ensure complete expression, semantic consistency, and decision traceability in the work order approval process. This effectively improves the quality of expression and accuracy of review in the maintenance work order flow, ensuring the transparency and compliance of the maintenance process.

[0052] Specifically, the steps for constructing the work order risk analysis input set using work orders as the analysis unit are as follows: Using work orders as the unit, summarize all approval node information in the current work order to construct the work order risk analysis input set: By retrieving the work order number field, extract all approval node entries contained in the current work order, and perform validity verification based on the approval response time field and approval processing time field recorded in the approval node to identify approval nodes with complete expression characteristics, and extract the total number of approval nodes and the number of valid approval nodes in the current work order accordingly; Based on the expression type labels generated by the approval opinion field and the approval conclusion field, count the number of approval nodes marked as expression deviation type in the current work order, and record them as the number of expression deviation type nodes; Based on the annotation results of the expression anomaly template matching label field, count the number of approval nodes that hit the expression anomaly template, and record them as the number of anomaly template hit nodes; Sequentially obtain the opinion and conclusion consistency label and the expression tendency clear label for each valid approval node, and combine them with the approver identifier field associated with the approval node to call its corresponding real-time behavior expression bias evaluation value, constructing a structured work order risk analysis input set, providing complete expression data support for subsequent work order anomaly risk value assessment.

[0053] This implementation plan constructs a structured input set for work order risk analysis, enabling the systematic summarization and quantification of the expression characteristics of approval nodes at the work order granularity. This effectively improves the integration efficiency of multi-dimensional data such as the number of nodes with expression deviation, the number of nodes hitting abnormal templates, the total number of approval nodes, the number of valid approval nodes, the consistency markers of opinions and conclusions, the markers of clear expression tendencies, and the evaluation values ​​of behavioral expression biases. This provides a highly complete and consistent input foundation for subsequent calculation of work order abnormal risk values ​​and risk level classification, enhancing the interpretability of maintenance work order expression status identification and the accuracy of expression risk assessment.

[0054] Specifically, the steps for assessing the abnormal risk value of each work order are as follows: Based on the constructed work order expression risk analysis input set, assess the abnormal risk value of each work order: First, calculate the sum of the number of expression deviation nodes and the number of nodes hit by the abnormal template in the current work order, as the joint measurement basis for the abnormal expression signal, and divide this result by the sum of the total number of approval nodes and 1 to obtain the structural abnormality amplification term, thereby achieving a normalized assessment of the abnormal node density; then, iterate through all valid approval nodes in the current work order one by one, and calculate the inverse value of the opinion and conclusion consistency flag of each valid approval node and the table. The product of the inverse values ​​of the tendency markers is multiplied by the reciprocal of the sum of the behavioral expression bias assessment value of the approver corresponding to that node and 1, to reflect the combined influence of single-node expression bias and approver biased behavior. The product results of all valid approval nodes are summed and divided by the number of valid approval nodes to obtain the mean node expression risk term, which serves as an aggregated quantitative indicator of the degree of expression bias of the approval node group. Finally, the structural anomaly amplification term is multiplied by the mean node expression risk term to obtain the final work order anomaly risk value, so as to achieve a comprehensive assessment of work order anomaly risk from both structural characteristics and node expression status dimensions.

[0055] The specific formula for calculating the abnormal risk value of a work order is as follows:

[0056]

[0057] In the formula, G represents the abnormal risk value of the work order, and N d N represents the number of nodes that deviate from the expected value. m N represents the number of nodes that were hit by the abnormal template. g D represents the total number of approval nodes, n represents the number of valid approval nodes, and D represents the number of approval nodes. j P represents the consistency marker between the opinions and conclusions at the j-th approval node. j B represents the expression tendency marker of the j-th approval node. j This represents the biased evaluation value of the approver's behavior corresponding to the j-th approval node.

[0058] In this embodiment, Table 2 is a work order anomaly risk value data table, which shows the quantitative results of the expression anomaly characteristics of 5 maintenance work orders in the approval process. Based on the results of approval node behavior analysis, this data table lists the number of expression deviation nodes, the number of nodes hit by the anomaly template, the total number of approval nodes, the number of valid approval nodes, and the finally calculated work order anomaly risk value for each work order, which is used to comprehensively assess the degree of potential expression deviation and risk hazards of the work orders in the approval process. Among them, work order A has 3 nodes with expression deviation, 2 nodes with abnormal template matching, a total of 10 approval nodes, and 9 valid approval nodes, resulting in a work order abnormality risk value of 0.08; work order B has 4 nodes with expression deviation, 3 nodes with abnormal template matching, a total of 12 approval nodes, and 11 valid approval nodes, resulting in a work order abnormality risk value of 0.17; work order C has 2 nodes with expression deviation, 1 node with abnormal template matching, a total of 8 approval nodes, and 7 valid approval nodes, resulting in a work order abnormality risk value of 0.08; work order D has 2 nodes with expression deviation, 1 node with abnormal template matching, a total of 11 approval nodes, and 10 valid approval nodes, resulting in a work order abnormality risk value of 0.13; and work order E has 3 nodes with expression deviation, 2 nodes with abnormal template matching, a total of 10 approval nodes, and 9 valid approval nodes, resulting in a work order abnormality risk value of 0.10.

[0059] Table 2. Work Order Abnormal Risk Value Data Table

[0060] Work order <![CDATA[N d ]]> <![CDATA[N m ]]> <![CDATA[N g ]]> n G Work Order A 3 2 10 9 0.08 Work Order B 4 3 12 11 0.17 Work order C 2 1 8 7 0.08 Work Order D 2 1 11 5 0.13 Work Order E 4 2 10 9 0.10

[0061] like Figure 4 The image shows a risk level distribution chart for work order anomalies. This chart displays the anomaly risk values ​​of five maintenance work orders and categorizes them according to risk level thresholds, labeling the risk level type of each work order. The colors of the dots in the chart reflect different risk levels: red indicates high-risk work orders, orange indicates medium-risk work orders, and green indicates low-risk work orders. The horizontal dashed lines in the chart indicate the first-level risk threshold and the second-level risk threshold, respectively. As can be seen from the chart: Work order B's anomaly risk value is higher than the first-level risk threshold, and it is classified as a high-risk work order; Work orders D and E's anomaly risk values ​​are between the two thresholds, and they are marked as medium-risk work orders; Work orders A and C's anomaly risk values ​​are lower than the second-level risk threshold, and they are classified as low-risk work orders. Figure 4 It clearly demonstrates the risk profile of maintenance work orders in terms of expression structure and node consistency, providing data support for review strategies and expression anomaly identification mechanisms during the approval process.

[0062] This implementation plan constructs a structural anomaly amplification item and a node expression risk mean item, and calculates the work order anomaly risk value accordingly, achieving a fusion assessment of the expression status of approval nodes and the behavioral bias characteristics of approvers in maintenance work orders. This method effectively improves the ability to identify anomaly expression patterns at the work order level, enhances the comprehensive utilization efficiency of the number of expression deviation nodes, the number of nodes hitting anomaly templates, the total number of approval nodes, the consistency markers of opinions and conclusions, the expression tendency markers, and the behavioral expression bias assessment value, ensuring the completeness and rigor of the work order anomaly risk value quantification process, and improving the accuracy of expression behavior anomaly identification and the intelligent response level of the maintenance approval process.

[0063] Specifically, the steps for classifying work orders by risk level based on the assessment results and matching them with the maintenance audit intervention mechanism are as follows: Based on the abnormal risk value of the work order, the work order risk level is classified: by setting a primary risk threshold and a secondary risk threshold, work orders are classified into high-risk, medium-risk, and low-risk levels, respectively. If the abnormal risk value of the work order is greater than or equal to the primary risk threshold, the work order is marked as a high-risk work order, and the risk level mark is written into the corresponding work unit data; if the abnormal risk value of the work order is less than the primary risk threshold but greater than the secondary risk threshold, the work order is marked as a medium-risk work order, and the medium-risk level mark is embedded in the work order's associated metadata; if the abnormal risk value of the work order is less than or equal to the secondary risk threshold, the work order is marked as a low-risk work order, and a low-risk level identifier field is written into the work order data structure. Based on the work order risk level results, an automatic risk warning strategy and maintenance review linkage mechanism are matched: For work orders with high expression risk, a high-risk expression warning is automatically triggered, and a mechanism for checking the structural integrity of approval opinions and a mechanism for verifying the semantics of key nodes are executed simultaneously to identify cases of vague approval semantics and missing opinions; if the current approval process for this work order does not have a secondary review node, an expression risk review node is automatically inserted, and the node attribute is updated in the approval node configuration table, while prompting reviewers to pay close attention to the identified abnormal expression patterns; For work orders with medium expression risk, a suggestion for expression standardization is triggered, and expression suggestion templates related to the semantic structure of the current work order are dynamically displayed in the opinion input area to assist approvers in completing a full semantic expression; For work orders with low expression risk, no expression intervention strategy is executed, and the original process continues to flow stably according to the preset approval path to ensure that normal work orders are not affected by redundant interventions.

[0064] This implementation plan establishes a work order risk level classification mechanism based on work order anomaly risk values, combined with a linkage mechanism between work unit data writing and maintenance review intervention, to achieve dynamic identification and tiered response to anomaly risks in the maintenance process. While ensuring the semantic integrity of approvals, it effectively improves the accuracy of work order risk management. By automatically triggering structural integrity checks and key node semantic verification mechanisms for high-risk work orders, and dynamically inserting expression risk review nodes in processes lacking secondary review nodes, it achieves flexible intervention in the work order review process. Through expression standardization suggestions and template displays for medium-risk work orders, it improves the accuracy and consistency of approvers' opinions. The non-intervention strategy for low-risk work orders ensures optimal scheduling of process resources. The maintenance review intervention linkage mechanism, driven by work order anomaly risk values, enhances the closed-loop control capability of the maintenance process against expression anomalies, improves the level of expression structure governance, and enhances the adaptability of the review path.

[0065] like Figure 2 As shown, the second aspect of this invention provides a multi-terminal integrated project and maintenance information management platform, including: a maintenance approval behavior data acquisition and preprocessing module, a maintenance approval behavior data structured parsing module, a maintenance approver expression behavior bias assessment module, and a maintenance project risk analysis and intelligent diagnosis module. The maintenance approval behavior data acquisition and preprocessing module is used to acquire raw maintenance approval behavior data and preprocess it to obtain preprocessed raw maintenance approval behavior data. The maintenance approval behavior data structured parsing module is used to extract approval semantic elements and establish a structured parsing module based on the preprocessed maintenance approval behavior data. The system includes a structural relationship module to quantify the consistency level of decision-making in approval behavior and generate expression type labels and expression anomaly template matching marks for approval behavior; a maintenance approver expression behavior bias assessment module to construct an expression behavior profile indicator set with approvers as the analysis unit, assess the expression bias characteristics of approvers, classify the expression behavior bias level of approvers based on the assessment results, and match maintenance process response strategies; and a maintenance project risk analysis and intelligent diagnosis module to construct a work order risk analysis input set with work orders as the analysis unit, assess the abnormal risk value of each work order, classify the risk level of work orders based on the assessment results, and match maintenance review intervention mechanisms.

[0066] This implementation plan constructs a multi-terminal integrated information management platform, including a maintenance approval behavior data collection and preprocessing module, a maintenance approval behavior data structure parsing module, a maintenance approver expression behavior bias assessment module, and a maintenance project risk analysis and intelligent diagnosis module. This platform achieves standardized processing, semantic structure parsing, behavioral feature assessment, and intelligent risk diagnosis of approval behavior data throughout the maintenance process, creating a closed-loop control system. The platform constructs an expression behavior profile indicator set using approvers as the analysis unit and a work order risk analysis input set using work orders as the analysis unit, ensuring improved accuracy in both data collection and risk identification. By quantifying the consistency level of approval behavior in decision-making and generating expression type labels and expression anomaly template matching markers, the platform enhances the structuring of approval semantic expression. By classifying and matching the approver's expression bias characteristics with maintenance process response strategies, it strengthens the dynamic adaptability of access control. By assessing abnormal risk values ​​and classifying risk levels in work orders, it enhances the risk identification and intervention response capabilities of the maintenance process, thereby improving the overall management transparency and operational security of the maintenance project execution process.

[0067] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0068] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A multi-terminal integrated project and maintenance information management method, characterized by, The method comprises the following steps: S1, obtaining maintenance approval behavior original data, and preprocessing the maintenance approval behavior original data to obtain preprocessed maintenance approval behavior original data; S2, based on the preprocessed maintenance approval behavior data, extracting approval semantic elements and establishing structural relationships, quantifying the expression decision consistency level of the approval behavior, and generating expression type labels and expression abnormal template matching marks of the approval behavior; The specific steps of extracting approval semantic elements and establishing structural relationships based on the preprocessed maintenance approval behavior data and quantifying the expression decision consistency level of the approval behavior are as follows: Based on the preprocessed maintenance approval behavior data, the approval opinion and approval conclusion fields are extracted, and a one-to-one correspondence relationship based on the work order number is established as the basis for structured analysis input. The length of the approval opinion field is counted. The consistency of the approval opinion and the approval is judged by text comparison. If they are consistent, 1 is recorded. If they are inconsistent, 0 is recorded. The consistency of the opinion and the conclusion is generated. The ratio of the length of the approval opinion field to the upper limit of the length of the approval opinion field is calculated, and the ratio is subtracted by 1 to obtain the approval text refinement degree item. The approval response time is divided by the sum of the approval response time and the approval processing time to obtain the approval response efficiency item. The approval text refinement degree item and the approval response efficiency item are added, and then the approval opinion and approval conclusion consistency mark is added to obtain the expression decision consistency evaluation value. S3, constructing an expression behavior portrait index set with the approver as the analysis unit, evaluating the expression bias characteristics of the approver, dividing the expression behavior bias level of the approver based on the evaluation result, and matching the maintenance process response strategy; The specific steps of constructing an expression behavior portrait index set with the approver as the analysis unit and evaluating the expression bias characteristics of the approver are as follows: Based on the approver, all the approval records corresponding to the approver are summarized to construct an expression behavior portrait index set: the number of records of expression tendency explicit type, expression fuzzy type and expression deviation type is extracted, and the expression deviation rate, expression fuzzy rate and expression explicit rate are calculated. Based on the expression decision consistency evaluation value of the approval record, the average value of the opinion and the approval conclusion consistency, the average value of the expression decision consistency evaluation value and the expression consistency evaluation value variance are calculated. The number of approval records that hit the expression abnormal template is counted to obtain the total number of times that hit the expression abnormal template; Based on the expression behavior portrait index set, the expression bias characteristics of the approver are evaluated: the sum of the expression explicit rate and the average value of the opinion and the conclusion consistency is calculated, and the sum is divided by the sum of the expression deviation rate and the expression fuzzy rate plus one to obtain the expression clarity proportion item; the average value of the expression decision consistency evaluation value is divided by the square root of the expression consistency evaluation value variance plus one to obtain the expression stability adjustment item; The reciprocal of the sum of one and the total number of times that hit the expression abnormal template is calculated to obtain the expression abnormal compression factor. The expression clarity proportion item, the expression stability adjustment item and the expression abnormal compression factor are multiplied to obtain the behavior expression bias evaluation value. S4, taking the work order as the analysis unit to construct a work order risk analysis input set, evaluating the abnormal risk value of each work order, classifying the work order according to the risk level based on the evaluation result, and matching the maintenance audit intervention mechanism.

2. The multi-terminal integrated item and maintenance information management method according to claim 1, characterized by: The specific steps of obtaining the maintenance approval behavior original data and preprocessing the maintenance approval behavior original data are as follows: Obtain maintenance approval behavior data, including work order number, approval response time, approval processing time, approval conclusion, approval opinion and upper limit of approval opinion field length; Through the time stamp alignment reconstruction algorithm, the maintenance approval behavior data is time sequence corrected to eliminate the response record delay offset; through the repeated submission identification and rejection frequency clustering algorithm, the maintenance approval behavior data is abnormally cleaned to remove irregular flow records and extreme behavior samples; through the approval opinion and approval conclusion semantic consistency determination algorithm, the effectiveness of the maintenance approval behavior data is judged to identify samples with no expressive opinion and semantic conflict; through the normalization compression algorithm, the maintenance approval behavior data is uniformly dimensionally converted to complete the standardization and scale unification processing of the maintenance approval behavior data.

3. The multi-terminal integrated item and maintenance information management method of claim 1, wherein: The specific steps of generating the expression type label and expression abnormal template matching mark of the approval behavior are as follows: Real-time comparison of expression decision consistency evaluation value and decision consistency threshold value, interval division of approval behavior, generation of expression type label: if the expression decision consistency evaluation value is greater than or equal to the first-level decision consistency threshold value, the approval behavior is marked as expression tendency clear type, and the expression tendency mark is 2; if the expression decision consistency evaluation value is less than the first-level decision consistency threshold value and greater than the second-level decision consistency threshold value, the approval behavior is marked as expression fuzzy type, and the expression tendency mark is 1; if the expression decision consistency evaluation value is less than or equal to the second-level decision consistency threshold value, the approval behavior is marked as expression deviation type, and the expression tendency mark is 0; For the approval behavior marked as expression deviation type, the corresponding approval opinion is extracted to construct an expression deviation type corpus set; fixed length sliding window segmentation and high frequency phrase clustering operation are performed in the corpus set, the main expression fragments with continuous appearance frequency exceeding the corpus threshold value are extracted, and the expression abnormal phrase set is generated; the expression template in the expression abnormal phrase set is matched with the historical approval opinion, if the matching is successful and the expression decision consistency evaluation value is continuously lower than the second-level decision consistency threshold value, the template is marked as expression abnormal template, the expression abnormal template matching mark is 1, otherwise it is 0.

4. The multi-terminal integrated item and maintenance information management method of claim 1, wherein: The specific steps of classifying the expression behavior bias level of the approver based on the evaluation result and matching the maintenance process response strategy are as follows: According to the behavior expression bias evaluation value, the expression behavior bias of the approver is classified, and a maintenance process response strategy is matched: if the behavior expression bias evaluation value is greater than or equal to a first-level expression behavior threshold value, the approver is marked as an expression clear and stable type, normal permission transfer is given in the maintenance process, and no intervention strategy is triggered; if the behavior expression bias evaluation value is between the first-level expression behavior threshold value and a second-level expression behavior threshold value, the approver is marked as an expression moderate deviation type, and a decision opinion structure verification prompt is enabled for the associated maintenance work order; if the behavior expression bias evaluation value is less than or equal to the second-level expression behavior threshold value, the approver is marked as an expression high-risk deviation type, the independent approval authority of the approver is limited, and an expression high-risk prompt and a multi-level review process linkage mechanism are started for the associated maintenance work order, and the work order is automatically transferred into a directional review path.

5. The multi-terminal integrated item and maintenance information management method of claim 1, wherein: The specific steps of constructing the work order risk analysis input set with the work order as the analysis unit are as follows: In the unit of work order, all the approval node information in the current work order is summarized to construct the work order risk analysis input set: the total number of approval nodes and the number of effective approval nodes of the current work order are extracted; the number of approval nodes of the expression deviation type in the current work order is counted, which is recorded as the number of expression deviation nodes; the number of approval nodes that hit the expression abnormal template is counted, which is recorded as the number of abnormal template hit nodes; the opinion and conclusion consistency mark, the expression tendency clear mark of each effective approval node, and the real-time behavior expression bias evaluation value of the corresponding approver are obtained.

6. The multi-terminal integrated item and maintenance information management method of claim 1, wherein: The specific steps of evaluating the abnormal risk value of each work order are as follows: Based on the work order expression risk analysis input set, the abnormal risk value of each work order is evaluated: the sum of the number of expression deviation nodes and the number of abnormal template hit nodes is calculated, and then divided by the total number of approval nodes plus one to obtain a structural abnormal amplification item; the product of the inverse value of the opinion and conclusion consistency mark and the inverse value of the expression tendency mark of each effective approval node in the current work order is calculated, and then multiplied by the inverse of the behavior expression bias evaluation value of the corresponding approver plus one, and then summed and divided by the total number of nodes to obtain a node expression risk average item; the structural abnormal amplification item and the node expression risk average item are multiplied to obtain the work order abnormal risk value.

7. The multi-terminal integrated item and maintenance information management method of claim 1, wherein: The specific steps of classifying the risk level of the work order based on the evaluation result and matching the maintenance review intervention mechanism are as follows: According to the work order abnormal risk value, the work order risk level is classified: if the work order abnormal risk value is greater than or equal to a first-level risk threshold value, the work order is marked as an expression high-risk work order; if the work order abnormal risk value is between the first-level risk threshold value and a second-level risk threshold value, the work order is marked as an expression medium-risk work order; if the work order abnormal risk value is less than or equal to the second-level risk threshold value, the work order is marked as an expression low-risk work order; the risk level mark is written into the work order metadata; According to the work order risk level result, the risk prompt strategy is automatically matched with the maintenance audit linkage mechanism: for the high-risk work order, the high-risk prompt of the work order expression is triggered, and the approval opinion structure integrity check and key node semantic verification mechanism are synchronously executed; if the current approval process of the work order is not set with a secondary review node, the expression risk review node is automatically inserted, and the review personnel is prompted to pay attention to the abnormal expression mode; for the medium-risk work order, the expression standardization suggestion prompt is triggered, and the expression suggestion template is dynamically displayed in the opinion input area; For the low-risk work order, no expression intervention strategy is executed, and the original process is maintained.

8. A multi-terminal integrated project and maintenance information management platform, applying the multi-terminal integrated project and maintenance information management method according to any one of claims 1-7. It comprises: a maintenance approval behavior data collection preprocessing module, a maintenance approval behavior data structured analysis module, a maintenance approver expression behavior bias evaluation module and a maintenance project risk analysis and intelligent diagnosis module, wherein: The maintenance approval behavior data collection preprocessing module is used for acquiring maintenance approval behavior raw data, preprocessing the maintenance approval behavior raw data, and acquiring preprocessed maintenance approval behavior raw data; The maintenance approval behavior data structured analysis module is used for extracting approval semantic elements and establishing structural relationships based on the preprocessed maintenance approval behavior data, quantifying the expression decision consistency level of the approval behavior, and generating expression type tags and expression abnormal template matching marks of the approval behavior; The maintenance approver expression behavior bias evaluation module is used for constructing an expression behavior portrait index set with the approver as the analysis unit, evaluating the expression bias characteristics of the approver, dividing the expression bias level of the approver based on the evaluation result, and matching the maintenance process response strategy; The maintenance project risk analysis and intelligent diagnosis module is used for constructing a work order risk analysis input set with the work order as the analysis unit, evaluating the abnormal risk value of each work order, dividing the risk level of the work order based on the evaluation result, and matching the maintenance audit intervention mechanism.

Images