Engineering comprehensive risk multi-source credibility fusion rule reasoning and evidence chain generation method

Abstract

The application discloses an engineering comprehensive risk multi-source credibility fusion rule reasoning and evidence chain generation method, and belongs to the technical field of intelligent construction and engineering management informatization. The application integrates multi-source business data to construct a unified fact wide table; confidence is calculated for field data, a preset conflict threshold is respectively used to generate effective fact values by adopting a confidence weighted average or a high confidence priority strategy; derived indexes are calculated and construction stages are identified; a rule engine calculates three-dimensional baseline risks, iterates rules and processes multi-rule conflicts through a ruling mechanism to form two-level processing results; comprehensive risk scores and risk grades are obtained by weighted aggregation and disposal suggestions are output; whole-process information is integrated to generate a structured evidence chain; when data or rules are changed, local incremental recalculation is executed and the evidence chain is updated. The application solves the problems of multi-source data conflict, rule conflict, untraceable results and low efficiency of full recalculation, and is suitable for engineering project risk early warning, audit review and auxiliary decision-making.

Description

Technical Field

[0001] This invention belongs to the field of intelligent construction and engineering management information technology, specifically relating to a method for multi-source credibility fusion rule reasoning and evidence chain generation for comprehensive engineering risks. Background Technology

[0002] With the development of intelligent construction and refined digital management of engineering projects, the demand for dynamic control of comprehensive risks related to progress, cost, and safety in project construction continues to increase. The overlapping of work processes and dynamic resource allocation at construction sites, coupled with external disturbances such as extreme weather and equipment changes, result in risks exhibiting phased, delayed, and coupled characteristics: schedule deviations induce rush work and overtime, simultaneously amplifying cost overruns and safety hazards; the switching between stages such as work stoppages, resumptions, capacity ramp-ups, and rush work further exacerbates the complexity of risk evolution, necessitating real-time risk assessment based on full-process data and the output of practical control recommendations.

[0003] Engineering projects generate massive amounts of heterogeneous data from multiple sources throughout their entire lifecycle, including progress, cost, inspection, and environmental incident data. This data originates from various channels such as business systems, supervision, and manual reporting by contractors, and commonly suffers from issues such as inconsistent definitions, misaligned timelines, missing or noisy data, and conflicting values ​​for the same indicators. Existing engineering risk assessment methods often rely on fixed threshold warnings, experience-weighted scoring, or single-system data analysis, which are difficult to adapt to scenarios with conflicting multi-source data. Directly using a single data source or simple mean calculations can easily lead to distorted risk assessments. Furthermore, the static and fixed rule system cannot dynamically match reasoning logic to differentiated construction scenarios such as work stoppages, resumptions, and expedited work, resulting in poor scenario adaptability.

[0004] The existing rule-based reasoning mechanism only focuses on outputting risk scores, lacking adjudication and governance logic when multiple rules are concurrently applied. This easily leads to overlapping risks and fluctuating conclusions. Furthermore, it fails to structurally record data sources, fusion processes, rule applications, and threshold comparison procedures, resulting in a lack of complete traceability for risk outcomes and hindering support for engineering audits, liability determination, and closed-loop review. In addition, the existing system uses a full data recalculation model to handle data updates, historical corrections, and rule iterations, resulting in high computational overhead and inference latency, failing to meet the real-time deployment requirements for continuous online monitoring at construction sites.

[0005] In summary, existing technologies are insufficient to achieve reliable fusion of multi-source heterogeneous data, phased adaptive rule reasoning, full-process auditability and traceability, and efficient incremental computation. They also cannot balance the stability, interpretability, and engineering implementation efficiency of risk assessment. The industry urgently needs a complete technical solution that is suitable for comprehensive risk management. Summary of the Invention

[0006] To address the shortcomings of existing technologies, the present invention aims to propose a method for multi-source credibility fusion rule-based reasoning and evidence chain generation for comprehensive engineering risks. This method employs field-level credibility fusion, phased rule arrangement, rule adjudication, structured evidence chains, and incremental recalculation mechanisms to generate multi-dimensional risk indicators and disposal recommendations. This approach improves the stability and auditability of risk outcomes under multi-source conflicts while reducing computational overhead, making it suitable for dynamic management scenarios in engineering projects.

[0007] The technical solution adopted in this invention: A method for multi-source credibility fusion rule-based reasoning and evidence chain generation for comprehensive engineering risks includes the following steps: S1. Collect business data to form a business source table. After completing the preprocessing rules, integrate the field data in each table with project identifier + date as the composite primary key to build a unified fact wide table. S2. Based on the unified fact wide table, construct a confidence calculation model and calculate the confidence parameter for each field of data; S3. Calculate the absolute difference between two fields of data in the unified fact wide table, and set a conflict threshold for each field. When the absolute difference is less than or equal to the conflict threshold, use a confidence-weighted average to generate a valid fact value. When the absolute difference is greater than the conflict threshold, execute the high-confidence priority selection strategy as the valid fact value. S4. Calculate derived indicators based on the data fields and valid fact values ​​of the unified fact wide table. These derived indicators include at least the inspection pass rate, schedule delay days, cost deviation rate, work stoppage flag, and capacity coefficient. The inspection pass rate is determined by the ratio of the number of qualified items inspected on a given day to the total number of items inspected on that day, or by using valid fact values ​​obtained by fusing inspection pass rates from multiple sources. The schedule delay days are determined by multiplying the positive difference between the cumulative planned progress and the cumulative actual progress by a preset duration factor. The cost deviation rate is determined by the ratio of the difference between the actual cumulative cost and the planned cumulative cost to the planned cumulative cost. The work stoppage flag is determined based on one of the following: work stoppage notice status, key process implementation status, and actual attendance. The capacity coefficient is calculated by weighting the attendance ratio of actual attendance to the normal attendance baseline, equipment availability rate, and key process implementation status. Based on these derived indicators, the current construction phase type of the project is comprehensively determined. The phase types include six categories: early warning preparation, complete work stoppage, resumption inspection, capacity recovery, expedited recovery, and normal construction. S5. Based on the unified fact wide table field data and construction phase, build a rule engine to calculate the three-dimensional baseline risks of progress, cost, and safety. Enter multiple rules into the rule engine and traverse them. When multiple rules conflict, the adjudication mechanism is activated to take into account multiple factors to adopt, retain or suppress the hit rules and record the corresponding information to form a two-level processing result of rule hit and rule adjudication. S6. The three-dimensional risks are linearly weighted using preset weighting coefficients to obtain a comprehensive risk score. Based on preset classification thresholds, three risk levels (low, medium, and high) are mapped, and corresponding on-site control and disposal suggestions are output according to the matching rules. S7. Integrate input fact snapshots, original values ​​and confidence levels from various sources, full records of conflict fusion, calculation process of derived indicators, stage identification labels, hit rules and adjudication records, baseline risk, rule risk increments, multidimensional risk results, comprehensive scores, risk levels, and disposal recommendations to construct an auditable and replayable structured chain of evidence. S8. When new business facts are detected, historical records are corrected, or rule versions are updated, the changed field data is identified, and only the affected part is recalculated locally and the evidence chain version is updated synchronously, without the need for a full data scan and recalculation.

[0008] Preferably, in step S1, the specific details are as follows: The data in each business source table includes the following fields: project identifier, business date, cumulative planned progress / actual progress, planned / actual cumulative cost, total number of inspections / number of qualified items on the day, cumulative actual progress / planned progress, environmental disaster type, warning level, work stoppage notice status, key process implementation status, actual number of employees present, equipment availability rate, key process implementation status, work stoppage identifier, environmental severity, resumption inspection status, work stoppage identifier, overtime hours, and update time metadata. A unified preprocessing rule is applied to the field data in each table. This preprocessing rule includes field mapping, date standardization, and deduplication within the source. Field mapping involves mapping custom business fields from each source to internal standard fields using configuration files. Date standardization involves converting all dates to the YYYY-MM-DD standard format. Deduplication within the source involves retaining the latest entry from multiple records of the same project and date. After preprocessing, the tables are integrated using the project identifier + date as the composite primary key to construct a unified wide fact table. Simultaneously, external environmental event data is integrated.

[0009] Preferably, in step S2, the specific details are as follows: Based on a unified fact wide table, a confidence score calculation model is constructed to verify whether each field in the table carries a confidence score parameter. When the field data in each table has a built-in confidence score parameter, it is directly reused; when the field data in each table does not have a confidence score parameter, the confidence score calculation model automatically calculates the confidence score parameter in the range of 0 to 1. The expression for the confidence calculation model is:

[0010] In the formula, To ensure the source is authoritative and rated, For timeliness scoring, Score for completeness.

[0011] Preferably, in step S3, the specific details are as follows: 1) Calculate the absolute difference between two field data. Its expression is:

[0012] 2) Conflict threshold determination: Three conflict thresholds are set for each field of data, namely the actual progress conflict threshold, the cumulative cost conflict threshold, and the inspection pass rate conflict threshold; when If the absolute difference between two data fields is less than or equal to the conflict threshold, it is determined that the difference is within a tolerable range. A confidence-weighted average is then used to generate valid fact values, the expression of which is: Valid fact value = (A value × A confidence level + B value × B confidence level) / (A confidence level + B confidence level); when > Conflict threshold: If the absolute difference between two fields is found to be significantly conflicting, a high confidence priority selection strategy is adopted, that is, the field with the higher confidence parameter value is selected as the valid fact value. Write all valid fact values ​​into the unified fact wide table to complete the expansion of the wide table.

[0013] Preferably, in step S4, the specific details are as follows: For the inspection pass rate, if the inspection pass rate field already exists in the unified fact wide table, it can be reused directly; if the inspection pass rate field does not exist, it can be obtained by calculating the ratio between the number of qualified items inspected on the day and the total number of items inspected on the day after merging in the table, or the result of the effective fact value after confidence fusion can be used as the inspection pass rate field. For the number of days of schedule delay, read the cumulative planned progress and cumulative actual progress in the unified fact wide table and compare the two values; when the cumulative planned progress is higher than the cumulative actual progress, it means that the current actual completion level is behind the planned completion level and the construction progress is lagging behind. Multiply the difference between the two by the schedule scale factor to obtain the number of days of schedule delay. The cost deviation rate is calculated using the expression "Cost Deviation Rate = (Actual Cumulative Cost - Planned Cumulative Cost) / Planned Cumulative Cost", which indicates the degree of deviation between the actual cumulative cost and the planned cumulative cost. The schedule performance index and cost performance index are used as auxiliary verification indicators, and their expressions are "Schedule performance index = Cumulative actual progress / Cumulative planned progress" and "Cost performance index = (Cumulative actual progress × Planned cumulative cost) / Actual cumulative cost", respectively. The work stoppage flag indicates whether the project segment is in a completely stopped state within the current time window. It is a binary variable, where 1 represents work stoppage and 0 represents no work stoppage. When the work stoppage flag is provided in the unified fact wide table, its value is used. If the wide table does not provide it directly, the determination is based on a comprehensive inference of the work stoppage notification status, the key process implementation status, and the actual number of employees present in the table. This includes: if any of the following conditions are met, the work stoppage notification status = 1, the key process implementation status = 0, or the actual number of employees present is lower than the preset work stoppage attendance threshold, the work stoppage flag is assigned the value 1, representing a complete work stoppage; otherwise, the value is assigned the value 0. The capacity coefficient represents the recovery level of current construction capacity relative to normal construction capacity. It is calculated using a weighted model, which includes: first, calculating the attendance ratio based on the actual number of employees and the normal attendance baseline; then, linearly combining the attendance ratio, equipment availability rate, and key process status according to preset weights, resulting in the capacity coefficient as expressed by the formula: "Capacity Coefficient = 0.45 × Attendance Ratio + 0.45 × Equipment Availability Rate + 0.10 × Key Process Status". When the actual number of employees, the normal attendance baseline, the equipment availability rate, or the key process status are missing, a degenerate estimation method is used, taking "1 - Stoppage Flag" as a substitute value for the capacity coefficient. The default value for the capacity coefficient is 0 in the stoppage state and 1 in the non-stoppage state. For each stage status, the system reads the shutdown flag, environmental severity, resumption inspection status, capacity coefficient, schedule delay days, and overtime hours derived from the unified fact wide table, and performs multi-condition status matching and identification, including: if the shutdown flag = 1, it is identified as a complete shutdown stage; if the environmental severity reaches the warning threshold and the capacity continues to decline, it is identified as a warning preparation stage; if it is in the resumption inspection stage and the inspection pass rate is lower than the set threshold, it is identified as a resumption inspection stage; if the capacity coefficient continues to rise but has not yet reached the normal level, it is identified as a capacity recovery stage; if the schedule delay days are large, accompanied by an increase in overtime hours and an increase in the capacity coefficient, it is identified as a rush work recovery stage; when the identification conditions of the complete shutdown stage, warning preparation stage, resumption inspection stage, capacity recovery stage, and rush work recovery stage are not met, it is marked as a normal construction stage. Through the above steps, the current construction stage of the project is comprehensively determined by combining derived indicators, including six construction stage statuses: early warning preparation, complete shutdown, resumption inspection, production capacity recovery, expedited recovery, and normal construction. All derived indicators are written into a unified fact wide table to complete the expansion of the fact layer of the wide table and form a complete fact dataset containing derived indicators.

[0014] Preferably, in step S5, the specific details are as follows: Based on derived indicators, the baseline risks for schedule, cost, and safety are calculated, with the following expressions:

[0015]

[0016]

[0017] In the formula, This is a parameter for measuring schedule delays. For cost and risk sensitivity coefficient, For safety pressure scale parameters; Multiple rules are entered into and iterated through in the rule engine. When multiple rules simultaneously trigger the same risk, an adjudication mechanism is activated to resolve the conflict between the multiple triggering rules. The adjudication mechanism comprehensively considers four evaluation factors: rule priority, specificity of conditions, temporal proximity, and source credibility. It then performs adoption, retention, or suppression on the triggering rules. For rules determined to be adopted, their triggering conditions, scope of action, risk increment, and suggested actions are recorded. For rules determined to be suppressed, the reasons for suppression, the corresponding adjudication basis, and the actions that did not take effect are recorded.

[0018] Preferably, in step S6, the specific details are as follows: The comprehensive risk score is obtained by linearly weighting the three-dimensional risks using preset weighting coefficients. The calculation formula is as follows: .

[0019] Compared with existing technologies, this invention proposes a method for multi-source credibility fusion rule-based reasoning and evidence chain generation for comprehensive engineering risks. The advantages of this method are: 1. This invention can perform difference detection and confidence fusion on similar business indicators from multiple sources such as progress systems, cost systems, inspection records, supervision records, and manual entries. When there are inconsistencies in the same fact across multiple sources, a unified fact input is formed after fusion, thereby reducing the risk judgment bias caused by differences in sources and improving the stability and credibility of the input facts.

[0020] 2. This invention can match and activate corresponding rules according to the stage of the project, so that different judgment logics are applied to the early warning preparation period, the shutdown period, the resumption inspection period, the capacity recovery period, and the rush period. This avoids the problem of insufficient adaptability or false triggering of fixed static rules in complex construction scenarios, thereby improving the matching degree between the comprehensive risk identification results and the actual operation status of the project.

[0021] 3. This invention uses a composite reasoning approach of "baseline risk plus rule increment" to continuously characterize progress, cost, and safety risks during the daily operation of a project, and to reflect risk transitions in a timely manner when there are changes in status such as typhoon warnings, work stoppages, resumptions, and rush work. Compared with simple fixed thresholds or single scoring methods, it can more realistically reflect the risk evolution process.

[0022] 4. This invention generates a structured chain of evidence simultaneously with the output of comprehensive risk results. It can completely record multi-source inputs, confidence fusion strategies, stage identification results, hit rules, risk increments, recommended measures, and final results. This makes the risk conclusions not only visible but also interpretable, traceable, and auditable, providing a basis for management review, responsibility analysis, and rule optimization.

[0023] 5. Based on the relationship between fields and rules, this invention can perform local evaluation of the subset of fields and rules affected by new facts, historical data corrections, or rule adjustments, thereby reducing unnecessary repetitive calculations, lowering system operating costs, and improving the system's response efficiency and deployment feasibility in continuous on-site monitoring, event-driven early warning, and online operating environments. Attached Figure Description

[0024] Figure 1 This is a schematic diagram of the overall process of the present invention; Figure 2 This is a flowchart of the conflict threshold determination process of the present invention. Detailed Implementation

[0025] The technical solutions of the embodiments of this application will be further described clearly and completely below with reference to the accompanying drawings. It should be noted that the described embodiments are only some embodiments of this application, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0026] To make the inventive objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be further described in detail below with reference to the accompanying drawings: In order to better understand the above-mentioned objectives, features, and advantages of this invention, the advantages of this invention will be further illustrated below by comparing the embodiments with the accompanying drawings and specific implementation methods.

[0027] This invention proposes a method for multi-source credibility fusion rule-based reasoning and evidence chain generation for comprehensive engineering risks, such as... Figure 1 The process of this method is described in detail below: This embodiment selects a building foundation pit project. The project's multi-source business data comes from six types of data sources, namely the progress management system, the on-site manual daily report filling module, the cost management system, the contractor cost reporting interface, the on-site safety inspection team record form, and the supervision and inspection form of the supervisor. The external environment is connected to the typhoon warning data of the meteorological platform.

[0028] S1. Collect business data from six sources: project progress system, cost system, daily site reports, contractor reports, inspection team records, and supervisor inspection forms to form a business source table. After completing the preprocessing rules, integrate the field data in each table with project identifier + date as the joint primary key to build a unified fact wide table. Specifically, in S1, the details are as follows: Six categories of standardized business source tables are collected, including the progress system table, the daily on-site labor progress table, the cost system table, the contractor cost reporting form, the inspection team quality inspection form, and the supervisory inspection form. The field data in each business source table includes project identifier, business date, cumulative planned progress / actual progress, planned / actual cumulative cost, total number of inspections / number of qualified items on the day, cumulative actual progress / planned progress, environmental disaster type, warning level, work stoppage notice status, key process implementation status, actual number of employees present, equipment availability rate, key process implementation status, work stoppage identifier, environmental severity, resumption inspection status, work stoppage identifier, overtime hours, update time, and metadata, etc. A unified preprocessing rule is applied to the field data in each table. This preprocessing rule includes field mapping, date standardization, and deduplication within the source. Field mapping maps custom business fields from each source to internal standard fields through configuration files. Date standardization converts all dates to the YYYY-MM-DD standard format. Deduplication within the source retains the latest entry for multiple records of the same project and date. After preprocessing, the tables are integrated using project identifier + date as the composite primary key to construct a unified fact wide table. Simultaneously, external environmental event data is integrated, and the text of typhoon, rainstorm, strong wind, and flood events is uniformly encoded. Blue / yellow / orange / red warning colors are mapped to numerical levels 1 to 4, generating a standard field for environmental severity.

[0029] S2. Based on the unified fact wide table, construct a confidence calculation model and calculate the confidence parameter for each field of data; Specifically, in S2, the details are as follows: Based on a unified fact wide table, a confidence score calculation model is constructed to verify whether each field in the table carries a confidence score parameter. When the field data in each table has a built-in confidence score parameter, it is directly reused; when the field data in each table does not have a confidence score parameter, the confidence score calculation model automatically calculates the confidence score parameter in the range of 0 to 1. The expression for the confidence calculation model is:

[0030] In the formula, (Source Authority) The source authority score indicates the prior reliability of the source entity. For example, different basic weights can be preset for the schedule system, cost system, inspection team, supervisor or manual entry. (Timeliness) is the timeliness score, which indicates the degree of delay between the source update time and the business date. The closer the update time is to the business date, the higher the timeliness score. (Completeness) is the completeness score. Full marks are awarded if there are no missing data in the current record's fields, and points are deducted for missing data in the corresponding fields.

[0031] S3. Calculate the absolute difference between two fields of data in the unified fact wide table, and set a conflict threshold for each field. When the absolute difference is less than or equal to the conflict threshold, use a confidence-weighted average to generate a valid fact value. When the absolute difference is greater than the conflict threshold, execute the high-confidence priority selection strategy as the valid fact value. Specifically, in S3, such as Figure 2 As shown, the details are as follows: 1) Calculate the absolute difference between two field data. Its expression is:

[0032] 2) Conflict threshold determination: Three conflict thresholds are set for each field of data: actual progress conflict threshold of 0.02, cumulative cost conflict threshold of 2000, and inspection pass rate conflict threshold of 0.04. when If the absolute difference between two data fields is less than or equal to the conflict threshold, it is determined that the difference is within a tolerable range. A confidence-weighted average is then used to generate valid fact values, the expression of which is: Valid fact value = (A value × A confidence level + B value × B confidence level) / (A confidence level + B confidence level); when > Conflict threshold: If the absolute difference between two fields is found to be significantly conflicting, a high confidence priority selection strategy is adopted, that is, the field with the higher confidence parameter value is selected as the valid fact value. Write all valid fact values ​​into the unified fact wide table to complete the expansion of the wide table.

[0033] S4. Calculate derived indicators based on the data fields and valid fact values ​​of the unified fact wide table. These derived indicators include at least the inspection pass rate, schedule delay days, cost deviation rate, work stoppage flag, and capacity coefficient. The inspection pass rate is determined by the ratio of the number of qualified items inspected on a given day to the total number of items inspected on that day, or by using valid fact values ​​obtained by fusing inspection pass rates from multiple sources. The schedule delay days are determined by multiplying the positive difference between the cumulative planned progress and the cumulative actual progress by a preset duration factor. The cost deviation rate is determined by the ratio of the difference between the actual cumulative cost and the planned cumulative cost to the planned cumulative cost. The work stoppage flag is determined based on one of the following: work stoppage notice status, key process implementation status, and actual attendance. The capacity coefficient is calculated by weighting the attendance ratio of actual attendance to the normal attendance baseline, equipment availability rate, and key process implementation status. Based on these derived indicators, the current construction stage type of the project is comprehensively determined. The stage types include six categories: early warning preparation, complete work stoppage, resumption inspection, capacity recovery, expedited recovery, and normal construction. Specifically, in S4, the details are as follows: For the inspection pass rate, if the inspection pass rate field already exists in the unified fact wide table, it can be reused directly; if the inspection pass rate field does not exist, it can be obtained by calculating the ratio between the number of qualified items inspected on the day and the total number of items inspected on the day after merging in the table, or the result of the effective fact value after confidence fusion can be used as the inspection pass rate field. For the number of days of schedule delay, read the cumulative planned progress and cumulative actual progress in the unified fact wide table and compare the two values; when the cumulative planned progress is higher than the cumulative actual progress, it means that the current actual completion level is behind the planned completion level and the construction progress is lagging behind. Multiply the difference between the two by the schedule scale factor to obtain the number of days of schedule delay. The schedule scale factor is used to convert the dimensionless schedule deviation ratio into time-dimensional delay duration. Its value is pre-configured based on the total project duration, phased planned duration, time window length, or critical path duration. In this embodiment, a value of 60.0 is preferred. The cost deviation rate is calculated using the expression "Cost Deviation Rate = (Actual Cumulative Cost - Planned Cumulative Cost) / Planned Cumulative Cost", which indicates the degree of deviation between the actual cumulative cost and the planned cumulative cost. The schedule performance index and cost performance index are used as auxiliary verification indicators, and their expressions are "Schedule performance index = Cumulative actual progress / Cumulative planned progress" and "Cost performance index = (Cumulative actual progress × Planned cumulative cost) / Actual cumulative cost", respectively. The work stoppage flag indicates whether the project segment is in a completely stopped state within the current time window. It is a binary variable, where 1 represents work stoppage and 0 represents no work stoppage. When the work stoppage flag is provided in the unified fact wide table, its value is used. If the wide table does not provide it directly, the determination is based on a comprehensive deduction of the work stoppage notification status, key process implementation status, and actual attendance, including: if any of the following conditions are met: work stoppage notification status = 1, key process implementation status = 0, or actual attendance is lower than the preset work stoppage attendance threshold, the work stoppage flag is assigned the value 1, representing a complete work stoppage; otherwise, it is assigned the value 0. The preset work stoppage attendance threshold is set in combination with the project scale and labor allocation. In this implementation, actual_attendance≤1 is preferred as an auxiliary judgment condition for work stoppage. The capacity coefficient represents the recovery level of current construction capacity relative to normal construction capacity. It is calculated using a weighted model, including: first, calculating the attendance ratio based on the actual number of employees and the baseline number of employees in the table. The baseline number of employees is determined based on the construction organization plan and historical normal attendance samples; this implementation preferably uses the 95th percentile of historical samples. The baseline number of employees is determined based on the construction organization plan, historical normal daily work attendance levels, planned labor allocation for the same stage, or historical sample quantiles. Then, the attendance ratio, equipment availability rate, and key process implementation status are linearly combined according to preset weights, resulting in the capacity coefficient as expressed by "Capacity Coefficient = 0.45 × Attendance Ratio + 0.45 × Equipment Availability Rate + 0.10 × Key Process Implementation Status". When the actual number of employees, the baseline number of employees, the equipment availability rate, or the key process implementation status are missing, a degenerate estimation method is used, taking "1 - Stoppage Flag" as a substitute value for the capacity coefficient. The capacity coefficient defaults to 0 in the stoppage state and 1 in the non-stoppage state, ensuring the continuous operation of subsequent rule inference links. For each stage status, the system reads the shutdown flag, environmental severity, resumption inspection status, capacity coefficient, schedule delay days, and overtime hours derived from the unified fact wide table, and performs multi-condition status matching and identification, including: if the shutdown flag = 1, it is identified as a complete shutdown stage; if the environmental severity reaches the warning threshold and the capacity continues to decline, it is identified as a warning preparation stage; if it is in the resumption inspection stage and the inspection pass rate is lower than the set threshold, it is identified as a resumption inspection stage; if the capacity coefficient continues to rise but has not yet reached the normal level, it is identified as a capacity recovery stage; if the schedule delay days are large, accompanied by an increase in overtime hours and an increase in the capacity coefficient, it is identified as a rush work recovery stage; when the identification conditions of the complete shutdown stage, warning preparation stage, resumption inspection stage, capacity recovery stage, and rush work recovery stage are not met, it is marked as a normal construction stage. Through the above steps, the current construction stage of the project is comprehensively determined by combining derived indicators, including six construction stage statuses: early warning preparation, complete shutdown, resumption inspection, production capacity recovery, expedited recovery, and normal construction. All derived indicators are written into a unified fact wide table to complete the expansion of the fact layer of the wide table and form a complete fact dataset containing derived indicators.

[0034] S5. Based on the unified fact wide table field data and construction phase, build a rule engine to calculate the three-dimensional baseline risks of progress, cost, and safety. Enter multiple rules into the rule engine and traverse them. When multiple rules conflict, the adjudication mechanism is activated to take into account multiple factors to adopt, retain or suppress the hit rules and record the corresponding information to form a two-level processing result of rule hit and rule adjudication. Specifically, in S5, the details are as follows: Based on derived indicators, three-dimensional baseline risks for schedule, cost, and safety are calculated. These baseline risks are used to normalize continuous indicators such as schedule deviations, cost deviations, and safety pressures to a preset risk range, preferably between 0 and 1. Schedule baseline risk can be determined based on the ratio of schedule delay days to a schedule delay metric parameter; cost baseline risk can be determined based on the cost deviation rate and the cost sensitivity coefficient; and safety baseline risk can be determined based on the ratio of overtime hours to a safety pressure metric parameter. The expressions are as follows:

[0035] In the formula, This is a parameter for measuring schedule delays. For cost and risk sensitivity coefficient, These are safety pressure scale parameters; the aforementioned scale parameters and sensitivity coefficients can be configured in conjunction with the total project duration, phased planned duration, cost control targets, historical sample distribution, and expert experience, and parameter calibration can be completed based on historical data playback; in this embodiment, the preferred configuration is... =50, =0.8, =20; Multiple rules are entered into and iterated through in the rule engine. When multiple rules simultaneously trigger the same risk, an adjudication mechanism is activated to resolve the conflict between the multiple triggering rules. The adjudication mechanism considers at least four evaluation factors, including rule priority, specificity of conditions, temporal proximity, and source credibility, and performs adoption, retention, or suppression on the triggering rules respectively. For rules determined to be adopted, their triggering conditions, scope of action, risk increment, and suggested actions are recorded. For rules determined to be suppressed, the reasons for suppression, the corresponding adjudication basis, and the ineffective actions are recorded.

[0036] S6. The three-dimensional risks are linearly weighted using preset weighting coefficients to obtain a comprehensive risk score. Based on preset classification thresholds, three risk levels (low, medium, and high) are mapped, and corresponding on-site control and disposal suggestions are output according to the matching rules. Specifically, in S6, the details are as follows: The comprehensive risk score is obtained by linearly weighting the three-dimensional risks using preset weighting coefficients. The calculation formula is as follows:

[0037] After obtaining the comprehensive risk score, the system maps any risk level (high, medium, or low) based on preset thresholds, and simultaneously outputs a three-dimensional risk vector and on-site control and disposal suggestions corresponding to the hit rules. Each rule is pre-bound with corresponding on-site control and disposal suggestions. The system summarizes the suggestions of all effective rules, removes duplicates, and generates standardized management suggestions for rectification, scheduling, and inspection, such as "adding mandatory rest windows to meet deadlines, on-site supervision of key processes, and rolling adjustments to material procurement plans."

[0038] S7. Integrate input fact snapshots, original values ​​and confidence levels from various sources, full records of conflict fusion, calculation process of derived indicators, stage identification labels, hit rules and adjudication records, baseline risk, rule risk increments, multidimensional risk results, comprehensive scores, risk levels, and disposal recommendations to construct an auditable and replayable structured chain of evidence. Specifically, in S7, the details are as follows: The evidence chain is a standardized structured text / data table that fully reproduces the entire calculation logic of risk outcomes. Its core comprises 10 types of information blocks, including the following: 1) Basic snapshot: Project identifier, business date, construction phase label, environmental severity, work stoppage label, capacity coefficient, overtime hours, and snapshot of core indicators after integration; 2) Fusion details: Original values ​​of all multi-source fields, corresponding confidence levels, differences, conflict thresholds, fusion strategies, and total number of conflicting fields for the day; 3) Calculate all values ​​for derived indicators; 4) Initial values ​​of three-dimensional baseline risk; 5) Rule reasoning details: All hit rule IDs, trigger conditions, risk increments, suppressed rules and reasons for suppression; 6) The final three-dimensional risk values ​​of schedule, cost, and safety after adding the increments; 7) Overall risk score and corresponding risk level; 8) List of on-site handling recommendations and total number of recommendations; The evidence chain is permanently stored, allowing auditors and project managers to trace back to the original basis of any risk value, supporting responsibility definition, monthly review, and rule parameter optimization.

[0039] S8. When new business facts are detected, historical records are corrected, or rule versions are updated, the changed field data is identified, and only the affected part is recalculated locally and the evidence chain version is updated synchronously, without the need for a full data scan and recalculation.

[0040] Specifically, in S8, the details are as follows: Construct a dependency mapping relationship of "field data - fusion process - derived indicators - rules" to achieve local incremental calculation, including the following: Triggering conditions: Adding new business data for the current day, correcting historical data for a certain date, updating rule thresholds / weight parameters; Execution logic: 1) Identify changed field data; 2) Retrieve all fusion calculations and derived indicators that depend on the data in this field; 3) Locate the subset of stage rules that depend on the derived indicators; 4) Only recalculate the three-dimensional risk, comprehensive score, and evidence chain locally for the associated links; keep the original results unchanged for unassociated historical data and rules; synchronously update the evidence chain version number, record the source of the change trigger, significantly reduce server computing power consumption, and meet the needs of real-time streaming early warning on site.

[0041] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0042] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for multi-source credibility fusion rule-based reasoning and evidence chain generation for comprehensive engineering risks, characterized in that, Includes the following steps: S1. Collect business data to form a business source table. After completing the preprocessing rules, integrate the field data in each table with project identifier + date as the composite primary key to build a unified fact wide table. S2. Based on the unified fact wide table, construct a confidence calculation model and calculate the confidence parameter for each field of data; S3. Calculate the absolute difference between two fields of data in the unified fact wide table, and set a conflict threshold for each field. When the absolute difference is less than or equal to the conflict threshold, use a confidence-weighted average to generate a valid fact value. When the absolute difference is greater than the conflict threshold, execute the high-confidence priority selection strategy as the valid fact value. 1) Calculate the absolute difference between two field data. Its expression is: ， 2) Conflict threshold determination: Three conflict thresholds are set for each field of data, namely the actual progress conflict threshold, the cumulative cost conflict threshold, and the inspection pass rate conflict threshold; when If the difference between two data fields is less than or equal to the conflict threshold, it is determined that the difference is within a tolerable range. A confidence-weighted average is then used to generate valid fact values, the expression of which is: Valid fact value = (A value × A confidence level + B value × B confidence level) / (A confidence level + B confidence level); when > Conflict threshold: If the difference between two fields is determined to be significantly conflicting, a high confidence priority selection strategy is adopted, that is, the field data with the higher confidence parameter value is selected as the valid fact value. Write all valid fact values ​​into the unified fact wide table to complete the expansion of the wide table; S4. Calculate derived indicators based on the data fields and valid fact values ​​of the unified fact wide table. These derived indicators include at least the inspection pass rate, schedule delay days, cost deviation rate, work stoppage flag, and capacity coefficient. The inspection pass rate is determined by the ratio of the number of qualified items inspected on a given day to the total number of items inspected on that day, or by using valid fact values ​​obtained by fusing inspection pass rates from multiple sources. The schedule delay days are determined by multiplying the positive difference between the cumulative planned progress and the cumulative actual progress by a preset duration factor. The cost deviation rate is determined by the ratio of the difference between the actual cumulative cost and the planned cumulative cost to the planned cumulative cost. The work stoppage flag is determined based on one of the following: work stoppage notice status, key process implementation status, and actual attendance. The capacity coefficient is calculated by weighting the attendance ratio of actual attendance to the normal attendance baseline, equipment availability rate, and key process implementation status. Based on these derived indicators, the current construction phase type of the project is comprehensively determined. The phase types include six categories: early warning preparation, complete work stoppage, resumption inspection, capacity recovery, expedited recovery, and normal construction. S5. Based on the unified fact wide table field data and construction phase, build a rule engine to calculate the three-dimensional baseline risks of progress, cost, and safety. Enter multiple rules into the rule engine and traverse them. When multiple rules conflict, the adjudication mechanism is activated to take into account multiple factors to adopt, retain or suppress the hit rules and record the corresponding information to form a two-level processing result of rule hit and rule adjudication. S6. The three-dimensional risks are linearly weighted using preset weighting coefficients to obtain a comprehensive risk score. Based on preset classification thresholds, three risk levels (low, medium, and high) are mapped, and corresponding on-site control and disposal suggestions are output according to the matching rules. S7. Integrate input fact snapshots, original values ​​and confidence levels from various sources, full records of conflict fusion, calculation process of derived indicators, stage identification labels, hit rules and adjudication records, baseline risk, rule risk increments, multidimensional risk results, comprehensive scores, risk levels, and disposal recommendations to construct an auditable and replayable structured chain of evidence. S8. When new business facts are detected, historical records are corrected, or rule versions are updated, the changed field data is identified, and only the affected part is recalculated locally and the evidence chain version is updated synchronously, without the need for a full data scan and recalculation.

2. The method for multi-source credibility fusion rule reasoning and evidence chain generation of comprehensive engineering risks according to claim 1, characterized in that, Specifically, in S1, the details are as follows: The data in each business source table includes the following fields: project identifier, business date, cumulative planned progress / actual progress, planned / actual cumulative cost, total number of inspections / number of qualified items on the day, cumulative actual progress / planned progress, environmental disaster type, warning level, work stoppage notice status, key process implementation status, actual number of employees present, equipment availability rate, key process implementation status, work stoppage identifier, environmental severity, resumption inspection status, work stoppage identifier, overtime hours, and update time metadata. A unified preprocessing rule is applied to the field data in each table. This preprocessing rule includes field mapping, date standardization, and deduplication within the source. Field mapping involves mapping custom business fields from each source to internal standard fields using configuration files. Date standardization involves converting all dates to the YYYY-MM-DD standard format. Deduplication within the source involves retaining the latest entry from multiple records of the same project and date. After preprocessing, the tables are integrated using the project identifier + date as the composite primary key to construct a unified wide fact table. Simultaneously, external environmental event data is integrated.

3. The method for multi-source credibility fusion rule reasoning and evidence chain generation of comprehensive engineering risks according to claim 1, characterized in that, Specifically, in S2, the details are as follows: Based on a unified fact wide table, a confidence score calculation model is constructed to verify whether each field in the table carries a confidence score parameter. When the field data in each table has a built-in confidence score parameter, it is directly reused; when the field data in each table does not have a confidence score parameter, the confidence score calculation model automatically calculates the confidence score parameter in the range of 0 to 1. The expression for the confidence calculation model is: ， In the formula, To ensure the source is authoritative and rated, For timeliness scoring, Score for completeness.

4. The method for multi-source credibility fusion rule reasoning and evidence chain generation of comprehensive engineering risks according to claim 1, characterized in that, Specifically, in S4: For the inspection pass rate, if the inspection pass rate field already exists in the unified fact wide table, it can be reused directly; if the inspection pass rate field does not exist, it can be obtained by calculating the ratio between the number of qualified items inspected on the day and the total number of items inspected on the day after merging in the table, or the result of the effective fact value after confidence fusion can be used as the inspection pass rate field. For the number of days of schedule delay, read the cumulative planned progress and cumulative actual progress in the unified fact wide table and compare the two values; when the cumulative planned progress is higher than the cumulative actual progress, it means that the current actual completion level is behind the planned completion level and the construction progress is lagging behind. Multiply the difference between the two by the schedule scale factor to obtain the number of days of schedule delay. The cost deviation rate is calculated using the expression "Cost Deviation Rate = (Actual Cumulative Cost - Planned Cumulative Cost) / Planned Cumulative Cost", which indicates the degree of deviation between the actual cumulative cost and the planned cumulative cost. The schedule performance index and cost performance index are used as auxiliary verification indicators, and their expressions are "Schedule performance index = Cumulative actual progress / Cumulative planned progress" and "Cost performance index = (Cumulative actual progress × Planned cumulative cost) / Actual cumulative cost" respectively. The work stoppage flag is used to indicate whether the project segment is in a completely stopped state during the current time window. It is a binary variable, where 1 indicates work stoppage and 0 indicates no work stoppage. When a stoppage flag is provided in the unified fact wide table, its value is used. If the wide table does not provide this information directly, the determination will be based on a comprehensive deduction of the stoppage notice status, key process operation status, and actual attendance in the table. This includes: if any of the following conditions are met, such as stoppage notice status = 1, key process operation status = 0, or actual attendance is lower than the preset stoppage attendance threshold, the stoppage flag will be assigned = 1, representing a complete stoppage; otherwise, the value will be assigned 0. The capacity coefficient represents the recovery level of current construction capacity relative to normal construction capacity. It is calculated using a weighted model, which includes: first, calculating the attendance ratio based on the actual number of employees and the normal attendance baseline; then, linearly combining the attendance ratio, equipment availability rate, and key process status according to preset weights, resulting in the capacity coefficient as expressed by "Capacity Coefficient = 0.45 × Attendance Ratio + 0.45 × Equipment Availability Rate + 0.10 × Key Process Status". When the actual number of employees, the normal attendance baseline, the equipment availability rate, or the key process status are missing, a degenerate estimation method is used, taking "1 - Stoppage Flag" as a substitute value for the capacity coefficient. The default value for the capacity coefficient is 0 in the stoppage state and 1 in the non-stoppage state. For each stage status, the system reads the shutdown flag, environmental severity, resumption inspection status, capacity coefficient, schedule delay days, and overtime hours derived from the unified fact wide table, and performs multi-condition status matching and identification, including: if the shutdown flag = 1, it is identified as a complete shutdown stage; if the environmental severity reaches the warning threshold and the capacity continues to decline, it is identified as a warning preparation stage; if it is in the resumption inspection stage and the inspection pass rate is lower than the set threshold, it is identified as a resumption inspection stage; if the capacity coefficient continues to rise but has not yet reached the normal level, it is identified as a capacity recovery stage; if the schedule delay days are large, accompanied by an increase in overtime hours and an increase in the capacity coefficient, it is identified as a rush work recovery stage; when the identification conditions of the complete shutdown stage, warning preparation stage, resumption inspection stage, capacity recovery stage, and rush work recovery stage are not met, it is marked as a normal construction stage. Through the above steps, the current construction stage of the project is comprehensively determined by combining derived indicators, including six construction stage statuses: early warning preparation, complete shutdown, resumption inspection, production capacity recovery, expedited recovery, and normal construction. All derived indicators are written into a unified fact wide table to complete the expansion of the fact layer of the wide table and form a complete fact dataset containing derived indicators.

5. The method for multi-source credibility fusion rule reasoning and evidence chain generation of comprehensive engineering risks according to claim 1, characterized in that, Specifically, in S5, the details are as follows: Based on derived indicators, the baseline risks for schedule, cost, and safety are calculated, with the following expressions: ， ， ， In the formula, This is a parameter representing the time delay. For cost and risk sensitivity coefficient, For safety pressure scale parameters; When multiple rules are entered into and iterated through the rule engine, and multiple rules simultaneously trigger the same risk, an adjudication mechanism is activated to resolve the conflict between the multiple triggering rules. The adjudication mechanism integrates four evaluation factors: rule priority, specificity of conditions, temporal proximity, and source credibility. For multiple rules that have been hit, it performs adoption, retention, or suppression respectively. For rules that are determined to be adopted, its hit conditions, scope of action, risk increment, and suggested actions are recorded. For rules that are determined to be suppressed, the reasons for suppression, the corresponding adjudication basis, and the content of the actions that have not taken effect are recorded.

6. The method for multi-source credibility fusion rule reasoning and evidence chain generation of comprehensive engineering risks according to claim 1, characterized in that, Specifically, in S6: The comprehensive risk score is obtained by linearly weighting the three-dimensional risks using preset weighting coefficients. The calculation formula is as follows: 。 7. An apparatus for implementing the method according to any one of claims 1 to 6, characterized in that, include: The module includes a multi-source data access and alignment module, a confidence scoring module, a conflict perception and fusion module, a derived fact and stage identification module, a staged rule reasoning and adjudication module, a risk aggregation output module, a structured evidence chain generation module, and an incremental recalculation scheduling module. Each module works in series, and the incremental recalculation scheduling module establishes a field dependency mapping relationship with all other modules to realize dynamic recalculation of local links.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the engineering comprehensive risk multi-source credibility fusion rule reasoning and evidence chain generation method according to any one of claims 1 to 6.

Images