A big data supervision method and system based on construction engineering quality detection

Abstract

The application discloses a kind of big data supervision methods based on construction engineering quality detection, including the following steps: acquisition construction engineering quality detection supervision business data, form standardization supervision data and object identification set;Reverse candidate space is clipped, and candidate detection fact unit set is formed;Extract detection fact state, form candidate detection fact state set;Original detection data is implemented specification back calculation processing, and detection conclusion can support state set is formed;Original detection time sequence is implemented Matrix Profile analysis, and curve reuse counterproof state set is formed;Fellegi-Sunter probability record linking method is used to calculate, and detection fact probability set is formed;Multi-candidate detection chain global decision is executed, and only credible detection chain is obtained, and consistency verification is implemented to only credible detection chain, and generates supervision early warning information.The application improves the early warning accuracy in construction engineering quality detection supervision.

Description

Technical Field

[0001] This invention relates to the field of engineering quality supervision technology, and in particular to a big data supervision method and system based on construction engineering quality testing. Background Technology

[0002] Construction project quality testing is a crucial component of project quality control and final acceptance. Currently, testing supervision typically relies on the separate storage and management of relevant data by the testing agency's business systems, testing equipment systems, sample transfer ledgers, testing report systems, and the regulatory platform of the competent authority. These various data types often differ in terms of collection methods, field formats, time representation, object identification, and business definitions. Therefore, it is usually necessary to confirm the correspondence between samples, equipment, original testing data, and testing reports through manual comparison, experience verification, or simple rule matching.

[0003] Based on this, existing technologies still have significant shortcomings: On the one hand, the lack of a unified, stable, and traceable object identification and candidate fact construction mechanism among data from different sources makes it difficult to form a credible integrated chain of testing facts among test reports, original test data, sample flow information, and equipment information, easily leading to problems such as incorrect chains, broken chains, and duplicate chains; on the other hand, existing regulatory methods mostly rely on static field comparison or manual review, lacking a back-calculation support judgment mechanism combined with testing specifications, making it difficult to determine whether the conclusions of the test report are truly derived from the corresponding original test data; at the same time, the ability to identify abnormal reuse, duplicate fragments, and forgery risks of original test time-series data is insufficient, making it difficult to detect hidden problems such as curve reuse and data splicing in a timely manner; in addition, when multiple candidate test chains exist simultaneously, existing technologies usually lack a global adjudication mechanism for conflict relationships, making it difficult to determine a unique credible test chain from multiple candidate results, thus affecting the accuracy of subsequent consistency verification and regulatory early warning.

[0004] Therefore, how to provide a big data supervision method and system based on construction project quality inspection is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] One objective of this invention is to propose a big data supervision method and system based on construction project quality inspection. This invention improves the accuracy of inspection fact identification, the reliability of inspection conclusions supporting judgment, and the precision of supervision and early warning in construction project quality inspection by reverse candidate space pruning, standard back-calculation to support judgment, curve reuse of original inspection time sequence for counter-evidence, enhanced inspection fact matching, and global adjudication of multi-candidate inspection chains.

[0006] A big data monitoring method based on construction project quality inspection according to an embodiment of the present invention includes the following steps: Collect construction project quality inspection and supervision business data, and perform standardization and object identification coding to form a standardized supervision data and object identification set; Based on standardized regulatory data, reverse candidate space pruning is performed, and cross-source mapping and grouping are carried out according to the object identifier set to form a candidate detection fact unit set; Based on the candidate detection fact unit set, the detection fact state is extracted to form the candidate detection fact state set; According to the testing specifications of the corresponding testing items, the original testing data in the candidate testing fact unit set are subjected to standardized back-calculation processing to form a state set that can support the testing conclusions; Matrix Profile analysis is performed on the original detection time series in the candidate detection fact unit set to form a curve reuse proof by contradiction state set; The candidate detection fact state set, the detection conclusion support state set, and the curve reuse proof state set are merged into the enhanced detection fact matching state set, and the Fellegi-Sunter probability record linking method is used to calculate and form the probability set of the detection fact. Based on the probability set of the detected facts, the conflict relationship between candidate detected fact units is constructed, and a global decision on the multi-candidate detection chain is executed to complete the rejection confirmation and unique attribution confirmation of the candidate detected fact unit set, obtain a unique trusted detection chain, and perform consistency verification on the unique trusted detection chain to generate regulatory warning information.

[0007] Optionally, the formation of the standardized regulatory data and object identifier set specifically includes: It receives multi-source business data from construction project quality inspection and supervision scenarios, collects and organizes the multi-source business data according to data source, business stage, object category and time order, and divides the collected data into data items to be processed and related index data items to form the original supervision dataset. Standardization processing is performed on the data items to be processed in the original regulatory dataset, and the standardized data items are reorganized and stored according to a unified field structure to form standardized regulatory data. The associated index data items in the original regulatory dataset are processed by object identification encoding to generate a unique corresponding object identifier for each regulatory object. The object identifier is then mapped to the corresponding data record in the standardized regulatory data and registered and stored to form an object identifier set.

[0008] Optionally, the formation of the candidate detection fact unit set specifically includes: Test report data is read from standardized regulatory data. According to the extraction rule that the same test report corresponds to the same anchoring source, the report anchoring information that represents the constraint range of the test report is extracted in sequence, and then merged and organized to form a report anchoring dataset. The report anchoring dataset is invoked, and reverse candidate space pruning is performed on the data records in the standardized regulatory data that are associated with each report anchoring information. The candidate range is reduced according to the consistency constraint of the detection project, the adjacency constraint of the detection time, the correlation constraint of the detection conclusion, and the continuity constraint of the business flow. Data records that meet the preset pruning conditions are retained, and data records that do not meet the preset pruning conditions are removed. The retained data records are then clustered and organized according to the corresponding report anchoring information to form a candidate retained record set. The object identifier set is invoked to perform cross-source mapping and grouping of each retained record in the candidate retained record set, and the retained records that meet the mapping conditions are combined according to the same report anchor information to form a candidate mapping set; For each candidate mapping group, a detection fact unit construction process is performed. The candidate mapping groups are associated according to the report anchoring information, and the group chain and integrity are screened according to the detection business sequence. The results of the screening are determined as candidate detection fact units, and all candidate detection fact units are summarized to form a candidate detection fact unit set.

[0009] Optionally, the formation of the candidate detection fact state set specifically includes: Read each candidate detection fact unit in the candidate detection fact unit set, extract, merge and organize the feature data that characterizes the internal correlation of each candidate detection fact unit to form basic correlation feature data; Based on the basic correlation feature data, the time-related data in each candidate detection fact unit are extracted, sorted and merged, and converted into process time interval data according to the preset time merging rules; The process time interval data is called, and each process time interval is aggregated according to the same candidate detection fact unit. Adjacent process time intervals or preset process time intervals to be compared are determined as pairs of intervals to be judged. Based on the start and end times of each pair of intervals to be judged, Allen interval algebra is used to determine the preceding relationship, the contiguous relationship, the overlapping relationship, the inclusion relationship or the synchronization relationship, and form interval relationship judgment data. The interval relationship determination data is compared item by item with the preset process relationship template to form process time sequence feature data; The basic association feature data and process time sequence feature data are called, and the basic association feature data and process time sequence feature data belonging to the same candidate detection fact unit are merged and stored in a unified manner to form a detection fact state. The detection fact states corresponding to each candidate detection fact unit are summarized to form a candidate detection fact state set.

[0010] Optionally, the detection conclusions that support the acquisition of the state set may specifically include: Extract the original detection data and detection report conclusions corresponding to the corresponding detection items from each candidate detection fact unit, and organize them according to the belonging relationship of the same candidate detection fact unit to form standardized back-calculation input data; Match the testing specifications corresponding to the corresponding testing items, determine the value requirements, operation order, rounding requirements and conclusion correspondence requirements for the standard back calculation of the original testing data from the testing specifications, and establish a correspondence between the testing specifications and the standard back calculation input data to form the basis for standard back calculation; Based on the standard back-calculation criteria, the original detection data in the standard back-calculation input data are screened item by item and valid values ​​are selected. The selected valid values ​​are then substituted sequentially and continuously calculated according to the operation order to form the standard back-calculation results corresponding to each candidate detection fact unit. The back-calculation results are uniformly rounded according to the rounding requirements to form the rounded back-calculation results. The rounded back-calculation results are then compared with the corresponding test report conclusions item by item according to the requirements of the conclusions. When the test report conclusions are expressed in numerical form, the numerical difference between the rounded back-calculation results and the test report conclusions is used as the comparison result. When the test report conclusions are expressed in interval form, the consistency between the interval to which the rounded back-calculation results belong and the interval to which the test report conclusions belong is used as the comparison result. Based on the comparison results, a conclusion support determination is performed on each candidate detection fact unit. Candidate detection fact units that meet the preset support conditions are assigned a supportable mark, while candidate detection fact units that do not meet the preset support conditions are assigned an unsupportable mark. The supportable and unsupportable marks corresponding to each candidate detection fact unit are summarized to form a detection conclusion supportable state set.

[0011] Optionally, obtaining the curve multiplexing proof state set specifically includes: Read the candidate detection fact unit set, extract the original detection time series related to the corresponding detection item from each candidate detection fact unit, and align the original detection time series according to the unified sampling order, unified time direction and unified data length to form the time series analysis input data; The input data for time series analysis is divided into sliding segments according to a preset sliding window length and a preset sliding step size, so that each original detection time series is converted into multiple time series subsequences arranged in chronological order. Each time series subsequence is numbered and registered according to its candidate detection fact unit to form a time series subsequence set. Matrix Profile analysis is performed on each time series subsequence in the time series subsequence set. The current time series subsequence is used as the target subsequence. The similarity distance between the target subsequence and the other time series subsequences is calculated in turn. The minimum similarity distance corresponding to the target subsequence is determined as the contour value of the current target subsequence. At the same time, the contour values ​​are arranged according to the order of each target subsequence in the original detection time series to form the contour sequence corresponding to each candidate detection fact unit. Repeated segment identification and abnormal segment identification are performed on the contour sequence. The time sequence subsequence with contour value below the preset repetition threshold and continuous occurrence length reaching the preset repetition length is identified as a repeated time sequence segment. The time sequence subsequence with contour value above the preset abnormal threshold and continuous occurrence length reaching the preset abnormal length is identified as an abnormal time sequence segment. The repeated time sequence segments and abnormal time sequence segments are merged according to their respective candidate detection fact units to form time sequence counter-evidence data. Based on the temporal proof data, curve reuse proof is performed on each candidate detection fact unit. When there are repeated temporal segments across candidate detection fact units or abnormal temporal segments in a single candidate detection fact unit in the temporal proof data, the corresponding candidate detection fact unit is assigned a curve reuse proof mark. When there are no repeated temporal segments across candidate detection fact units and no abnormal temporal segments in a single candidate detection fact unit in the temporal proof data, the corresponding candidate detection fact unit is assigned a non-proof mark. The curve reuse proof marks and non-proof marks corresponding to each candidate detection fact unit are summarized to form a curve reuse proof state set.

[0012] Optionally, obtaining the probability set of the detected facts specifically includes: Read the candidate detection fact state set, the detection conclusion support state set, and the curve reuse rebuttal state set. According to the correspondence of the same candidate detection fact unit, align, associate, and merge the three types of state data. Aggregate the state data belonging to the same candidate detection fact unit into the same enhanced state item to form an enhanced detection fact matching state set. Each enhanced state item in the enhanced detection fact matching state set is matched and encoded. The state results of each enhanced state item, which reflect the degree of correlation of candidate detection fact units, the degree of conclusion support, and the degree of curve reuse and counter-evidence, are converted into corresponding state values. The matching comparison results are then formed according to the preset state arrangement order to form the matching comparison data corresponding to the candidate detection fact units. The Fellegi-Sunter probability record link calculation is performed on the matching comparison data. For each state value in each candidate detection fact unit, the probability of the current state value appearing under the matching condition and the probability of the current state value appearing under the non-matching condition are determined respectively. The logarithm of the ratio of the probability of the current state value appearing under the matching condition to the probability of the current state value appearing under the non-matching condition is determined as the matching weight corresponding to the state value. The matching weights corresponding to each state value in the same candidate detection fact unit are accumulated to form the comprehensive matching weight corresponding to the candidate detection fact unit. The comprehensive matching weight is subjected to probability transformation processing. The comprehensive matching weight is jointly calculated with the preset prior matching probability to form the probability of the detection fact corresponding to each candidate detection fact unit. The joint calculation is to combine the matching advantage corresponding to the comprehensive matching weight with the prior advantage corresponding to the preset prior matching probability, and convert the combined result into a probability value between zero and one. The probability of the detection fact corresponding to each candidate detection fact unit is registered, merged and summarized. The probability of the detection fact belonging to the same candidate detection fact unit is stored as the probability result of the candidate detection fact unit, forming a set of detection fact probability.

[0013] Optionally, obtaining the regulatory warning information specifically includes: Read the probability of each candidate detected fact unit in the probability set of the detected fact, and collect and compare each candidate detected fact unit according to the preset conflict judgment conditions, identify the candidate detected fact unit pairs that cannot be established at the same time, and register them to form a conflict relationship set; Based on the conflict relationship set and the probability set of the detection facts, a global decision on the multi-candidate detection chain is performed on each candidate detection fact unit to form a candidate credible detection chain combination. Perform exclusion confirmation and unique attribution confirmation on candidate trusted detection chain combinations, eliminate candidate detection fact units that do not meet the preset exclusion constraints, and determine the candidate trusted detection chain combinations after exclusion confirmation and unique attribution confirmation as the unique trusted detection chain; Extract the original detection data, detection report, and detection conclusion support status corresponding to each candidate detection fact unit in the unique credible detection chain, perform consistency verification on the original detection data and detection report, check the object correspondence, result correspondence, and time sequence correspondence, and form a consistency verification result; Based on the consistency verification results, the unique trusted detection chain that passes the consistency verification is registered as passed, and the unique trusted detection chain that fails the consistency verification is marked as abnormal. Regulatory warning information is generated according to the attribution information corresponding to the abnormal mark.

[0014] A big data monitoring system based on construction project quality inspection according to an embodiment of the present invention includes: The regulatory data processing module is used to collect business data on construction project quality inspection and supervision, and to perform standardization and object identification coding to form a standardized regulatory data and object identification set. The fact unit construction module is used to perform reverse candidate space pruning based on standardized regulatory data and to perform cross-source mapping and assembly according to the object identifier set to form a candidate detection fact unit set; The fact state generation module is used to extract the detection fact state based on the candidate detection fact unit set to form a candidate detection fact state set. The conclusion support judgment module is used to perform standardized back-calculation processing on the original test data in the candidate test fact unit set according to the test specifications of the corresponding test items, and form a set of test conclusion support states; The curve rebuttal identification module is used to perform MatrixProfile analysis on the original detection time sequence of the candidate detection fact unit set to form a curve reuse rebuttal state set; The probability matching calculation module is used to merge the candidate detection fact state set, the detection conclusion support state set, and the curve reuse proof state set into the enhanced detection fact matching state set, and to calculate the probability set of the detection fact using the Fellegi-Sunter probability record linking method. The link adjudication and early warning module is used to construct conflict relationships between candidate detection fact units based on the probability set of the detection facts, execute global adjudication of multiple candidate detection chains, complete the rejection confirmation and unique attribution confirmation of the candidate detection fact unit set, obtain a unique trusted detection chain, perform consistency verification on the unique trusted detection chain, and generate regulatory early warning information.

[0015] The beneficial effects of this invention are: This invention standardizes, identifies, encodes, prunes, and maps cross-source data for construction project quality inspection and supervision, thereby unifying relevant information that was originally scattered across different business processes and data sources into candidate test fact units. This improves the accuracy of the correlation between samples, original test data, test reports, and related business information, and reduces the probability of misaligned, broken, and duplicated links.

[0016] This invention extracts the detection fact state from candidate detection fact units and performs standardized back-calculation processing on the original detection data in conjunction with detection standards to form a set of states that can support the detection conclusion. This allows the determination of whether the conclusion of the detection report has corresponding original data support, thereby improving the reliability of the verification of detection conclusions and the credibility of regulatory results.

[0017] This invention performs Matrix Profile analysis on the original detection time series to form a curve reuse counter-evidence state set, which can identify repeated and abnormal segments in the original detection curve, and enhance the ability to identify the reuse, splicing or abnormal generation behavior of the original detection data.

[0018] This invention improves the ability to quantitatively judge the validity of detected facts by incorporating the candidate detection fact state set, the detection conclusion support state set, and the curve reuse proof state set into the enhanced detection fact matching state set, and combining the Fellegi-Sunter probability record linking method to form the detection fact validity probability set.

[0019] This invention constructs conflict relationships based on the probability set of detected facts and performs global adjudication of multiple candidate detection chains. It can determine a unique credible detection chain among multiple candidate chains and then generate regulatory early warning information by combining consistency verification. This improves the ability to identify anomalies, the accuracy of early warnings, and the overall regulatory efficiency in the supervision of construction project quality inspection. Attached Figure Description

[0020] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart of a big data-based monitoring method for construction project quality inspection proposed in this invention; Figure 2 This diagram illustrates the fact-finding and trust chain generation of a big data-based monitoring method for construction project quality inspection proposed in this invention. Detailed Implementation

[0021] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0022] refer to Figures 1-2 A big data-driven regulatory method for construction project quality inspection includes the following steps: Collect construction project quality inspection and supervision business data, and perform standardization and object identification coding to form a standardized supervision data and object identification set; Based on standardized regulatory data, reverse candidate space pruning is performed, and cross-source mapping and grouping are carried out according to the object identifier set to form a candidate detection fact unit set; Based on the candidate detection fact unit set, the detection fact state is extracted to form the candidate detection fact state set; According to the testing specifications of the corresponding testing items, the original testing data in the candidate testing fact unit set are subjected to standardized back-calculation processing to form a state set that can support the testing conclusions; Matrix Profile analysis is performed on the original detection time series in the candidate detection fact unit set to form a curve reuse proof by contradiction state set; The candidate detection fact state set, the detection conclusion support state set, and the curve reuse proof state set are merged into the enhanced detection fact matching state set, and the Fellegi-Sunter probability record linking method is used to calculate and form the probability set of the detection fact. Based on the probability set of the detected facts, the conflict relationship between candidate detected fact units is constructed, and a global decision on the multi-candidate detection chain is executed to complete the rejection confirmation and unique attribution confirmation of the candidate detected fact unit set, obtain a unique trusted detection chain, and perform consistency verification on the unique trusted detection chain to generate regulatory warning information.

[0023] In this embodiment, the formation of standardized regulatory data and object identifier sets specifically includes: The system receives multi-source business data from construction project quality testing and supervision scenarios, collects and organizes the multi-source business data according to data source, business stage, object category and time order, and divides the collected data into data items to be processed and related index data items to form the original supervision dataset. The multi-source business data includes testing entrustment data, sample circulation data, testing equipment data, original testing data, testing report data and supervision and disposal data. Standardization processing is performed on the data items to be processed in the original regulatory dataset. The standardization processing includes unifying field names, converting field formats, padding field values, merging duplicate records, unifying time formats, unifying units of measurement, removing outliers, and verifying source consistency. The standardized data items are then reorganized and stored according to a unified field structure to form standardized regulatory data. The associated index data items in the original regulatory dataset are processed by object identification encoding. The object identification encoding process includes extracting object type information, source identification information, business association information, sequence identification information and time association information, and combining and encoding them according to a preset encoding order to generate a unique corresponding object identifier for each regulatory object. The object identifier is then mapped to the corresponding data record in the standardized regulatory data and registered and stored to form an object identifier set.

[0024] In this embodiment, the formation of the candidate detection fact unit set specifically includes: Test report data is read from standardized regulatory data. According to the extraction rule that the same test report corresponds to the same anchoring source, the report anchoring information that represents the constraint scope of the test report is extracted in sequence and then merged and organized to form a report anchoring dataset. The report anchoring information is used to represent the scope of test objects, the scope of test time, the scope of test conclusions, and the scope of business associations that have a direct correspondence or a correspondence to be confirmed with the test report. The report anchoring dataset is invoked, and reverse candidate space pruning is performed on data records in the standardized regulatory data that are associated with each report anchoring information. The candidate range is reduced according to the following constraints: consistency of detection items, adjacency of detection time, association of detection conclusions, and continuity of business flow. Data records that simultaneously meet the preset pruning conditions are retained, while those that do not are removed. The retained data records are then clustered according to their corresponding report anchoring information to form a candidate retained record set. The consistency of detection items constraint is used to determine the correspondence between candidate records and report anchoring information in terms of detection items; the adjacency of detection time constraint is used to determine the adjacency of candidate records and report anchoring information within the time association range; the association of detection conclusions constraint is used to determine the association between candidate records and report anchoring information in terms of detection conclusion features; and the continuity of business flow constraint is used to determine the continuous connection between candidate records and report anchoring information in the detection business chain. The preset pruning conditions are the candidate retention conditions constituted by the above constraints. The object identifier set is invoked to perform cross-source mapping and combination on each retained record in the candidate retained record set. The cross-source mapping and combination includes identifying retained records from different sources but pointing to the same regulatory object or the same business chain position based on the correspondence relationship of object identifiers, identifying retained records with sequential connection relationship based on the association order between object identifiers, identifying retained records with detection implementation association based on the business flow continuity relationship, and combining retained records that meet the mapping conditions according to the same report anchoring information to form a candidate mapping set. For each candidate mapping group, a detection fact unit construction process is performed. The candidate mapping groups are associated according to the report anchoring information, and the chain and integrity screening are performed according to the order of the detection business. The results of the screening are determined as candidate detection fact units. All candidate detection fact units are summarized to form a candidate detection fact unit set. The chain and integrity screening refers to the sequential association of data items that have corresponding relationships after cross-source mapping, according to the formation order of the construction engineering quality detection business, so as to form a data chain that can reflect the same detection activity. After the data chain is formed, the integrity of the data chain is verified according to the integrity conditions to determine whether the constituent data items in the data chain are complete and whether the association relationship is closed.

[0025] In this embodiment, the formation of the candidate detection fact state set specifically includes: Read each candidate detection fact unit in the candidate detection fact unit set, extract, merge and organize the feature data that characterizes the internal correlation of each candidate detection fact unit to form basic correlation feature data; Based on the basic correlation feature data, the time-related data in each candidate detection fact unit are extracted, sorted and merged, and converted into process time interval data according to the preset time merging rules; The process time interval data is called, and each process time interval is aggregated according to the same candidate detection fact unit. Adjacent process time intervals or preset process time intervals to be compared are determined as pairs of intervals to be judged. Based on the start and end times of each pair of intervals to be judged, Allen interval algebra is used to determine the preceding relationship, the contiguous relationship, the overlapping relationship, the inclusion relationship or the synchronization relationship, and form interval relationship judgment data. The interval relationship determination data is compared item by item with the preset process relationship template to form process time sequence feature data. The preset process relationship template is a set of process relationship rules pre-established based on the construction engineering quality inspection business process. It is used to characterize the interval relationship types and their combination methods that should be satisfied between the process time intervals corresponding to each process stage in the candidate inspection fact unit, and is used to match and verify the interval relationship determination data. The basic association feature data and process time sequence feature data are called, and the basic association feature data and process time sequence feature data belonging to the same candidate detection fact unit are merged and stored in a unified manner to form a detection fact state. The detection fact states corresponding to each candidate detection fact unit are summarized to form a candidate detection fact state set.

[0026] This invention improves the accuracy of identifying the state of detection facts and the reliability of judging the process chain in construction engineering quality inspection by jointly extracting the internal correlation features and process time sequence features of candidate detection fact units, and combining process time interval transformation, Allen interval algebraic relationship judgment and process relationship template matching. It also enhances the integrity and temporal consistency of multi-source data correlation results, and provides a stable foundation for subsequent detection conclusion support judgment, curve reuse proof analysis and probability calculation of the validity of detection facts.

[0027] In this embodiment, the detection conclusions that support the acquisition of the state set specifically include: The original detection data and detection report conclusions corresponding to the corresponding detection items are extracted from each candidate detection fact unit, and are organized according to the belonging relationship of the same candidate detection fact unit to form standardized back-calculation input data. The detection items are the categories of detection items determined according to the report anchoring information corresponding to the candidate detection fact unit. Match the testing specifications corresponding to the corresponding testing items, determine the value requirements, operation order, rounding requirements and conclusion correspondence requirements for the standard back calculation of the original testing data from the testing specifications, and establish a correspondence between the testing specifications and the standard back calculation input data to form the basis for standard back calculation; Based on the standard back-calculation criteria, the original detection data in the standard back-calculation input data are screened item by item and valid values ​​are selected. The selected valid values ​​are then substituted sequentially and continuously calculated according to the operation order to form the standard back-calculation results corresponding to each candidate detection fact unit. The back-calculation results are uniformly rounded according to the rounding requirements. The uniform rounding process applies a uniform rounding standard to the results obtained from the back-calculation, according to the number of digits to retain, the number of significant figures, the rounding rules, or the rounding rules specified in the applicable testing specifications for the corresponding testing items. This ensures that the back-calculation results within the same candidate testing fact unit have a consistent numerical expression that can be directly compared with the testing report conclusions, forming the rounded back-calculation results. The rounded back-calculation results are then compared item by item with the corresponding testing report conclusions according to the requirements for corresponding conclusions. When the testing report conclusions are expressed in numerical form, the numerical difference between the rounded back-calculation results and the testing report conclusions is used as the comparison result. When the testing report conclusions are expressed in interval form, the consistency between the interval to which the rounded back-calculation results belong and the interval to which the testing report conclusions belong is used as the comparison result. Based on the comparison results, a conclusion support determination is performed on each candidate detection fact unit. Candidate detection fact units that meet the preset support conditions are assigned a supportable mark, while candidate detection fact units that do not meet the preset support conditions are assigned an unsupportable mark. The supportable and unsupportable marks corresponding to each candidate detection fact unit are summarized to form a detection conclusion supportable state set.

[0028] This invention effectively determines whether the conclusions of the test report have standardized support by performing standardized back-calculation, unified rounding, and corresponding comparison of the original test data. This improves the accuracy and consistency of conclusion verification in construction project quality testing, enhances the credibility of test results, and provides reliable support for subsequent fact confirmation, anomaly identification, and regulatory early warning.

[0029] In this embodiment, obtaining the state set of proof by contradiction using curve reuse specifically includes: Read the candidate detection fact unit set, extract the original detection time series related to the corresponding detection item from each candidate detection fact unit, and align the original detection time series according to the unified sampling order, unified time direction and unified data length to form the time series analysis input data; The input data for time series analysis is divided into sliding segments according to a preset sliding window length and a preset sliding step size, so that each original detection time series is converted into multiple time series subsequences arranged in chronological order. Each time series subsequence is numbered and registered according to its candidate detection fact unit to form a time series subsequence set. Matrix Profile analysis is performed on each time series subsequence in the time series subsequence set. The current time series subsequence is used as the target subsequence. The similarity distance between the target subsequence and the other time series subsequences is calculated in turn. The minimum similarity distance corresponding to the target subsequence is determined as the contour value of the current target subsequence. At the same time, the contour values ​​are arranged according to the order of each target subsequence in the original detection time series to form the contour sequence corresponding to each candidate detection fact unit. Repeated segment identification and abnormal segment identification are performed on the contour sequence. The time sequence subsequence with contour value below the preset repetition threshold and continuous occurrence length reaching the preset repetition length is identified as a repeated time sequence segment. The time sequence subsequence with contour value above the preset abnormal threshold and continuous occurrence length reaching the preset abnormal length is identified as an abnormal time sequence segment. The repeated time sequence segments and abnormal time sequence segments are merged according to their respective candidate detection fact units to form time sequence counter-evidence data. Based on the temporal proof data, curve reuse proof is performed on each candidate detection fact unit. When there are repeated temporal segments across candidate detection fact units or abnormal temporal segments in a single candidate detection fact unit in the temporal proof data, the corresponding candidate detection fact unit is assigned a curve reuse proof mark. When there are no repeated temporal segments across candidate detection fact units and no abnormal temporal segments in a single candidate detection fact unit in the temporal proof data, the corresponding candidate detection fact unit is assigned a non-proof mark. The curve reuse proof marks and non-proof marks corresponding to each candidate detection fact unit are summarized to form a curve reuse proof state set.

[0030] This invention effectively detects hidden problems such as the reuse of original curves, abnormal splicing, or abnormal generation during the quality inspection of construction projects by uniformly aligning the original detection time sequence, segmenting subsequences, performing contour analysis, and identifying duplicate and abnormal segments. It improves the ability to identify the authenticity of detection data and time sequence anomalies, enhances the reliability of curve verification results, and provides more credible support for subsequent fact matching, abnormal link confirmation, and regulatory early warning.

[0031] In this embodiment, obtaining the probability set of the detected facts specifically includes: Read the candidate detection fact state set, the detection conclusion support state set, and the curve reuse rebuttal state set. According to the correspondence of the same candidate detection fact unit, align, associate, and merge the three types of state data. Aggregate the state data belonging to the same candidate detection fact unit into the same enhanced state item to form an enhanced detection fact matching state set. Each enhanced state item in the enhanced detection fact matching state set is matched and encoded. The state results of each enhanced state item, which reflect the degree of correlation of candidate detection fact units, the degree of conclusion support, and the degree of curve reuse and counter-evidence, are converted into corresponding state values. The matching comparison results are then formed according to the preset state arrangement order to form the matching comparison data corresponding to the candidate detection fact units. The Fellegi-Sunter probability record link calculation is performed on the matching comparison data. For each state value in each candidate detection fact unit, the probability of the current state value appearing under the matching condition and the probability of the current state value appearing under the non-matching condition are determined respectively. The logarithm of the ratio of the probability of the current state value appearing under the matching condition to the probability of the current state value appearing under the non-matching condition is determined as the matching weight corresponding to the state value. The matching weights corresponding to each state value in the same candidate detection fact unit are accumulated to form the comprehensive matching weight corresponding to the candidate detection fact unit. The comprehensive matching weight is subjected to probability transformation processing. The comprehensive matching weight is jointly calculated with the preset prior matching probability to form the probability of the detection fact corresponding to each candidate detection fact unit. The joint calculation is to combine the matching advantage corresponding to the comprehensive matching weight with the prior advantage corresponding to the preset prior matching probability, and convert the combined result into a probability value between zero and one. The probability of the detection fact corresponding to each candidate detection fact unit is registered, merged and summarized. The probability of the detection fact belonging to the same candidate detection fact unit is stored as the probability result of the candidate detection fact unit, forming a set of detection fact probability.

[0032] This invention integrates and encodes the candidate detection fact state, the supportable state of the detection conclusion, and the curve reuse rebuttal state, and combines them with probability record linking to calculate the probability of the detection fact being established. This can improve the accuracy of multi-source evidence collaborative judgment in construction engineering quality inspection, enhance the quantitative assessment capability of the validity of detection facts, improve the credibility of data association results in complex detection scenarios, and enhance the reliability of subsequent conflict adjudication, anomaly identification, and regulatory early warning.

[0033] In this implementation method, obtaining regulatory early warning information specifically includes: Read the probability of each candidate detected fact unit in the probability set of the detected fact, and collect and compare each candidate detected fact unit according to the preset conflict judgment conditions, identify the candidate detected fact unit pairs that cannot be established at the same time, and register them to form a conflict relationship set; Based on the conflict relationship set and the probability set of the detection facts, a global decision on the multi-candidate detection chain is performed on each candidate detection fact unit. The global decision on the multi-candidate detection chain retains the candidate detection fact units with a high probability of success and that meet the link continuity requirement, and removes the candidate detection fact units with conflict relationships and a low probability of success, thus forming a candidate trusted detection chain combination. Perform exclusion confirmation and unique attribution confirmation on candidate trusted detection chain combinations, eliminate candidate detection fact units that do not meet the preset exclusion constraints, and determine the candidate trusted detection chain combinations after exclusion confirmation and unique attribution confirmation as the unique trusted detection chain; Extract the original detection data, detection report, and detection conclusion support status corresponding to each candidate detection fact unit in the unique credible detection chain, perform consistency verification on the original detection data and detection report, check the object correspondence, result correspondence, and time sequence correspondence, and form a consistency verification result; Based on the consistency verification results, the unique trusted detection chain that passes the consistency verification is registered as passed, and the unique trusted detection chain that fails the consistency verification is marked as abnormal. Regulatory warning information is generated according to the attribution information corresponding to the abnormal mark.

[0034] This invention performs conflict identification, global adjudication, exclusion confirmation, unique attribution confirmation, and consistency verification on candidate detection fact units. This enables the accurate determination of a unique and credible detection chain from multiple candidate detection chains, improving the accuracy of abnormal link identification in construction engineering quality inspection, the pertinence of regulatory early warning generation, and the credibility of regulatory results. It also enhances the ability to handle multi-source data conflicts and confirm detection facts in complex detection scenarios.

[0035] A big data monitoring system based on construction project quality inspection includes: The regulatory data processing module is used to collect business data on construction project quality inspection and supervision, and to perform standardization and object identification coding to form a standardized regulatory data and object identification set. The fact unit construction module is used to perform reverse candidate space pruning based on standardized regulatory data and to perform cross-source mapping and assembly according to the object identifier set to form a candidate detection fact unit set; The fact state generation module is used to extract the detection fact state based on the candidate detection fact unit set to form a candidate detection fact state set. The conclusion support judgment module is used to perform standardized back-calculation processing on the original test data in the candidate test fact unit set according to the test specifications of the corresponding test items, and form a set of test conclusion support states; The curve rebuttal identification module is used to perform MatrixProfile analysis on the original detection time sequence of the candidate detection fact unit set to form a curve reuse rebuttal state set; The probability matching calculation module is used to merge the candidate detection fact state set, the detection conclusion support state set, and the curve reuse proof state set into the enhanced detection fact matching state set, and to calculate the probability set of the detection fact using the Fellegi-Sunter probability record linking method. The link adjudication and early warning module is used to construct conflict relationships between candidate detection fact units based on the probability set of the detection facts, execute global adjudication of multiple candidate detection chains, complete the rejection confirmation and unique attribution confirmation of the candidate detection fact unit set, obtain a unique trusted detection chain, perform consistency verification on the unique trusted detection chain, and generate regulatory early warning information.

[0036] Example 1: To verify the feasibility of this invention in practice, it was applied to a construction project quality inspection and supervision scenario conducted simultaneously in a certain region. This scenario covers multiple building construction and supporting engineering projects, involving main structural material testing, on-site physical testing, witnessed sampling testing, and test report review. Before application, the supervision method in this region mainly relied on testing institutions submitting test reports separately, supervisors randomly checking original records in batches, and manually reviewing anomalies. The data standards between different systems were inconsistent, and sample flow information, original testing sequence, test reports, and business attribution were scattered across different platforms and devices. This resulted in problems such as reports being available but the original testing chain being difficult to close quickly, the risk of duplicate calls to original testing curves being difficult to identify in a timely manner, and multiple candidate testing chains coexisting within a similar time frame, leading to supervisory judgments relying on experience.

[0037] In this scenario, after integrating this invention, data from the regulatory platform, testing agency business systems, equipment acquisition terminals, and sample transfer terminals are first aggregated. The aggregated business data undergoes standardized processing and object identification encoding to form standardized regulatory data and object identifier sets. Subsequently, a reverse candidate space pruning is performed around the business anchor point corresponding to the test report, reducing the originally scattered large amount of regulatory data to a range of candidate data that may be related to the target testing item. Then, cross-source mapping and assembly are performed using the object identifier set, organizing sample transfer information, testing implementation information, original testing time sequence, and report information into candidate testing fact units. Internal correlation features are further extracted from the formed candidate testing fact units, and process-related times are converted into process time intervals. The reasonableness of the time sequence is identified by combining process relationship templates. Simultaneously, standard back-calculation is performed on the original testing data in each candidate testing fact unit according to the corresponding testing specifications, ensuring that the conclusions in the test report can form a supportable or unsupportable judgment result with the original data. For matters involving continuous curves, loading curves, or other detection time-series data, a Matrix Profile analysis is performed on the original detection time-series data to identify duplicate and abnormal segments, forming curve reuse rebuttal states. Then, the detection fact state, the supportable detection conclusion state, and the curve reuse rebuttal state are combined into an enhanced detection fact matching state set, and the probability of the detection fact being true is calculated using the Fellegi-Sunter probabilistic record linking method. When multiple highly similar candidate chains appear under the same regulatory matter, the system no longer simply retains a single locally optimal result. Instead, it combines conflict relationships, exclusion constraints, and link continuity requirements to perform a global adjudication of multiple candidate detection chains, ultimately outputting a unique trusted detection chain. Based on this trusted chain, consistency verification of object correspondence, result correspondence, and time-series correspondence is performed, automatically generating regulatory warning information.

[0038] To demonstrate the beneficial effects of this invention, it is compared with traditional regulatory methods. Traditional regulatory methods refer to a regulatory scheme in which, during the quality inspection of construction projects, the main reliance is on testing institutions submitting test reports, original records, and sample transfer information. Supervisory personnel then judge the correspondence between test data, the supporting relationships of test conclusions, and anomalies through manual sampling, experience comparison, and static rule verification. Specific comparative data are shown in Table 1. Table 1. Comparison of key performance characteristics between the method of this invention and traditional regulatory methods.

[0039] As shown in Table 1, the method of this invention demonstrates significant advantages over traditional regulatory methods in construction project quality inspection and supervision scenarios, particularly in candidate space compression, multi-source data matching, test conclusion support judgment, original curve reuse identification, unique credible detection chain formation, and early warning effectiveness. Specifically, the average number of candidates per report decreased from 14.7 to 2.8, indicating that this invention can significantly narrow the candidate range; the accuracy rate of multi-source data matching increased to 96.8%, and the accuracy rate of test conclusion support judgment increased to 95.1%, indicating that this invention has higher reliability in identifying test facts and verifying conclusion support; the detection rate of original curve reuse issues increased from 38.7% to 89.4%, showing that this invention has stronger capabilities in identifying abnormal time sequences; simultaneously, the unique credible detection chain formation rate and early warning effectiveness reached 94.8% and 91.55%, respectively, and the time for handling a single anomaly was shortened to 8.7 minutes, reducing manual review time to 108.0 hours, further proving that this invention can effectively improve the accuracy and efficiency of supervision.

[0040] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A big data-driven monitoring method for construction project quality inspection, characterized in that, Includes the following steps: Collect construction project quality inspection and supervision business data, and perform standardization and object identification coding to form a standardized supervision data and object identification set; Based on standardized regulatory data, reverse candidate space pruning is performed, and cross-source mapping and grouping are carried out according to the object identifier set to form a candidate detection fact unit set; Based on the candidate detection fact unit set, the detection fact state is extracted to form the candidate detection fact state set; According to the testing specifications of the corresponding testing items, the original testing data in the candidate testing fact unit set are subjected to standardized back-calculation processing to form a state set that can support the testing conclusions; Matrix Profile analysis is performed on the original detection time series in the candidate detection fact unit set to form a curve reuse proof by contradiction state set; The candidate detection fact state set, the detection conclusion support state set, and the curve reuse proof state set are merged into the enhanced detection fact matching state set, and the Fellegi-Sunter probability record linking method is used to calculate and form the probability set of the detection fact. Based on the probability set of the detected facts, the conflict relationship between candidate detected fact units is constructed, and a global decision on the multi-candidate detection chain is executed to complete the rejection confirmation and unique attribution confirmation of the candidate detected fact unit set, obtain a unique trusted detection chain, and perform consistency verification on the unique trusted detection chain to generate regulatory warning information.

2. The big data supervision method based on construction project quality inspection according to claim 1, characterized in that, The formation of the standardized regulatory data and object identifier set specifically includes: It receives multi-source business data from construction project quality inspection and supervision scenarios, collects and organizes the multi-source business data according to data source, business stage, object category and time order, and divides the collected data into data items to be processed and related index data items to form the original supervision dataset. Standardization processing is performed on the data items to be processed in the original regulatory dataset, and the standardized data items are reorganized and stored according to a unified field structure to form standardized regulatory data. The associated index data items in the original regulatory dataset are processed by object identification encoding to generate a unique corresponding object identifier for each regulatory object. The object identifier is then mapped to the corresponding data record in the standardized regulatory data and registered and stored to form an object identifier set.

3. The big data supervision method based on construction project quality inspection according to claim 1, characterized in that, The formation of the candidate detection fact unit set specifically includes: Test report data is read from standardized regulatory data. According to the extraction rule that the same test report corresponds to the same anchoring source, the report anchoring information that represents the constraint range of the test report is extracted in sequence, and then merged and organized to form a report anchoring dataset. The report anchoring dataset is invoked, and reverse candidate space pruning is performed on the data records in the standardized regulatory data that are associated with each report anchoring information. The candidate range is reduced according to the consistency constraint of the detection project, the adjacency constraint of the detection time, the correlation constraint of the detection conclusion, and the continuity constraint of the business flow. Data records that meet the preset pruning conditions are retained, and data records that do not meet the preset pruning conditions are removed. The retained data records are then clustered and organized according to the corresponding report anchoring information to form a candidate retained record set. The object identifier set is invoked to perform cross-source mapping and grouping of each retained record in the candidate retained record set, and the retained records that meet the mapping conditions are combined according to the same report anchor information to form a candidate mapping set; For each candidate mapping group, a detection fact unit construction process is performed. The candidate mapping groups are associated according to the report anchoring information, and the group chain and integrity are screened according to the detection business sequence. The results of the screening are determined as candidate detection fact units, and all candidate detection fact units are summarized to form a candidate detection fact unit set.

4. The big data supervision method based on construction project quality inspection according to claim 1, characterized in that, The formation of the candidate detection fact state set specifically includes: Read each candidate detection fact unit in the candidate detection fact unit set, extract, merge and organize the feature data that characterizes the internal correlation of each candidate detection fact unit to form basic correlation feature data; Based on the basic correlation feature data, the time-related data in each candidate detection fact unit are extracted, sorted and merged, and converted into process time interval data according to the preset time merging rules; The process time interval data is called, and each process time interval is aggregated according to the same candidate detection fact unit. Adjacent process time intervals or preset process time intervals to be compared are determined as pairs of intervals to be judged. Based on the start and end times of each pair of intervals to be judged, Allen interval algebra is used to determine the preceding relationship, the contiguous relationship, the overlapping relationship, the inclusion relationship or the synchronization relationship, and form interval relationship judgment data. The interval relationship determination data is compared item by item with the preset process relationship template to form process time sequence feature data; The basic association feature data and process time sequence feature data are called, and the basic association feature data and process time sequence feature data belonging to the same candidate detection fact unit are merged and stored in a unified manner to form a detection fact state. The detection fact states corresponding to each candidate detection fact unit are summarized to form a candidate detection fact state set.

5. The big data supervision method based on construction project quality inspection according to claim 1, characterized in that, The detection conclusions that support the acquisition of the state set specifically include: Extract the original detection data and detection report conclusions corresponding to the corresponding detection items from each candidate detection fact unit, and organize them according to the belonging relationship of the same candidate detection fact unit to form standardized back-calculation input data; Match the testing specifications corresponding to the corresponding testing items, determine the value requirements, operation order, rounding requirements and conclusion correspondence requirements for the standard back calculation of the original testing data from the testing specifications, and establish a correspondence between the testing specifications and the standard back calculation input data to form the basis for standard back calculation; Based on the standard back-calculation criteria, the original detection data in the standard back-calculation input data are screened item by item and valid values ​​are selected. The selected valid values ​​are then substituted sequentially and continuously calculated according to the operation order to form the standard back-calculation results corresponding to each candidate detection fact unit. The back-calculation results are uniformly rounded according to the rounding requirements to form the rounded back-calculation results. The rounded back-calculation results are then compared with the corresponding test report conclusions item by item according to the requirements of the conclusions. When the test report conclusions are expressed in numerical form, the numerical difference between the rounded back-calculation results and the test report conclusions is used as the comparison result. When the test report conclusions are expressed in interval form, the consistency between the interval to which the rounded back-calculation results belong and the interval to which the test report conclusions belong is used as the comparison result. Based on the comparison results, a conclusion support determination is performed on each candidate detection fact unit. Candidate detection fact units that meet the preset support conditions are assigned a supportable mark, while candidate detection fact units that do not meet the preset support conditions are assigned an unsupportable mark. The supportable and unsupportable marks corresponding to each candidate detection fact unit are summarized to form a detection conclusion supportable state set.

6. The big data supervision method based on construction project quality inspection according to claim 1, characterized in that, The specific steps for obtaining the curve multiplexing proof state set include: Read the candidate detection fact unit set, extract the original detection time series related to the corresponding detection item from each candidate detection fact unit, and align the original detection time series according to the unified sampling order, unified time direction and unified data length to form the time series analysis input data; The input data for time series analysis is divided into sliding segments according to a preset sliding window length and a preset sliding step size, so that each original detection time series is converted into multiple time series subsequences arranged in chronological order. Each time series subsequence is numbered and registered according to its candidate detection fact unit to form a time series subsequence set. Matrix Profile analysis is performed on each time series subsequence in the time series subsequence set. The current time series subsequence is used as the target subsequence. The similarity distance between the target subsequence and the other time series subsequences is calculated in turn. The minimum similarity distance corresponding to the target subsequence is determined as the contour value of the current target subsequence. At the same time, the contour values ​​are arranged according to the order of each target subsequence in the original detection time series to form the contour sequence corresponding to each candidate detection fact unit. Repeated segment identification and abnormal segment identification are performed on the contour sequence. The time sequence subsequence with contour value below the preset repetition threshold and continuous occurrence length reaching the preset repetition length is identified as a repeated time sequence segment. The time sequence subsequence with contour value above the preset abnormal threshold and continuous occurrence length reaching the preset abnormal length is identified as an abnormal time sequence segment. The repeated time sequence segments and abnormal time sequence segments are merged according to their respective candidate detection fact units to form time sequence counter-evidence data. Based on the temporal proof data, curve reuse proof is performed on each candidate detection fact unit. When there are repeated temporal segments across candidate detection fact units or abnormal temporal segments in a single candidate detection fact unit in the temporal proof data, the corresponding candidate detection fact unit is assigned a curve reuse proof mark. When there are no repeated temporal segments across candidate detection fact units and no abnormal temporal segments in a single candidate detection fact unit in the temporal proof data, the corresponding candidate detection fact unit is assigned a non-proof mark. The curve reuse proof marks and non-proof marks corresponding to each candidate detection fact unit are summarized to form a curve reuse proof state set.

7. The big data supervision method based on construction project quality inspection according to claim 1, characterized in that, The specific steps to obtain the probability set of the detected facts include: Read the candidate detection fact state set, the detection conclusion support state set, and the curve reuse rebuttal state set. According to the correspondence of the same candidate detection fact unit, align, associate, and merge the three types of state data. Aggregate the state data belonging to the same candidate detection fact unit into the same enhanced state item to form an enhanced detection fact matching state set. Each enhanced state item in the enhanced detection fact matching state set is matched and encoded. The state results of each enhanced state item, which reflect the degree of correlation of candidate detection fact units, the degree of conclusion support, and the degree of curve reuse and counter-evidence, are converted into corresponding state values. The matching comparison results are then formed according to the preset state arrangement order to form the matching comparison data corresponding to the candidate detection fact units. The Fellegi-Sunter probability record link calculation is performed on the matching comparison data. For each state value in each candidate detection fact unit, the probability of the current state value appearing under the matching condition and the probability of the current state value appearing under the non-matching condition are determined respectively. The logarithm of the ratio of the probability of the current state value appearing under the matching condition to the probability of the current state value appearing under the non-matching condition is determined as the matching weight corresponding to the state value. The matching weights corresponding to each state value in the same candidate detection fact unit are accumulated to form the comprehensive matching weight corresponding to the candidate detection fact unit. The comprehensive matching weight is subjected to probability transformation processing. The comprehensive matching weight is jointly calculated with the preset prior matching probability to form the probability of the detection fact corresponding to each candidate detection fact unit. The joint calculation is to combine the matching advantage corresponding to the comprehensive matching weight with the prior advantage corresponding to the preset prior matching probability, and convert the combined result into a probability value between zero and one. The probability of the detection fact corresponding to each candidate detection fact unit is registered, merged and summarized. The probability of the detection fact belonging to the same candidate detection fact unit is stored as the probability result of the candidate detection fact unit, forming a set of detection fact probability.

8. A big data monitoring method for construction project quality inspection according to claim 1, characterized in that, The acquisition of the regulatory early warning information specifically includes: Read the probability of each candidate detected fact unit in the probability set of the detected fact, and collect and compare each candidate detected fact unit according to the preset conflict judgment conditions, identify the candidate detected fact unit pairs that cannot be established at the same time, and register them to form a conflict relationship set; Based on the conflict relationship set and the probability set of the detection facts, a global decision on the multi-candidate detection chain is performed on each candidate detection fact unit to form a candidate credible detection chain combination. Perform exclusion confirmation and unique attribution confirmation on candidate trusted detection chain combinations, eliminate candidate detection fact units that do not meet the preset exclusion constraints, and determine the candidate trusted detection chain combinations after exclusion confirmation and unique attribution confirmation as the unique trusted detection chain; Extract the original detection data, detection report, and detection conclusion support status corresponding to each candidate detection fact unit in the unique credible detection chain, perform consistency verification on the original detection data and detection report, check the object correspondence, result correspondence, and time sequence correspondence, and form a consistency verification result; Based on the consistency verification results, the unique trusted detection chain that passes the consistency verification is registered as passed, and the unique trusted detection chain that fails the consistency verification is marked as abnormal. Regulatory warning information is generated according to the attribution information corresponding to the abnormal mark.

9. A big data monitoring system based on construction project quality inspection, comprising the big data monitoring method based on construction project quality inspection as described in any one of claims 1 to 8, characterized in that, include: The regulatory data processing module is used to collect business data on construction project quality inspection and supervision, and to perform standardization and object identification coding to form a standardized regulatory data and object identification set. The fact unit construction module is used to perform reverse candidate space pruning based on standardized regulatory data and to perform cross-source mapping and assembly according to the object identifier set to form a candidate detection fact unit set; The fact state generation module is used to extract the detection fact state based on the candidate detection fact unit set to form a candidate detection fact state set. The conclusion support judgment module is used to perform standardized back-calculation processing on the original test data in the candidate test fact unit set according to the test specifications of the corresponding test items, and form a set of test conclusion support states; The curve rebuttal identification module is used to perform MatrixProfile analysis on the original detection time sequence of the candidate detection fact unit set to form a curve reuse rebuttal state set; The probability matching calculation module is used to merge the candidate detection fact state set, the detection conclusion support state set, and the curve reuse proof state set into the enhanced detection fact matching state set, and to calculate the probability set of the detection fact using the Fellegi-Sunter probability record linking method. The link adjudication and early warning module is used to construct conflict relationships between candidate detection fact units based on the probability set of the detection facts, execute global adjudication of multiple candidate detection chains, complete the rejection confirmation and unique attribution confirmation of the candidate detection fact unit set, obtain a unique trusted detection chain, perform consistency verification on the unique trusted detection chain, and generate regulatory early warning information.

Images