Constructional engineering compliance examination method and system based on Ai decision tracing

Abstract

The invention discloses a building engineering compliance examination method and system based on Ai decision tracing, and relates to the technical field of data processing. Standard data are determined, building drawing samples are labeled according to the standard data, and the standard data and the labeled building drawing samples are stored to serve as a labeling rule base; setting a matching ring in the labeling rule base; and dividing the standard data and the labeled architectural drawing sample into a plurality of data segments, storing the data segments in a matching sub-unit, when the matching sub-unit is abnormal, destroying the data segments in the matching sub-unit, and performing data segment recovery on the abnormal matching sub-unit based on the supplementary interval. According to the method, the safety and the completeness of the data can be guaranteed through the matchers, when the data is abnormal, the adjacent matchers can help the abnormal matchers to be quickly recovered, the response is quick, loss can be stopped in time, more data is prevented from being influenced, and the method has better data storage stability.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and particularly to a building engineering compliance review method and system based on AI decision traceability. Background Art

[0002] In the field of building engineering, the traditional method for reviewing the compliance of construction drawings mainly relies on manual review, but this method has many drawbacks. On the one hand, the manual review process is cumbersome and inefficient, requiring a large amount of human and time costs, and it is difficult to meet the requirements of the fast-paced and high-efficiency advancement of modern building engineering projects. On the other hand, manual review is extremely vulnerable to subjective factors. Different reviewers have differences in the understanding and implementation of specification standards, which directly leads to the omission or error of the review results, and thus cannot effectively meet the increasingly complex design requirements and stricter specification standards of current building engineering.

[0003] With the booming development of artificial intelligence (AI) technology, its application in the field of engineering compliance review has gradually emerged. Training an AI model to automatically identify violations in construction drawings has become a new way to improve the review efficiency and accuracy. However, in the traditional building engineering compliance review method, the security and integrity guarantee mechanism at the data level is relatively weak. Once the data is tampered with, due to the lack of an effective monitoring and warning mechanism, the system is difficult to quickly detect and respond, which will affect more associated data and has no good data protection effect. Summary of the Invention

[0004] The purpose of the present invention is to provide a building engineering compliance review method and system based on AI decision traceability to solve the deficiencies in the background art.

[0005] To achieve the above purpose, the present invention provides the following technical solutions: A building engineering compliance review method based on AI decision traceability, including the following steps: Determine the specification data, label the building drawing samples according to the specification data, and store the specification data and the labeled building drawing samples as a labeling rule library; Set a matching loop in the labeling rule library, where the matching loop consists of two sub-loops, both of which are composed of multiple matching elements connected together, and open intervals and supplementary intervals are set for the corresponding matching elements; Divide the specification data and the labeled building drawing samples into multiple data segments, store the data segments in the matching elements, and when an abnormality occurs in the matching element, destroy the data segments in the matching element, and recover the data segments of the abnormal matching element based on the supplementary interval; Based on the open interval, the deep learning model is trained with the specification data of the annotation rule library and the annotated architectural drawing samples to obtain a trained violation recognition model, the architectural drawing data is determined, and the architectural drawing data is input into the trained violation recognition model to output the violation information, wherein the violation information includes the violation components and the corresponding violation degree; The violation information is bound to the corresponding building drawing data as an analysis log, and the analysis log is stored.

[0006] In a preferred embodiment, the step of storing the specification data and the annotated architectural drawing samples as an annotation rule library includes: A plurality of architectural drawings are obtained as architectural drawing samples, and component information in the architectural drawing samples is read, wherein the component information includes geometric properties and annotation information of the components.

[0007] Mark the component information in the architectural drawing samples according to the specification data; The specification data and the annotated architectural drawing samples are stored to obtain an annotation rule library.

[0008] In a preferred embodiment, the step of setting a matching ring in the annotation rule library includes: Two sub-rings are configured in the annotation rule base, and both sub-rings are composed of multiple identical matching sub-rings connected in a chain-like manner. Divide the matching sub-interval into multiple parallel sub-intervals, connect the multiple sub-intervals to each other, set corresponding matching codes for the multiple sub-intervals respectively, and bind the matching codes to the corresponding sub-intervals; Connect the two sub-chains to the matching sub-chains one-to-one, and use the matching codes corresponding to the matching sub-chains as the connection pairing conditions; A subinterval matching the preset data in the middle part of the subinterval is selected as an open interval, and the subintervals on both sides of the open interval are respectively made to correspond to the supplementary intervals of the adjacent subintervals.

[0009] In a preferred embodiment, the step of recovering the data segment of the abnormal matching sub-item based on the supplementary interval includes: Divide the specification data and the annotated architectural drawing samples into a number of data segments equal to the number of matching sub-segments; The data segments are matched one by one with the matching sub-elements, and the data segments are randomly divided into different data volumes according to the number of open intervals in the corresponding matching sub-elements to obtain multiple sub-data blocks, and the multiple sub-data blocks are correspondingly stored in the open intervals of the matching sub-elements; The data segment is copied simultaneously and divided into two parts with different data amounts to obtain two parts of supplementary data, and the supplementary data are respectively stored in supplementary intervals in adjacent matching sub-parts; Sort the matching codes of the matching sub-elements in descending order according to the data volume in the corresponding sub-intervals, and use the matching codes of the matching sub-elements sorted in descending order of the two sub-chains as the pairing conditions; Regard the matching sub-elements with changed matching codes as abnormal matching sub-elements, destroy the data segments in the open intervals of the abnormal matching sub-elements, and restore the data segments of the abnormal matching sub-elements based on the supplementary intervals in the adjacent matching sub-elements.

[0010] In a preferred embodiment, the step of storing the supplementary data in the supplementary intervals of adjacent matching sub-elements respectively includes: Divide the supplementary data into different data volumes according to the number of supplementary intervals in the adjacent matching sub-elements respectively; Store the divided supplementary data in the corresponding supplementary intervals respectively.

[0011] In a preferred embodiment, the step of inputting the building drawing data into the trained violation recognition model to output the violation information includes: Obtain data segments through the open intervals, and obtain the standard data of the annotation rule library and the annotated building drawing samples through the data segments in the entire matching loop; Train the deep learning model with the standard data of the annotation rule library and the annotated building drawing samples to obtain a violation recognition model; Input the building drawing data that needs to be analyzed currently into the violation recognition model to output the violation components and the corresponding violation degrees.

[0012] In a preferred embodiment, the step of binding the violation information with the corresponding building drawing data as an analysis log and storing the analysis log includes: Obtain the analysis time of the building drawing data; Bind and store the violation information and the corresponding building drawing data according to the analysis time as an analysis log.

[0013] The present invention also provides a building engineering compliance review system based on AI decision tracing, including: A storage module, used to determine the standard data, annotate the building drawing samples according to the standard data, and store the standard data and the annotated building drawing samples as an annotation rule library; A setting module, connected to the storage module, used to set a matching loop in the annotation rule library, wherein the matching loop consists of two sub-loops, both of which are composed of multiple matching sub-elements connected, and open intervals and supplementary intervals are set for the corresponding matching sub-elements; A data processing module, connected to the setting module, is used to divide the specification data and the annotated building drawing samples into multiple data segments, store the data segments in the matching sub, destroy the data segments in the matching sub when there is an abnormality in the matching sub, and recover the data segments of the abnormal matching sub based on the supplementary interval; An analysis module, connected to the data processing module, is used to train a deep learning model based on the specification data of the annotation rule library and the annotated building drawing samples in the open interval to obtain a trained violation recognition model, determine the building drawing data, and input the building drawing data into the trained violation recognition model to output the violation information. The violation information includes the violated components and the corresponding violation degrees; A recording module, connected to the analysis module, is used to bind the violation information with the corresponding building drawing data as an analysis log and store the analysis log.

[0014] In the above technical solution, the technical effects and advantages provided by the present invention are as follows: The present invention can ensure the security and integrity of data through the matching sub. When an abnormality occurs in the data, the adjacent matching sub can help the abnormal matching sub to recover quickly, with a fast response and the ability to stop losses in time, avoiding affecting more data, and having good data storage stability. Description of the Drawings

[0015] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required in the embodiments. Obviously, the drawings in the following description are only some embodiments recorded in the present invention. For those of ordinary skill in the art, other drawings can also be obtained based on these drawings.

[0016] Figure 1 It is the method flow chart of the present invention.

[0017] Figure 2 It is the system block diagram of the present invention. Detailed Embodiments

[0018] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0019] Embodiment 1, please refer to Figure 1 As shown, a building engineering compliance review method based on Ai decision traceability in this embodiment includes the following steps: S1. Determine the specification data, annotate the architectural drawing samples according to the specification data, and store the specification data and the annotated architectural drawing samples as an annotation rule library. S2. Set a matching loop in the annotation rule library. The matching loop consists of two sub-loops, and both sub-loops are composed of multiple matching elements connected. Open intervals and supplementary intervals are set for the corresponding matching elements. S3. Divide the specification data and the annotated architectural drawing samples into multiple data segments, store the data segments in the matching elements. When there is an abnormality in the matching element, destroy the data segments in the matching element, and recover the data segments of the abnormal matching element based on the supplementary interval. S4. Train a deep learning model with the specification data and the annotated architectural drawing samples in the annotation rule library based on the open interval to obtain a trained violation recognition model. Determine the architectural drawing data, input the architectural drawing data into the trained violation recognition model to output violation information. The violation information includes the violated components and the corresponding violation degrees. S5. Bind the violation information with the corresponding architectural drawing data as an analysis log, and store the analysis log.

[0020] In one embodiment, step S1 of storing the specification data and the annotated architectural drawing samples as an annotation rule library includes: S11. Obtain multiple architectural drawings as architectural drawing samples, and read the component information in the architectural drawing samples. The component information includes the geometric attributes and annotation information of the components.

[0021] S12. Annotate the component information in the architectural drawing samples according to the specification data. S13. Store the specification data and the annotated architectural drawing samples to obtain an annotation rule library.

[0022] As described in the above steps S11 - S13, obtain multiple architectural drawings (CAD drawings) as architectural drawing samples, read the component information in the architectural drawing samples, read the drawing files through the CAD software interface, use the layer information to distinguish different components (such as beams, columns, walls), extract the geometric attributes (dimensions, positions) and annotation information (material grades, load parameters) of each component, and then annotate the component information in the architectural drawing samples according to the specification data. Here, what is annotated is whether the component information meets the specification data. For example, annotate that the clear width of the stairwell should be ≥ 1.1m. If the actual value is 1.05m, it is marked as a violation. During the annotation process, distinguish between serious violations (such as structural safety) and minor violations (such as non-critical dimensional deviations). The specification data here is the design specification version in architecture. Store the specification data and the annotated architectural drawing samples to obtain an annotation rule library, which has good data preparation work.

[0023] In one embodiment, step S2 of setting a matching loop in the annotation rule library includes: S21. Configure two sub-loops in the annotation rule library. Both sub-loops are formed by connecting the same multiple matching sub-chains in a loop; S22. Divide the matching sub into multiple parallel sub-intervals. The multiple sub-intervals are connected to each other. Corresponding matching codes are set for the multiple sub-intervals respectively, and the matching codes are bound to the corresponding sub-intervals; S23. Connect the matching sub one-to-one through two sub-chains, using the matching code corresponding to the matching sub as the connection pairing condition; S24. Select the sub-interval of the preset data in the middle part of the matching sub as the open interval, and the sub-intervals on both sides of the open interval are respectively corresponding to the supplementary intervals of the adjacent sub-intervals.

[0024] As described in the above steps S21 - S23, the matching loop is set in the annotation rule library. Both sub-loops are formed by connecting the same multiple matching sub-chains in a loop. The matching sub is a virtual machine, which is equivalent to a hand-in-hand loop connection between multiple virtual machines. Since the matching sub is a virtual machine for storage and connection, the matching sub can be divided into multiple parallel sub-intervals. The multiple sub-intervals are connected to each other. Corresponding matching codes are set for the multiple sub-intervals respectively, and the matching codes are bound to the corresponding sub-intervals. Dividing the sub-intervals is for the subsequent planning of the open interval and the supplementary interval, and at the same time, corresponding matching codes can be bound and set. One sub-interval corresponds to one matching code, and the code corresponding to one matching sub is a combination of the matching codes of multiple sub-intervals. The sorting of the matching codes is carried out according to the size of the data volume in the sub-interval, and the sorting of the matching codes is only carried out when storing the standardized data and the annotated building drawing samples later. The current matching loop is in an initial connection state (connecting the matching sub one-to-one through two sub-chains, using the matching code corresponding to the matching sub as the connection pairing condition). Select the sub-interval of the preset data in the middle part of the matching sub as the open interval. The open interval is in an open state to the outside. The sub-intervals on both sides of the open interval are respectively corresponding to the supplementary intervals of the adjacent sub-intervals. The supplementary interval is in a closed state to the outside and is only used for subsequent supplementation of the adjacent matching sub, which can ensure the accuracy of the data.

[0025] In one embodiment, step S3 of recovering the data segment of the abnormal matching sub based on the supplementary interval includes: S31. Divide the standardized data and the annotated building drawing samples into multiple data segments with the same number as the matching sub; S32. One-to-one correspondence is established between the data segments and the matching substrings. The data segments are randomly divided into multiple sub-data blocks with different data volumes according to the number of open intervals in the corresponding matching substrings, and the multiple sub-data blocks are correspondingly stored in the open intervals of the matching substrings; S33. The data segments are simultaneously copied and divided into two parts with different data volumes to obtain two supplementary data, and the supplementary data are respectively stored in the supplementary intervals of adjacent matching substrings; S34. The matching codes of the matching substrings are sorted from large to small according to the data volume in the corresponding sub-intervals, and the matching codes of the matching substrings after sorting from large to small of the two sub-chains are used as the pairing conditions; S35. The matching substrings with changed matching codes are used as abnormal matching substrings, the data segments in the open intervals of the abnormal matching substrings are destroyed, and the data segments of the abnormal matching substrings are restored based on the supplementary intervals in adjacent matching substrings.

[0026] In one embodiment, step S32 of storing the supplementary data in the supplementary intervals of adjacent matching substrings respectively includes: 321. The supplementary data are respectively divided into different data volumes according to the number of supplementary intervals in adjacent matching substrings; 322. The divided supplementary data are respectively stored in the corresponding supplementary intervals.

[0027] As described in the above steps S31 - S35, the specification data and the annotated architectural drawing samples are divided into multiple data segments with the same number as the number of matching subunits. The data segments can be stored one by one in the matching subunits. The data segments are randomly divided with different data volumes according to the number of open intervals in the corresponding matching subunits. The different data volumes represent different orderings of the matching codes, resulting in multiple sub - data blocks. The multiple sub - data blocks are correspondingly stored in the open intervals of the matching subunits. The open intervals are open to the outside and can export data for training a deep - learning model. The data segments are copied simultaneously and divided into two parts with different data volumes to obtain two supplementary data sets. The supplementary data sets are respectively stored in the supplementary intervals of adjacent matching subunits. The supplementary intervals cannot be accessed, and data recovery can only be performed on adjacent matching subunits in the annotation rule library. Specifically: the supplementary data sets are respectively divided with different data volumes according to the number of supplementary intervals in the adjacent matching subunits, and the divided supplementary data sets are respectively stored in the corresponding supplementary intervals. The matching codes of the supplementary intervals in the matching subunits and the matching codes of the open intervals are combined into codes. The matching codes of the matching subunits are sorted from large to small according to the data volume in the corresponding sub - intervals (the sorted matching codes here are the codes). Taking the matching codes of the matching subunits sorted from large to small as the pairing condition for two sub - chains, the two sub - chains can be matched and connected through the matching codes of the matching subunits sorted from large to small. In this way, the matching subunits with changed matching codes are regarded as abnormal matching subunits (a changed matching code means that the matching subunit has been accessed and tampered with by an external network, resulting in a change in the data volume. Even if the change time is short, it will cause a short - term change in the sorting of the matching codes). The connection between the abnormal matching subunit on one sub - chain and the corresponding matching subunit on the other sub - chain can be quickly disconnected, with a fast response. The data segments in the open interval of the abnormal matching subunit are destroyed, which can prevent unsafe data from affecting the data in more matching subunits. Based on the supplementary intervals in the adjacent matching subunits, the data segments of the abnormal matching subunit are restored. After the abnormal data segments are destroyed, the data segments of the abnormal matching subunit can be restored through the adjacent matching subunits. The restored matching subunit is connected again to the corresponding matching subunit in the other sub - chain through the corresponding matching code sorted from large to small, which can ensure the safe storage of data. For example, there are two sub - chains each with four matching subunits. The matching subunits in one sub - chain are respectively denoted as a1, a2, a3, and a4;The matching elements in another sub-chain are respectively denoted as b1, b2, b3, and b4. For example, each matching element is denoted as c1, c2, c3, and c4 due to four sub-intervals. c2 and c3 can be used as open intervals, c1 and c4 as supplementary intervals. c2 and c3 are used to store the first data segments with different data volumes. c1 is used to store the data segment of a partial data volume in one adjacent matching element, and c4 is used to store the data segment of a partial data volume in the other adjacent matching element. a1 and b1 are correspondingly connected through the matching codes of the matching elements sorted from large to small, a2 and b2 are correspondingly connected through the matching codes of the matching elements sorted from large to small, a2 and b2 are correspondingly connected through the matching codes of the matching elements sorted from large to small, a2 and b2 are correspondingly connected through the matching codes of the matching elements sorted from large to small. During the process of storing data, the storage security index can be evaluated. The calculation formula of the storage security index is:; , where, is the storage security index, is the number of times the matching code of the matching element changes, is the frequency of change of the matching code of the matching element, is the number of matching elements, is a constant greater than zero. It should be noted that the larger the value of the storage security index, the greater the potential hidden danger of data storage security. Through the matching elements, the security and integrity of the data can be guaranteed. When the data is abnormal, the adjacent matching elements can help the abnormal matching element to recover quickly, with fast response and timely loss prevention, avoiding affecting more data, having good data storage stability, and ensuring that accurate data can train an accurate model in subsequent training.

[0028] In one embodiment, the step S4 of inputting the building drawing data into the trained violation recognition model and outputting the violation information includes: S41. Obtain the data segment through the open interval, and obtain the standard data of the annotation rule library and the annotated building drawing samples through the data segments in the entire matching loop; S42. Train the deep learning model with the standard data of the annotation rule library and the annotated building drawing samples to obtain the violation recognition model; S43. Input the building drawing data that needs to be analyzed currently into the violation recognition model and output the violation components and the corresponding violation degrees.

[0029] As described in the above steps S41 - S43, the open interval can export data for the training of the model. Here, the deep learning model is a convolutional neural network, which can obtain the standardized data of the annotation rule library and the annotated building drawing samples by matching the data segments in the entire matching loop. The deep learning model is trained with the standardized data of the annotation rule library and the annotated building drawing samples to obtain a violation recognition model. Then, the model can be put into use, and the building drawing data to be analyzed currently is input into the violation recognition model, and the output is the violation components and the corresponding violation degrees as violation information, which can complete the compliance review of the construction project.

[0030] In one embodiment, step S5 of binding the violation information with the corresponding building drawing data as an analysis log and storing the analysis log includes: S51. Obtain the analysis time of the building drawing data; S52. Bind and store the violation information and the corresponding building drawing data according to the analysis time as an analysis log.

[0031] As described in the above steps S51 and S52, since the building drawing data needs to be modified if there are non - compliant situations after analysis, and then the building drawing data is analyzed again until it is compliant. There is an analysis time during the compliance review of the building drawing data. Therefore, for subsequent data traceability and responsibility division of the building drawing data, it is necessary to bind and store the violation information and the corresponding building drawing data according to the analysis time as an analysis log.

[0032] Embodiment 2, please refer to Figure 2 As shown, a construction project compliance review system based on Ai decision - making traceability in this embodiment includes: A storage module, used to determine the standardized data, annotate the building drawing samples according to the standardized data, and store the standardized data and the annotated building drawing samples as an annotation rule library; A setting module, connected to the storage module, used to set a matching loop in the annotation rule library. Among them, the matching loop is composed of two sub - loops, and both sub - loops are composed of multiple matching elements connected. Open intervals and supplementary intervals are set for the corresponding matching elements; A data processing module, connected to the setting module, used to divide the standardized data and the annotated building drawing samples into multiple data segments, store the data segments in the matching elements. When there is an abnormality in the matching element, destroy the data segments in the matching element, and restore the data segments of the abnormal matching element based on the supplementary interval; An analysis module, connected to the data processing module, is used to train a deep learning model based on the specification data in the annotation rule library and the architectural drawing samples after annotation in an open interval to obtain a trained violation recognition model, determine the architectural drawing data, and input the architectural drawing data into the trained violation recognition model to output violation information, where the violation information includes the violated components and the corresponding violation degrees; A recording module, connected to the analysis module, is used to bind the violation information with the corresponding architectural drawing data as an analysis log and store the analysis log.

[0033] It should be noted that the matching sub can ensure the security and integrity of the data. When the data is abnormal, the adjacent matching sub can help the abnormal matching sub to recover quickly, with fast response and timely loss prevention, avoiding affecting more data, and having good data storage stability.

[0034] The above is only the specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present application can easily think of changes or substitutions, which should all be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A building engineering compliance review method based on AI decision traceability, characterized in that, The following steps are involved: Determine the specification data, annotate the architectural drawing samples according to the specification data, and store the specification data and the annotated architectural drawing samples as an annotation rule library; A matching ring is set in the annotation rule library, wherein the matching ring is composed of two sub-rings, both sub-rings are composed of a plurality of matching sub-rings, and an open interval and a supplementary interval are set for the matching sub-rings; Divide the specification data and the annotated architectural drawing samples into multiple data segments, store the data segments in the matching sub-sub ... Based on the open interval, the deep learning model is trained with the specification data of the annotation rule library and the annotated architectural drawing samples to obtain a trained violation recognition model, the architectural drawing data is determined, and the architectural drawing data is input into the trained violation recognition model to output the violation information, wherein the violation information includes the violation components and the corresponding violation degree; The violation information is bound to the corresponding building drawing data as an analysis log, and the analysis log is stored.

2. The method for compliance review of construction projects based on AI decision tracing according to claim 1, wherein: The step of storing the specification data and the annotated architectural drawing samples as an annotation rule library includes: Acquiring a plurality of architectural drawings as architectural drawing samples, reading component information in the architectural drawing samples, wherein the component information includes geometric properties and annotation information of the components; Mark the component information in the architectural drawing samples according to the specification data; The specification data and the annotated architectural drawing samples are stored to obtain an annotation rule library.

3. A method for compliance review of construction projects based on AI decision tracing according to claim 1, characterized in that: The step of setting a matching ring in the annotation rule library includes: Two sub-rings are configured in the annotation rule base, and both sub-rings are composed of multiple identical matching sub-rings connected in a chain-like manner. Divide the matching sub-interval into multiple parallel sub-intervals, connect the multiple sub-intervals to each other, set corresponding matching codes for the multiple sub-intervals respectively, and bind the matching codes to the corresponding sub-intervals; Connect the two sub-chains to the matching sub-chains one-to-one, and use the matching codes corresponding to the matching sub-chains as the connection pairing conditions; A subinterval matching the preset data in the middle part of the subinterval is selected as an open interval, and the subintervals on both sides of the open interval are respectively made to correspond to the supplementary intervals of the adjacent subintervals.

4. A method for compliance review of construction projects based on AI decision tracing according to claim 1, characterized in that: The step of recovering the data segment of the abnormal matching sub-item based on the supplementary interval includes: Divide the specification data and the annotated architectural drawing samples into a number of data segments equal to the number of matching sub-segments; The data segments are matched one by one with the matching sub-elements, and the data segments are randomly divided into different data volumes according to the number of open intervals in the corresponding matching sub-elements to obtain multiple sub-data blocks, and the multiple sub-data blocks are correspondingly stored in the open intervals of the matching sub-elements; The data segment is copied simultaneously and divided into two parts with different data amounts to obtain two parts of supplementary data, and the supplementary data are respectively stored in supplementary intervals in adjacent matching sub-parts; Sort the matching codes of the matching sub-zones from large to small according to the amount of data in the corresponding sub-zones, and use the matching codes of the matching sub-zones after sorting the two sub-chains from large to small as the pairing condition; Take the matching sub with a changed matching code as an abnormal matching sub, destroy the data segments in the open interval of the abnormal matching sub, and restore the data segments of the abnormal matching sub based on the supplementary interval in the adjacent matching sub.

5. The method for compliance review of construction projects based on AI decision traceability according to claim 4, wherein: The step of storing the supplementary data in the supplementary intervals of adjacent matching subs respectively includes: Dividing the supplementary data into different data volumes according to the number of supplementary intervals in the adjacent matching subs respectively; Storing the divided supplementary data in the corresponding supplementary intervals respectively.

6. The method for compliance review of construction projects based on AI decision traceability according to claim 1, wherein: The step of inputting the building drawing data into the trained violation recognition model to output the violation information includes: Obtaining the data segments through the open interval, and obtaining the standard data of the annotation rule library and the annotated building drawing samples through the data segments in the entire matching loop; Training the deep learning model with the standard data of the annotation rule library and the annotated building drawing samples to obtain a violation recognition model; Inputting the building drawing data that needs to be analyzed currently into the violation recognition model to output the violation components and the corresponding violation degrees.

7. The method for compliance review of construction projects based on AI decision tracing according to claim 1, wherein: The step of binding the violation information with the corresponding building drawing data as an analysis log and storing the analysis log includes: Obtaining the analysis time of the building drawing data; Binding and storing the violation information and the corresponding building drawing data according to the analysis time as an analysis log.

8. A building engineering compliance review system based on AI decision traceability, which is used to implement a building engineering compliance review method according to any one of claims 1-7, characterized in that, Including: A storage module, used to determine the standard data, annotate the building drawing samples according to the standard data, and store the standard data and the annotated building drawing samples as an annotation rule library; A setting module, connected to the storage module, used to set a matching loop in the annotation rule library, where the matching loop consists of two sub-loops, both of which are composed of multiple matching subs connected, and an open interval and a supplementary interval are set for the corresponding matching subs; A data processing module, connected to the setting module, used to divide the standard data and the annotated building drawing samples into multiple data segments, store the data segments in the matching subs, and when there is an abnormality in the matching sub, destroy the data segments in the matching sub, and restore the data segments of the abnormal matching sub based on the supplementary interval; An analysis module, connected to the data processing module, used to train the deep learning model with the standard data of the annotation rule library and the annotated building drawing samples based on the open interval to obtain a trained violation recognition model, determine the building drawing data, and input the building drawing data into the trained violation recognition model to output the violation information, where the violation information includes the violation components and the corresponding violation degrees; A recording module, connected to the analysis module, used to bind the violation information with the corresponding building drawing data as an analysis log and store the analysis log.

Images