An ai-driven construction drawing compliance review system and method

Abstract

The application discloses an AI-driven construction drawing compliance review system and method, relates to the technical field of construction drawing review, and initializes a construction element database based on input construction drawings and a BIM model, automatically generates a construction sequence diagram by using an AI engine, and synchronously generates simulation scene parameters; a graph construction module automatically loads matched construction process specifications and process standards to construct a specification graph, converts the specification graph into a standardized graph format, and automatically calls the specification graph by using a simulation engine to compare a component behavior and a space state in real time; if there is a construction anomaly in the simulation process, a corresponding process node is automatically marked to form a violation event set. The review system realizes the paradigm change of construction drawing review from "static rule matching" to "dynamic process evolution", and significantly enhances the ability to understand complex construction logic and space behavior and the real-time response capability.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of construction drawing review, in particular to an AI-driven construction drawing compliance review system and method. BACKGROUND

[0002] With the increasing complexity of construction projects, the continuous compression of design cycle, and the improvement of national requirements for engineering quality and safety supervision, the traditional construction drawing compliance review method relying on manual work has been difficult to meet the efficient, accurate and standardized review needs. Manual review not only has the disadvantages of low efficiency, strong subjectivity and easy omission, but also is difficult to adapt to the needs of multi-specialty collaborative review and massive specification text rapid matching. Under this background, the development of artificial intelligence technology provides a new path for construction drawing review.

[0003] The prior art has the following defects:

[0004] Traditional construction drawing review usually relies on manual experience and static rules to compare construction drawings item by item. This method not only depends heavily on professional drawing reviewers, but also is prone to omissions and deviations when dealing with large-scale engineering projects or complex construction scenarios. In particular, when dynamic elements such as construction sequence, component arrangement, and equipment operation path are involved, manual review cannot accurately predict possible timing conflicts, site overlaps, or specification violations in the construction phase, greatly restricting the comprehensiveness and forward-looking nature of the review. In addition, the specification documents are diverse, complex in expression, and cross-stage applicable, making it difficult for humans to achieve dynamic fusion and semantic understanding of multi-dimensional rules, resulting in lag and inconsistency in specification implementation.

[0005] Based on this, the present application proposes an AI-driven construction drawing compliance review system and method, which realizes the paradigm shift of construction drawing review from "static rule matching" to "dynamic process evolution", significantly enhancing the ability to understand complex construction logic and spatial behavior and real-time response capability. SUMMARY

[0006] The purpose of the present application is to provide an AI-driven construction drawing compliance review system and method to solve the problems in the background art.

[0007] To achieve the above purpose, the present application provides the following technical solution: an AI-driven construction drawing compliance review system, comprising an initialization module, a graph construction module, a simulation review module, and a decision output module;

[0008] The initialization module: based on the input construction drawing and BIM model, a construction element database is established, an AI engine is used to automatically generate a construction sequence diagram, and a simulation scene parameter is generated synchronously. The construction element database, construction sequence diagram, and simulation scene parameter are sent to the simulation review module;

[0009] The specification construction module: after automatically loading the matched construction process specification and process standard, a specification graph is constructed, the specification graph is converted into a standardized graph format, and the specification graph is sent to the simulation review module;

[0010] The simulation review module: the specification graph is automatically called by the simulation engine, the behavior of the component and the spatial state are compared in real time, if there is construction anomaly in the simulation process, the corresponding process node is automatically marked to form a violation event set, and the violation event set is sent to the decision output module;

[0011] The decision output module: according to the time point and spatial position of the violation event set, a violation cause path is automatically traced and generated, and a suggestion decision is output.

[0012] In a preferred embodiment, the simulation review module drives the dynamic evolution of the construction component in three-dimensional space-time according to the construction drawing, the BIM model and the construction element database through the simulation engine, generates a component behavior sequence, and each component state in the sequence is represented by a spatial envelope box.

[0013] In the simulation running process, the real-time specification graph matches the behavior of the current component with the rules in the specification graph, determines whether there is a spatial conflict between components through a spatial conflict determination function, and determines whether the construction of the component meets the dependent sequence through a process time sequence constraint determination function.

[0014] In a preferred embodiment, the decision output module extracts component number, violation time point, spatial coordinates, violation type, violation rule number information, and reconstructs an event graph with a topological structure;

[0015] Based on the event graph structure, a time backtracking analysis algorithm is called to backtrace the influence path of each violation event, identify the cause-effect path, and generate a violation score for the cause-effect path through a path scoring function;

[0016] When the violation cause-effect path is established, a rule avoidance scheme is automatically generated according to the rule items in the specification graph and the rule mapping logic engine of the AI engine.

[0017] In a preferred embodiment, the decision output module generates a violation score for the cause-effect path through a path scoring function, and the calculation expression is: Where: φ is the violation score of the violation cause-effect path, α j represents the influence factor of component j in the cause-effect path, Risk(C j ) represents the risk probability of component C j triggering a violation in historical data, and n represents the number of upstream components involved in the cause-effect path before the violation of the component.

[0018] In a preferred embodiment, the decision output module calls a time backtracking algorithm to trace the impact path of each violation event forward, identify its causal path, generate the violation causal path according to the time dimension and construction logic based on the topological sorting and path identification method, and set a certain component C i The violation occurs at time t, and the causal path is represented as: P i = {C i-n ,…,C i-1 ,C i}, t i-n <…<t i , wherein: P i represents the causal path of component C i violation, C i-n represents the earliest upstream process component that causes the current violation, the components satisfy the time dependence of , n represents the number of upstream components involved in the causal path before the component violates, t represents the simulation start time of process i, and t represents the completion time of process j.

[0019] In a preferred embodiment, the simulation review module advances the behavior simulation of each component based on time in dynamic simulation evolution, automatically compares the current state of the behavior node with the space specification and process order atlas rule at each time step, and if a violation is detected, marks the behavior node as a violation process and generates a violation event. The violation event records component ID, timestamp, violation type, involved object, violation rule number and severity level.

[0020] In a preferred embodiment, the atlas construction module loads the construction process specification text and process standard manual of the project type and stage, and performs semantic analysis to extract the elements in the standard, including process name, construction conditions, equipment requirements and space restrictions;

[0021] The extraction results are converted into nodes and edges in the graph structure, where nodes represent construction entities and edges represent process logic, spatial constraints, sequence or dependency rules between entities.

[0022] In a preferred embodiment, the initialization module performs semantic recognition and object analysis on the input two-dimensional construction drawing and three-dimensional BIM model, including component type, size, material, position and connection relationship;

[0023] Using an AI engine, the construction sequence graph is automatically constructed based on the structural dependency relationship between components and the process annotations in the drawing, where nodes represent components and arrows represent construction sequence dependency relationships;

[0024] Based on the construction sequence diagram and the geometric information of the components, scene parameters for construction simulation are automatically generated, including path planning, component approach time window, equipment arrangement area, and stacking position.

[0025] In a preferred embodiment, the operation of the AI engine comprises the following steps:

[0026] Obtain the structural dependency assignment, spatial overlap, and construction process matching degree between components, normalize the structural dependency assignment, spatial overlap, and construction process matching degree, map the value range of the structural dependency assignment, spatial overlap, and construction process matching degree to [0, 1], obtain the normalized values of the structural dependency assignment, spatial overlap, and construction process matching degree, and sum the normalized values to obtain the dependency weight between components.

[0027] An AI-driven construction drawing compliance review method, the review method comprising the following steps:

[0028] The review system establishes a construction element database based on the input construction drawing and BIM model, automatically generates a construction sequence diagram using an AI engine, and synchronously generates simulation scene parameters;

[0029] After automatically loading the matched construction process specification and process standard, a specification atlas is constructed, and the specification atlas is converted into a standardized atlas format;

[0030] Through the simulation engine, the specification atlas is automatically called to compare the component behavior and spatial state in real time, and if there is a construction anomaly during the simulation process, the corresponding process node is automatically marked to form a violation event set;

[0031] After automatically tracing the violation cause path according to the time point and spatial position of the violation event set, a suggestion decision is output.

[0032] In the above technical solution, the present application provides technical effects and advantages:

[0033] The application establishes a construction element database based on the input construction drawing and the BIM model through the initialization module, automatically generates a construction sequence diagram using an AI engine, and synchronously generates simulation scene parameters. The atlas construction module automatically loads and matches construction process specifications and process standards to construct a specification atlas, converts the specification atlas into a standardized atlas format, and automatically calls the specification atlas through a simulation engine to compare the component behavior and spatial state in real time. If there is a construction anomaly during the simulation process, the corresponding process node will be automatically marked to form a violation event set. The decision output module automatically traces the violation cause path based on the time point and spatial position of the violation event set and outputs a suggestion decision. The review system realizes the paradigm shift from "static rule matching" to "dynamic process evolution" in construction drawing review, significantly enhancing the ability to understand complex construction logic and spatial behavior and the real-time response capability. BRIEF DESCRIPTION OF DRAWINGS

[0034] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments described in the present application, and other drawings can also be obtained by those skilled in the art based on these drawings.

[0035] Figure 1 The system architecture diagram of the review system of the present application.

[0036] Figure 2 The mind map of the review system of the present application.

[0037] Figure 3 The method flowchart of the review method of the present application.

[0038] Figure 4 The brain map of the review method of the present application. DETAILED DESCRIPTION

[0039] In order to make the purpose, technical scheme and advantages of the embodiments of the present application more clear, the technical scheme in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0040] Embodiment 1: Please refer to Figures 1-2 and Figure 4 The AI-driven construction drawing compliance review system of the present embodiment includes an initialization module, an atlas construction module, a simulation review module and a decision output module.

[0041] The initialization module: based on the input construction drawing and BIM model, a construction element database is established, and the construction sequence diagram is automatically generated by using the AI engine, and the simulation scene parameters are synchronously generated, and the construction element database, the construction sequence diagram and the simulation scene parameters are sent to the simulation review module;

[0042] The atlas construction module: after automatically loading the matched construction process specification and process standard, a specification atlas is constructed, the specification atlas is converted into a standardized atlas format, and the specification atlas is sent to the simulation review module;

[0043] The simulation review module: the specification atlas is automatically called by the simulation engine, the component behavior and the space state are compared in real time, if there is construction abnormality in the simulation process, the corresponding process node is automatically marked to form a violation event set, and the violation event set is sent to the decision output module;

[0044] The decision output module: after automatically tracing the violation cause path according to the time point and the space position of the violation event set, a suggestion decision is output.

[0045] The initialization module of the application establishes a construction element database based on the input construction drawing and BIM model, generates a construction sequence diagram by using the AI engine, and synchronously generates simulation scene parameters, the atlas construction module automatically loads the matched construction process specification and process standard to construct a specification atlas, converts the specification atlas into a standardized atlas format, the simulation review module automatically calls the specification atlas by the simulation engine, compares the component behavior and the space state in real time, if there is construction abnormality in the simulation process, the corresponding process node is automatically marked to form a violation event set, and the decision output module outputs a suggestion decision after automatically tracing the violation cause path according to the time point and the space position of the violation event set. The review system realizes the paradigm shift of the construction drawing review from “static rule matching” to “dynamic process evolution”, and significantly enhances the ability to understand complex construction logic and space behavior and the real-time response ability.

[0046] Embodiment 2: the initialization module establishes a construction element database based on the input construction drawing and BIM model, generates a construction sequence diagram by using the AI engine, and synchronously generates simulation scene parameters, and the construction element database, the construction sequence diagram and the simulation scene parameters are sent to the simulation review module.

[0047] The initialization module is the starting link of the AI-driven construction drawing compliance review system, and the core target is to automatically convert the input construction drawing and BIM model into a calculable and simulative construction element data structure, and to provide clear construction sequence logic and three-dimensional scene parameter support for the subsequent simulation and review process.

[0048] Firstly, semantic recognition and object analysis are performed on the input two-dimensional construction drawing (such as a CAD drawing) and three-dimensional BIM model (such as a Revit model). Mainly including component type, size, material, position, connection relationship and other attributes.

[0049] Table 1: Construction element database structure (simplified example)

[0050]

[0051] This table 1 is automatically generated by the initialization module, and the component information is derived from BIM element attributes and spatial relationship reasoning. The component state is initially "to be installed" or "installed", which is used to generate the construction sequence logic subsequently.

[0052] The initialization module uses an AI engine to automatically construct a construction sequence diagram based on the structural dependency relationship between components (such as columns before beams, foundations before walls) and the process annotations in the drawing (such as construction section division). The running steps of the AI engine are as follows:

[0053] Get the structural dependency assignment, spatial overlap degree, and construction process matching degree between components, normalize the structural dependency assignment, spatial overlap degree, and construction process matching degree, map the value range of the structural dependency assignment, spatial overlap degree, and construction process matching degree to [0, 1], get the normalized value of the structural dependency assignment, spatial overlap degree, and construction process matching degree, sum the normalized value of the structural dependency assignment, spatial overlap degree, and construction process matching degree to get the dependency weight between components. Assuming that the dependency weight from component i to component j is larger, it indicates that the construction of component j is more dependent on the completion of the construction of component i.

[0054] Structural dependency assignment D ij Reflects the support relationship between components in the mechanical or structural system, such as columns supporting beams and beams supporting plates. Data sources:

[0055] Component connection information in BIM model (such as Revit's structural connection constraints);

[0056] Structural relationship of IFC file (such as IfcRelConnectsStructuralElement) calculation logic: use Boolean mapping + weight scoring mechanism:

[0057]

[0058] Example:

[0059] Column (C001) supports beam (C002) → D C001,C002 = 1.0;

[0060] Beam (C002) indirectly supports the suspended ceiling → D C002,Cx =0.7.

[0061] Structural dependency assignment is used to describe the mechanical dependence relationships and load transfer sequence between components in a structural system. It reflects whether a component depends on another component to achieve its structural stability or determine its installation position. This dependency is a strong constraint that must be given priority when generating construction sequence diagrams.

[0062] For example, in a multi-story frame structure, beam installation must be based on the columns or walls to which they are attached; otherwise, problems such as structural inability to bear load and incorrect hoisting points will arise. If structural dependency assignment is ignored in an AI engine, it is highly likely to generate a construction sequence that violates the principles of mechanics, leading to distorted simulation results and even incorrectly determining that certain "pre-installation" operations are reasonable. Therefore, structural dependency assignment is the most critical foundational constraint in the construction sequence diagram, determining the logical order of load-bearing for components.

[0063] Furthermore, this parameter can also be used to dynamically assess the impact of construction schedule adjustments caused by the absence of certain components or design changes. For example, if the construction of a critical column component is delayed, the system can automatically determine its scope of impact based on the dependency weight diagram and adjust the installation time of subsequent dependent components.

[0064] Spatial overlap S ij The degree of spatial overlap between component i and component j is quantified, reflecting the potential for construction conflicts and the closeness of their dependencies. Data source:

[0065] The three-dimensional coordinate bounding box of the component;

[0066] When components have spatial overlap: In the formula, V overlap V represents the overlapping volume of two components. i V j Example of bounding box volume representing components i and j:

[0067] Beam (C002) and slab (C003) are laid together in close contact → S = 1.0;

[0068] The distance between the column (C001) and the far wall (C005) is relatively large → S = 0.2.

[0069] Spatial overlap describes the proximity or potential interference between components in three-dimensional space, and it has a critical impact on construction accessibility, safety, and construction path planning. Even if components are not structurally dependent, if they are spatially close or overlap, problems such as installation conflicts, operational interference, or overlapping construction equipment paths may occur during construction.

[0070] If the AI engine does not consider spatial overlap when generating construction sequence diagrams or simulating construction processes, it may arrange multiple adjacent components to be constructed simultaneously, causing serious congestion in the construction site and even safety hazards (such as overlapping hoisting paths). Therefore, spatial overlap is a core parameter that affects the rationality of construction "operation sequences" and is widely used to determine whether there are spatial overlap installation conflicts, predict construction interference caused by spatial proximity between components, optimize construction paths and component stacking logic (such as the stacking order of component loading and unloading areas), and, in addition, spatial overlap can also be used for construction site resource scheduling and equipment deployment, such as determining whether multiple hoisting operations can be performed simultaneously to avoid resource conflicts.

[0071] Construction process matching degree represents the degree of fit between the construction sequence of components and the standard process,

[0072] Establish standard construction sequence, as shown in Table 2:

[0073] Table 2: Rule mapping score table

[0074] Component combination Standard sequence Matching degree Column → beam Yes 1.0 Beam → plate Yes 1.0 Plate → beam No 0.0 Wall → window Yes 0.8

[0075] Data sources:

[0076] Project construction process specification atlas (e.g., hoisting sequence standards);

[0077] Industry knowledge graph, specification documents (such as "Building Construction Manual" and "Concrete Structure Construction Specification").

[0078] Construction process matching degree is an important indicator that reflects whether the construction sequence between components conforms to existing industry construction specifications, technical standards, or enterprise process atlases. Its essence is to structure and calculate empirical engineering knowledge data and embed it into AI simulation models to ensure that the generated construction timing diagrams are not only logically reasonable but also have real operability and engineering practical significance.

[0079] In many cases, structural dependencies and spatial constraints may allow for flexible adjustment of component parallelism or sequential order, but truly reasonable installation procedures also depend on technical details, construction safety standards, and material installation methods. For example: the sticking order of insulation boards must follow the principle of "from bottom to top, from edge to center"; the template and reinforcement of the elevator shaft must be completed before the internal mechanical and electrical pre-burying of the shaft wall; pipeline installation requires main trunk first, branch second, gravity first, pressure second, large diameter first, small diameter second. These industry knowledge cannot be directly derived from geometric or structural models, but can be modeled or mapped by knowledge graph through construction process matching degree, to realize the knowledge compliance review of construction behavior. The AI engine relies on this parameter to automatically avoid "process errors" in the generated construction process, such as unreasonable installation order and violation of standard operating procedures. It can also be applied to cross-project transfer learning to help AI models generate more practical and reasonable process order based on past project experience.

[0080] Construction sequence diagram direction (example):

[0081]

[0082] In the above construction sequence direction, nodes represent components, and arrows represent construction sequence dependencies.

[0083] Based on the construction sequence diagram and the geometric information of the components, the scene parameters for construction simulation are automatically generated, such as path planning, component arrival time window, equipment layout area, and stacking position, as shown in Table 2:

[0084] Table 3: Simulation scene parameter table (schematic)

[0085] Component number Hoisting equipment type Hoisting start point Hoisting end point Hoisting time window Stacking position Installation duration estimate C001 Tower crane A (5,5,0) (12.5,8.2,4.0) 08:00-10:00 (4,4,0) 25 minutes C002 Tower crane A (5,5,0) (12.5,8.2,0.0) 07:30-08:30 (4,3,0) 15 minutes

[0086] The data in Table 3 is derived from path calculation (based on A* or Dijkstra algorithm) and the combination of component geometric information and site layout rules, to automatically generate scheduling basis for simulation.

[0087] The construction element database, construction sequence diagram, and simulation scene parameters built are sent by the system internal pipeline to the simulation review module for subsequent dynamic simulation. The module data organization and sending are as follows:

[0088]

[0089]

[0090] This pseudo code shows how the module organizes data and sends it to the downstream simulation review module, which provides complete support for subsequent simulation logic.

[0091] Through the above steps, the initialization module not only realizes the full-automatic conversion from drawings and BIM models to data structures and simulation scenes, but also avoids misjudgment or omissions under human intervention by calculating construction sequence logic and dynamic parameters through AI, providing accurate, structured, and evolvable input conditions for the simulation review module. This process lays a solid foundation for subsequent violation identification and path backtracking, ensuring high precision and high real-time performance of the review results.

[0092] The atlas construction module automatically loads the matched construction process specifications and process standards to construct the specification atlas, converts the specification atlas into a standardized atlas format, and sends the specification atlas to the simulation review module.

[0093] The atlas construction module automatically loads the construction process specification text and process standard manual that conforms to the project type and stage, performs semantic analysis, and extracts key elements (such as process name, construction conditions, equipment requirements, space restrictions, etc.) and their logical relationships in the standard. This process can use language models such as BERT or RoBERTa combined with dependency syntax trees for entity and relationship extraction.

[0094] The extraction results are converted into nodes and edges in a graph structure. Nodes represent construction entities (such as components, equipment, personnel actions) or abstract specification concepts, and edges represent process logic, spatial constraints, sequence, or dependency rules between entities. The graph structure is extensible and supports multi-dimensional nesting.

[0095] The atlas rules can be abstracted as a triple representation: R = (E i ,R ij ,E j ), where: E i ,E j represents two construction entity nodes (such as components, processes, operating conditions, etc.), R ij represents the process relationship between entities, which can be "preposition", "space in the middle", "need temporary protection", "equipment sharing", etc., and R represents a rule relationship that can form edges in the atlas and be used to regulate simulation behavior.

[0096] To interface with the simulation review module, the constructed atlas is uniformly converted into a standard atlas format (such as RDF / OWL representation or Neo4j JSON graph structure) and annotated with node weights, rule priority, and other metadata.

[0097] The specification atlas is ultimately sent to the simulation review module through an interface, which calls subgraphs from the atlas according to the simulation scene and completes process compliance comparison and timing logic verification in real time. The following is a simplified pseudo-code for atlas construction (using Neo4j graph database):

[0098] frompy2neoimportGraph,Node,Relationship

[0099] #Connection graph database

[0100] graph=Graph("bolt: / / localhost:7687",auth=("neo4j","password"))

[0101] # Constructing Standard Entity Nodes

[0102] n1 = Node("Process", name = "Component hoisting")

[0103] n2 = Node("Constraints", name = "Lifting radius < 30m")

[0104] n3 = Node("Protective Measures", name = "Edge Fence")

[0105] #Add process specification side

[0106] r1 = Relationship(n1,"must satisfy",n2)

[0107] r2 = Relationship(n1,"Required",n3)

[0108] # Write to graph database

[0109] graph.create(n1|n2|n3|r1|r2)

[0110] The code above indicates that the component hoisting process in the diagram must meet the requirement that the hoisting radius is less than 30 meters and that edge guardrails must be installed. This format can be directly read by the AI ​​compliance engine and used to determine whether the hoisting path is legal and whether the protection is in place during simulation.

[0111] The following table shows a schematic diagram of the graph format (standard graph node / edge attributes):

[0112] Table 4: Standard Graph Nodes

[0113] Node ID Type Name Attribute description N001 Process Component hoisting Corresponding construction task node N002 Restriction condition Lifting radius < 30 m Space and equipment constraints N003 Protective measures Edge guardrail Safety measure specification requirements

[0114] Table 5: Edge Attributes

[0115] Edge ID Start node End node Relationship type Weight R001 N001 N002 Must be met 0.9 R002 N001 N003 Must be set 0.95

[0116] In the standard format of the graphs in Tables 4 and 5, the relationship type of the edges will be used to dynamically trigger rule determination in the simulation; the node attributes will be dynamically bound to the construction element database to achieve accurate mapping and comparison.

[0117] The atlas construction module realizes the automatic conversion process from unstructured specifications to computable atlases through language analysis, graph structure modeling, multi-dimensional rule scoring, and standard format packaging. The result is to provide a knowledge-driven model with context logic, rule weight, and multi-dimensional association for the subsequent simulation review module, enabling the system to automatically perform "atlas-behavior" comparison without human intervention, and complete the whole process of dynamic compliance judgment.

[0118] The simulation review module automatically calls the specification atlas through the simulation engine to compare the behavior of the components and the state of the space in real time. If there is an abnormal construction during the simulation process, it will automatically mark the corresponding process node to form a violation event set, and the violation event set is sent to the decision output module.

[0119] The simulation review module is the core component of the AI-driven construction drawing compliance review system, and is responsible for real-time comparison between the construction simulation process and the construction process specification atlas, thereby accurately identifying abnormal or illegal behavior in the construction process. Its main role is to compare the construction behavior simulation and standardized specifications to achieve "automatic supervision" of the construction process, especially in terms of spatial arrangement, construction sequence, safety specifications, and other aspects to provide real-time review support.

[0120] First, in the simulation initialization phase, the system generates the behavior sequence B_{t} of the components, i.e., the spatial state and behavior action of the components at each time step, by driving the dynamic evolution of the construction components in three-dimensional space-time through the simulation engine (such as Unity3D or AnyLogic) based on the construction drawing, BIM model, and construction element database. Each component state is represented by a spatial envelope box G_{i}(t).

[0121] During the simulation process, the system matches the behavior of the current component with the rules in the atlas generated by the atlas construction module in real time. Taking spatial conflict as an example, the system compares based on the following spatial conflict determination function: Where G i (t) and G j (t) are the geometric space envelopes of components i and j at time t, Vol(*) represents the volume calculation function (such as no more than 5% volume intersection), and δ is the conflict threshold. When C ij (t) = 1, the system determines that there is an illegal spatial conflict, such as overlapping component stacking, hoisting path insertion, etc.

[0122] Next, the simulation review module further verifies whether the time logic of the construction process meets the process dependency relationship. This process uses the following time sequence constraint determination formula:

[0123] Here, Denote the simulation start time of process i, Denote the completion time of process j. When F ij (t)=0, it indicates that the non-compliance phenomenon of "the process constructed first has not been completed while the subsequent process has started" is detected. For example, hoisting the upper component before the concrete has hardened, which is a typical "construction sequence violation".

[0124] In the dynamic simulation evolution, the simulation engine advances the behavior simulation of each component based on time. The system automatically compares the current state with the atlas rules such as spatial specifications and process sequences at each time step. Once a violation is detected, the behavior node is marked as a violation process, and a violation event is generated. This event set records detailed component ID, timestamp, violation type, involved object, violation rule number and severity level. For example, if component L-L3201 has an unacceptable spatial intersection with hoisting equipment D01 at 10:13:21 on May 8, 2025, the system will record the following information:

[0125] {

[0126] "component_id":"Beam-L3201",

[0127] "timestamp":"2025-05-08T10:13:21",

[0128] [[ID= nineteen]]"violation_type":"Spatial conflict",

[0129] "conflict_with":"Hoisting equipment-D01",

[0130] "location":{"x":125.5,"y":34.2,"z":9.8},

[0131] "rule_violated":"Hoisting radius is less than the equipment turning radius",

[0132] "severity_level":"Medium",

[0133] "related_rule_id":"R203-A"

[0134] }

[0135] Associate the above events with relevant components and time nodes, write them into the violation event set uniformly, and finally send this clustered data structure to the decision output module.

[0136] In this way, the simulation review module realizes a complete closed loop from component behavior modeling, dynamic calling of specification atlas, to simulation comparison and violation marking. It significantly solves the pain points of relying on manual comparison of construction drawings and ignoring the relationship between construction timing and spatial evolution in the past, and realizes the upgrade of "dynamic, real-time, and structured" construction drawing review, providing a solid foundation for digital construction management.

[0137] The decision output module automatically generates a violation causal path based on the time point and spatial location of the violation event set, and outputs a recommended decision.

[0138] The decision output module is the terminal processing core of the AI-driven construction drawing compliance review system. Its function is to intelligently analyze and reason the violation event set generated by the simulation review module, automatically identify the cause of the violation, and propose executable optimization suggestions or adjustment strategies. The entire process does not require human intervention and is completely driven by internal algorithms, ensuring rapid response and objective consistency.

[0139] First, the received violation event set is parsed to extract component number, violation time point, spatial coordinates, violation type, violation rule number, and other information, and reconstructed into an event atlas with a topological structure. For example: nodes represent components and their process status; edges represent the dependency relationship between components, spatial proximity relationship, or concurrent relationship within the same time period; each edge is labeled with a causal relationship tag, such as "construction sequence", "spatial overlap", "path intersection". This process can be illustrated by the following pseudo code:

[0140] for event in violation_events:

[0141] G.add_node(event.component_id,timestamp=event.timestamp,type=event.violation_type)

[0142] For related in event.related_components:

[0143] G.add_edge(event.component_id,related,relation='conflict')

[0144] This step establishes the event atlas, enabling subsequent analysis to reason the cause of the violation based on the time causal chain.

[0145] On the basis of the event graph structure, the system calls the time backtracking analysis algorithm, and identifies the cause-effect path of each violation event by backtracking the influence path forward. This process is based on the topological sorting and key path identification method, and generates the violation cause-effect path according to the time dimension and construction logic recursion. Assuming that a component C i violates at time t, its cause-effect path can be expressed as:

[0146] P i ={C i-n ,...,C i-1 ,C i},t i-n <…<t i , in the formula, P i represents the cause-effect path of the component C i violation, C i-n represents the earliest upstream process component that causes the current violation, the components satisfy the time dependence of , n represents the number of upstream components involved in the cause-effect path before the component violates, and the system generates a violation score for the cause-effect path through a path scoring function, which is defined as follows:

[0147] In the formula, φ is the violation score of the cause-effect path, α j represents the influence factor of the component C j in the cause-effect path (generated according to the process key degree, construction position, etc.), and Risk(C j ) represents the risk probability of the component C j triggering a violation in historical data. The higher the score, the more likely it is that the corresponding component in the path is the main cause of the violation.

[0148] α j represents the influence factor of the component in the cause-effect path (generated according to the process key degree, construction position, etc.), in this application, the number of process dependencies (i.e., the number of all processes that need to use the component) and the number of component accidents (i.e., the number of accidents of the same component in the historical construction process) of the component in the cause-effect path are obtained, and the normalized process dependency number and the normalized component accident number are summed to obtain the influence factor. The larger the influence factor, the more important the component is in the current cause-effect path.

[0149] The historical risk probability of the component is used to characterize the frequency and severity of the component triggering a violation in historical data, and is an important basis for the path scoring and priority backtracking of the decision output module. The calculation formula is as follows: Parameter definition: Risk(C j ) represents the risk probability of the component C j triggering a violation in historical data, Vij Representative component C j The number of occurrences of the i-th type of violation in historical cases (such as violation of time sequence, spatial conflict, insufficient safety clearance, etc.), T j Representative component C j The total number of times included in the historical case statistics, i.e., the total number of times simulated, is used for normalization to quantify the historical violation of the component as a risk probability indicator, facilitating the system to assess the violation tendency of the component through statistical methods. In causal tracing, this probability is multiplied by the impact factor to obtain the violation contribution value of the component in the path, thereby promoting the system to lock the main violation source.

[0150] When the violation causal path is established, the system automatically generates avoidance schemes according to the rule items in the specification atlas, combined with the rule mapping logic engine of the AI engine. For example, if the violation is caused by path intersection, the system will suggest adjusting the construction sequence or replacing the construction path; if the violation is caused by spatial overlap, the system will automatically suggest adjusting the component stacking point or optimizing the hoisting timing; if the violation is caused by specification provisions (such as insufficient clearance), the system will propose suggestions for rearranging the site layout.

[0151] Finally, the system outputs all suggestions to the BIM model platform and highlights the violation nodes in the three-dimensional view and visualizes the path. Each suggestion is bound to a time point, component number, and optimization target, facilitating subsequent system re-simulation or automatic modification of the construction plan. For example, construction personnel can see the red-marked violation component on the BIM interface, and after clicking, a prompt of "suggest adjusting hoisting time to avoid path intersection" will pop up, enhancing intuitiveness and execution efficiency.

[0152] In summary, the decision output module realizes the complete chain from data structure analysis, causal tracing, automatic optimization generation to three-dimensional visualization output, solving the defects of traditional compliance review such as vague violation location, lack of systematic adjustment suggestions, and reliance on manual experience, greatly improving the intelligent level and implementation efficiency of the construction plan.

[0153] Embodiment 3: Please refer to Figure 3 As shown in the figure, the AI-driven construction drawing compliance review method of the present embodiment includes the following steps:

[0154] Based on the input construction drawing and BIM model, various components, materials, and their spatial positions and associated relationships are automatically identified, and a preliminary construction element database is established. On this basis, the AI engine uses the construction logic reasoning model to automatically generate a construction sequence diagram, including component installation sequence, process dependency relationship, key path labeling, etc., and simultaneously generates simulation scene parameters such as material transportation path, stockyard planning, and equipment hoisting area, laying a data foundation for simulation modeling.

[0155] According to the project type, construction phase and location, the matching construction process specification and process standard are automatically loaded, including component stacking density, safety distance, construction clearance, operation time window and other rules, and are converted into a standardized graph format that can be parsed by a rule engine. This rule graph not only has a structure level, but also can establish a semantic mapping relationship with the space area and construction equipment in the construction drawing, forming a constraint boundary and verification logic.

[0156] The simulation engine is started to drive the whole process of component approach, handling, hoisting, installation, etc. on the time axis, and automatically calls the specification graph at each time to compare the component behavior and space state in real time. During the simulation process, if there is a conflict between the component hoisting path and other components, the stacking position exceeds the safety distance, or the construction sequence violates the logical dependence, etc., the corresponding process node will be automatically marked to form a set of violation events.

[0157] According to the time point and space position of the violation event, the violation cause path is automatically traced back to clearly indicate whether it is due to improper construction sequence setting, path planning overlap, or specification constraint not met, and output the suggested modified process or arrangement parameters. Through this backtracking mechanism, not only the problem is pointed out, but also the scheme optimization is assisted.

[0158] All review results are output in a structured form, and the system automatically highlights the violation position and posts it back to the original BIM model and construction drawing view, while generating a comprehensive review report containing construction node information, violation rule source, risk level, etc.

[0159] This method realizes the whole process review process from drawings to dynamic construction logic to specification comparison, and the data at each stage is closely related to the next stage, ensuring the integrity and accuracy of the review logic. The whole process does not require human intervention, completely relying on the modeling, simulation and comparison ability of the AI engine, greatly improving the efficiency and coverage depth of the construction drawing compliance review.

[0160] In the description of the present specification, the description referring to the terms "one embodiment", "example", "specific example" and the like means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In the present specification, the illustrative description of the above terms does not necessarily refer to the same embodiment or example. Moreover, the described specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0161] The preferred embodiments of the application disclosed above are only to facilitate the elucidation of the application. The preferred embodiments do not describe all the details of the application and limit the application to the specific embodiments. Obviously, many modifications and variations can be made in light of the teachings above. The description is chosen and described in order to provide the best illustration of the application and its practical application to those skilled in the art and to enable those skilled in the art to best utilize the application. The application is limited only by the claims and their full scope and equivalents.

Claims

1. An AI-driven construction drawing compliance review system, characterized in that: The system comprises an initialization module, a graph construction module, a simulation review module, and a decision output module. The initialization module: based on the input construction drawing and BIM model, a construction element database is established, and the AI engine is used to automatically generate a construction sequence diagram and synchronize the generation of a simulation scene parameter; the construction element database, construction sequence diagram, and simulation scene parameter are sent to the simulation review module; The graph construction module: after automatically loading the matched construction process specification and process standard, a specification graph is constructed, and the specification graph is converted into a standardized graph format; the specification graph is sent to the simulation review module; The simulation review module: through the simulation engine, the specification graph is automatically called, the behavior of the component and the space state are compared in real time, and if there is a construction anomaly in the simulation process, the corresponding process node is automatically marked to form a violation event set; the violation event set is sent to the decision output module; The decision output module: according to the time point and space position of the violation event set, a violation cause and effect path is automatically traced and generated, and a suggestion decision is output; The decision output module extracts the component number, violation time point, space coordinate, violation type, and violation rule number information, and reconstructs them into an event graph with a topological structure; Based on the event graph structure, a time backtracking analysis algorithm is called to backtrace the influence path of each violation event, identify the cause and effect path, and generate a violation score for the cause and effect path through a path scoring function; After the violation cause and effect path is established, a rule item in the specification graph is automatically generated according to the rule mapping logic engine of the AI engine to generate an avoidance scheme; The decision output module generates a violation score for the causal path through a path scoring function, with the calculation expression being: wherein: is the violation score of the violation causal path, represents an impact factor of component j in the causal path, represents a component a risk probability of triggering a violation in historical data, and n represents a number of upstream components involved in the causal path of the component before the component violates.

2. The AI-driven construction drawing compliance review system of claim 1, wherein: The simulation review module drives the dynamic evolution of the construction component in three-dimensional space-time through the simulation engine based on the construction drawing, BIM model, and construction element database, generates a component behavior sequence, and represents each component state with a space envelope box. In the simulation running process, the real-time specification graph matches the behavior of the current component with the rules in the specification graph, determines whether there is a space conflict between components through a space conflict determination function, and determines whether the construction of the component meets the dependent sequence through a process time sequence constraint determination function.

3. The AI-driven construction drawing compliance review system of claim 2, wherein: The decision output module calls a time backtracking algorithm, backtracks the influence path of each violation event, identifies the cause-effect path, generates the violation cause-effect path according to the time dimension and construction logic recursion based on the topological sorting and path identification method, and sets a component In time A violation occurs, and the cause-effect path is represented as: , wherein: represents the component of the violation cause-effect path, represents the earliest upstream process component that causes the current violation, the components satisfy the time dependence of , n represents the number of upstream components involved in the cause-effect path before the component occurs, and represents the simulation start time of process i, represents the completion time of process j.

4. The AI-driven construction drawing compliance review system of claim 3, wherein: The simulation review module simulates the behavior of each component based on time in the dynamic simulation evolution, automatically compares the current state of the behavior node with the space specification and process sequence graph rules at each time step, marks the behavior node as a violation process if a violation is detected, and generates a violation event; the violation event records the component ID, timestamp, violation type, involved object, violation rule number, and severity level.

5. The AI-driven construction drawing compliance review system of claim 4, wherein: The graph construction module loads the construction process specification text and process standard manual of the project type and stage, and performs semantic analysis to extract the elements in the standard, including process name, construction condition, equipment requirement, and space limitation; The extraction results are converted into nodes and edges in a graph structure, where the nodes represent construction entities and the edges represent the process logic, space constraints, sequence, or dependency rules between entities.

6. The AI-driven construction drawing compliance review system of claim 5, wherein: The initialization module performs semantic recognition and object analysis on the input two-dimensional construction drawing and three-dimensional BIM model, including component type, size, material, position and connection relationship; Using the AI engine, a construction sequence diagram is automatically constructed based on the structural dependency relationship between components and the process annotation in the drawing, where nodes represent components and arrows represent construction dependency relationship; Based on the construction sequence diagram and the geometric information of the components, scene parameters for construction simulation are automatically generated, including path planning, component arrival time window, equipment arrangement area and stacking position.

7. The AI-driven construction drawing compliance review system of claim 6, wherein: The operation of the AI engine includes the following steps: Obtain the structural dependency assignment, spatial overlap degree and construction process matching degree between components, normalize the structural dependency assignment, spatial overlap degree and construction process matching degree, map the value range of the structural dependency assignment, spatial overlap degree and construction process matching degree to [0, 1], obtain the normalized value of the structural dependency assignment, the normalized value of the spatial overlap degree and the normalized value of the construction process matching degree, and sum the normalized value of the structural dependency assignment, the normalized value of the spatial overlap degree and the normalized value of the construction process matching degree to obtain the dependency weight between components.

8. An AI-driven construction drawing compliance review method, implemented by the review system of any one of claims 1-7, characterized in that: The review method includes the following steps: The review system establishes a construction element database based on the input construction drawing and BIM model, automatically generates a construction sequence diagram using an AI engine, and synchronously generates simulation scene parameters; After automatically loading the matched construction process specification and process standard, a specification map is constructed, and the specification map is converted into a standardized map format; Through the simulation engine, the specification map is automatically called to compare the component behavior and spatial state in real time, and if there is a construction exception in the simulation process, the corresponding process node is automatically marked to form a violation event set; According to the time point and spatial position of the violation event set, a violation cause path is automatically traced and generated, and a suggestion decision is output.

Images