Building operation and maintenance supervision method and system based on BIM

Abstract

The invention discloses a BIM-based building operation and maintenance supervision method and system, and the method comprises the steps: S1, obtaining building operation and maintenance data, and dividing the building operation and maintenance data into n data partitions according to the layout and structure system of a building; s2, establishing a corresponding data partition index for each data partition, and outputting a prediction index table by using the building component diagram model; s3, determining an initial data partition of the building operation and maintenance data based on the prediction index table, and introducing a random forest model to predict a next collided data partition; and S4, carrying out collision detection on the next collided data partition by utilizing the optimized collision detection algorithm, judging whether the component space bounding boxes are intersected or not, obtaining a collision detection result, screening the data partitions which are possibly collided when collision detection is carried out, then executing the collision detection algorithm in the screened data partitions, and carrying out collision detection on the data partitions which are possibly collided when collision detection is carried out. Therefore, the unnecessary calculation amount is reduced, and the collision detection efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data operation and maintenance supervision, and in particular to a BIM-based building operation and maintenance supervision method and system. Background Art

[0002] Covers a large amount of information throughout the life cycle of a building, including geometric information (such as component shape, size, and spatial position), physical information (such as component weight, density, and strength), and material information (such as material type and model). As the size of a building increases and the time it is used increases, the amount of data continues to expand. For example, a large commercial complex or high-rise office building contains a large number of components, and new data (such as equipment operation data and maintenance records) is continuously generated during the operation of the building, resulting in a sharp increase in the amount of data;

[0003] Among them, collision detection plays an important role in ensuring the safety and functionality of building facilities in building operation and maintenance. However, the expansion of data volume prolongs the collision detection time. This is because when performing collision detection, the system needs to compare the spatial position relationship of various components in the building (such as pipes, lines, structural frames, etc.). The large amount of data leads to an exponential increase in the amount of calculation, which makes it impossible to timely discover and resolve potential collision risks, and increases the risk of building operation and maintenance supervision. Therefore, a BIM-based building operation and maintenance supervision method and system are proposed to solve the above problems. Summary of the invention

[0004] The purpose of the present invention is to solve the shortcomings of the prior art that the amount of data is too large, resulting in an exponential increase in the amount of calculation, which makes it impossible to timely discover and solve potential collision risks, and to propose a building operation and maintenance supervision method and system based on BIM.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] A building operation and maintenance supervision method based on BIM, the steps of the method are as follows:

[0007] S1: Obtain building operation and maintenance data, and divide the building operation and maintenance data into n data partitions according to the layout and structural system of the building;

[0008] S2: Establish a corresponding data partition index for each data partition, and use the building component graph model to output a prediction index table;

[0009] S3: Determine the initial data partition of the building operation and maintenance data based on the prediction index table, and introduce the random forest model to predict the next data partition where the collision will occur;

[0010] S4: Use the optimized collision detection algorithm to perform collision detection on the next data partition where the collision occurs, determine whether the component space bounding boxes intersect, and obtain the collision detection results. Based on the collision detection results, a real-time warning is issued for the detected collision event on the building operation and maintenance supervision interface.

[0011] Preferably, the process of dividing the building operation and maintenance data into multiple data partitions is:

[0012] Obtain a large amount of information data covering the entire life cycle of the building from the real-time building operation and maintenance database, including geometric information, physical information, and material information. Let the acquired building operation and maintenance data be D;

[0013] According to the functional layout and spatial layout of the building, the data set is divided into n data partitions, denoted as D = (P1...P2, .., P n );

[0014] Preferably, the data set is divided into n data partitions by functional area division and spatial layout.

[0015] Preferably, the prediction index table includes a data partition component set, a component quantity, and a set of adjacent data partitions.

[0016] Preferably, the implementation process of outputting the corresponding prediction index table through the building component graph model is:

[0017] Construct a building component diagram model;

[0018] The preferred construction process of the building component diagram model is:

[0019] Extract all the relevant information of components from the building operation and maintenance data, including the geometric information, physical information, spatial location information and the connection relationship between components. Suppose the component set is C = (c1, c2, .., c N ), where N is the number of components;

[0020] Treat each component as a node in a graph model, and the attributes of the component as the attributes of the node;

[0021] Assume that the basic framework of the building component graph model is G = (C, E), where E is the edge set, expressed as E = {e jk │c j , c k}, specific, e jk It consists of two components c j , c k The connection relationship between them is determined, and 1≤j≤N, 1≤k≤N, j≠k;

[0022] Preferably, the output process of the prediction index table is:

[0023] Traverse all nodes in the building component graph model and obtain the component type t of each node;

[0024] Through a data partition Traversing data partitions All nodes of the node are counted, and the component types of all nodes are counted to obtain a component type set T = (t1, t2, .., t N );

[0025] For a data partition P v , let the corresponding graph model node set be C v , traverse the node set C v Nodes in Will be with the node The data partition P where all nodes connected by the edge are located α , added to a set of adjacent data partitions A, which contains a series of data partitions connected by edges;

[0026] The component type set, component quantity, and adjacent partition set are integrated to form a prediction index table, expressed as θ = {N, T, A}.

[0027] Preferably, the step of determining the initial data partition of the building operation and maintenance data based on the prediction index table is:

[0028] Traverse the prediction index table and set the component type set involved in the collision detection request as T C For each partition and its corresponding prediction index table, check whether each component type in the component type set T belongs to T C ;

[0029] Define a matching function F T (P i ):

[0030]

[0031] in, Indicates the first Type of component, when F T (P i )>0, the corresponding data partition P i As the initial data partition, an initial data partition set S = {P i1 , P i2 , .., P in}.

[0032] Preferably, the implementation process of predicting the next data partition where a collision occurs is:

[0033] A random forest model is introduced, and the collision data in the historical operation and maintenance data is used as training data for model training;

[0034] Starting from the initial data partition, for the new data partition, the component type and the components in the adjacent partitions are input into the trained random forest model. Extract the characteristic values ​​of the initial data partition, that is, the component type in the partition and the component type in the adjacent partition, and obtain a series of characteristic data The feature data Input into the trained random forest model to obtain the prediction result Y;

[0035] Output the final collision probability based on the prediction results Among them, the final collision probability prediction result is the average value of all decision tree prediction results. It represents the predicted probability value of the collision between component type t and components in adjacent data partition P. After traversing the collision probability after all initial data partition inputs, a collision threshold σ is set. When , the adjacent data partitions of the initial data partition in the input data that are larger than the collision threshold σ are marked as the next data partitions that collide.

[0036] Preferably, the steps for obtaining the collision detection result are:

[0037] Define a data partition P S ∈S next Partition, data partition P S The component set in C S =(c S1 , c S2 , .., c Sq ), where q represents the data partition P S The number of components within;

[0038] By C S Any two components c Sα , c Sβ , α∈q, β∈q and α≠β, let component c Sα The bounding box is B Sα =(x maxα , x minα ,y maxα ,y minα , z maxα , z minα ), where for component c Sα , its spatial bounding box B Sα Defined by six parameters, x maxα and x minα Respectively represent the minimum and maximum coordinate values ​​of the component in the axial direction; ymaxα and minα Respectively represent the minimum and maximum coordinate values ​​of the component in the axial direction; z maxα and z minα Respectively represent the minimum and maximum coordinate values ​​of the component in the axial direction;

[0039] Component c Sβ The bounding box is B Sβ =(x maxβ , x minβ ,y maxβ ,y minβ , z maxβ , z minβ ), where x maxβ and x minβ Respectively represent the minimum and maximum coordinate values ​​of the component in the axial direction; y maxβ and minβ Respectively represent the minimum and maximum coordinate values ​​of the component in the axial direction; z maxβ and z minαβ Respectively represent the minimum and maximum coordinate values ​​of the component in the axial direction;

[0040] A formula is used to determine whether two spatial bounding boxes intersect, expressed as:

[0041]

[0042] When all conditions are met, the judgment result is that the spatial bounding boxes of the two components intersect, otherwise they do not intersect.

[0043] A BIM-based building operation and maintenance supervision system, including:

[0044] Data partitioning module: obtains building operation and maintenance data and divides the building operation and maintenance data into n data partitions according to the layout and structural system of the building;

[0045] Data index establishment module: establishes corresponding data partition index for each data partition, and outputs a prediction index table using the building component graph model;

[0046] Data partition prediction module: Determine the initial data partition of building operation and maintenance data based on the prediction index table, and introduce the random forest model to predict the next data partition where collision will occur;

[0047] Data detection module: Use the optimized collision detection algorithm to perform collision detection on the next data partition where a collision occurs, determine whether the component space bounding boxes intersect, and obtain the collision detection results. Based on the collision detection results, a real-time warning is issued for the detected collision event on the building operation and maintenance supervision interface.

[0048] The present invention has the following beneficial effects:

[0049] 1. In the present invention, through data partition management and prediction index table, the area where collision may occur can be quickly located, and unnecessary data processing can be reduced. In large buildings, only the data partition related to the specific component type or the initial collision area needs to be paid attention to, and there is no need to conduct a comprehensive detection of the entire building data, which significantly shortens the detection time. At the same time, the random forest model is used to predict the next data partition where collision will occur, further narrowing the detection range; the model is based on the pattern learned from historical data, and can accurately predict the potential collision area, avoiding the detection of all data partitions one by one, improving the detection efficiency, and by optimizing the collision detection algorithm, the spatial bounding box is first preliminarily screened, a large number of non-intersecting components are quickly excluded, and then the components that may collide are accurately detected, avoiding complex geometric calculations for all components, greatly improving the speed of collision detection, and based on the prediction results and the detection results, the building operation and maintenance management personnel can make more scientific and reasonable decisions. When a data partition with a high collision risk is predicted, detailed data conditions can be obtained in advance through the BIM building operation and maintenance data to prevent it. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 A method step diagram of a BIM-based building operation and maintenance supervision method and system proposed by the present invention.

[0051] Figure 2 This is a system block diagram of a BIM-based building operation and maintenance supervision method and system proposed by the present invention. DETAILED DESCRIPTION

[0052] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0053] Embodiment 1

[0054] like Figure 1 to Figure 2 As shown, the present invention proposes a BIM-based building operation and maintenance supervision method and system, including:

[0055] S1: Obtain building operation and maintenance data, and divide the building operation and maintenance data into n data partitions according to the layout and structural system of the building;

[0056] The process of dividing building operation and maintenance data into multiple data partitions is as follows:

[0057] Obtain a large amount of information data covering the entire life cycle of the building from the real-time building operation and maintenance database, including geometric information, physical information and material information. Let the acquired building operation and maintenance data be D;

[0058] Among them, the building operation and maintenance database is the main source of data, which continuously collects and stores various types of data throughout the life cycle of the building, including but not limited to:

[0059] Geometric information: such as the shape, size, spatial position, etc. of building components and the coordinates of each component in three-dimensional space;

[0060] Physical information: physical properties of building materials, such as weight, density, strength, etc. of components;

[0061] Material information: including detailed data of various materials used in the building, such as the type of material (concrete, steel, wood, etc.) and model;

[0062] According to the functional layout and spatial layout of the building, the data set is divided into n data partitions, denoted as D = (P1...P2, .., P n );

[0063] Among them, according to the functional area division: buildings are usually composed of areas with different functions, such as office areas, commercial areas, residential areas, equipment areas, etc. The building components in these functional areas are similar in terms of usage functions and operation and maintenance requirements;

[0064] For example, the office area may contain a large number of office furniture, electrical equipment, and network lines; the equipment area contains large equipment such as air-conditioning units and elevator rooms. Based on the functional characteristics, the building component data belonging to the same functional area is divided into the same data partition;

[0065] Partitioning by spatial layout: Data partitioning is performed based on the spatial structure of the building, for example, by floor. For multi-story buildings, the building components of each floor are relatively independent in the vertical direction, and the usage and operation and maintenance requirements of each floor may be different. Suppose the building has m floors, and each floor corresponds to a data partition P. i (i=1,2,..,m),P i Represents the i-th partition.

[0066] S2: Establish a corresponding data partition index for each data partition, and use the building component graph model to output a prediction index table;

[0067] The prediction index table includes a data partition component set, a component quantity, and an adjacent data partition set;

[0068] The implementation process of outputting the corresponding prediction index table through the building component graph model is as follows:

[0069] Construct a building component diagram model;

[0070] Among them, the construction process of the building component diagram model is:

[0071] Extract all the relevant information of components from the building operation and maintenance data, including the geometric information (such as shape and size), physical information (such as material and weight), spatial position information (3D coordinates) and the connection relationship between components. Suppose the component set is C = (c1, c2, .., c N ), where N is the number of components;

[0072] Treat each component as a node in a graph model, and the attributes of the component as the attributes of the node;

[0073] For example, if the geometric shape of component c1 is G1 and the spatial position is (x1, y1, z1), then the attributes of the node in the graph model can be expressed as c1 = (G1, x1, y1, z1);

[0074] Assume that the basic framework of the building component graph model is G = (C, E), where E is the edge set, expressed as E = {e jk │c j , c k}, specific, e jk It consists of two components c j , c k The connection relationship between them is determined, and 1≤j≤N, 1≤k≤N, j≠k;

[0075] Among them, the connection relationship mainly refers to the physical connection and logical association between building components;

[0076] The output process of the prediction index table is:

[0077] Traverse all nodes in the building component graph model and obtain the component type t of each node;

[0078] Through a data partition Traversing data partitions All nodes of the node are counted, and the component types of all nodes are counted to obtain a component type set T = (t1, t2, .., t N );

[0079] A counter array count[N] is set through the component set. N represents the number of all possible component types in the building. The array is used to record the number of times each component type appears. The initial value of each element in the array is set to 0. Before starting the statistics, the number of occurrences of each component type is considered to be 0. All components in the data partition are traversed. When a component of type t is encountered, i When a component is created, the corresponding counter is increased by 1;

[0080] Specifically, it is represented in the program as count[i]++, where i is the same as the component type t. i The corresponding index,Finally, the component type set T is determined by checking which elements in the counter array have values ​​greater than 0. For each element in the counter array count, if count[j]>0, then the corresponding component type belongs to the component type set T;

[0081] For a data partition P v , let the corresponding graph model node set be C v , traverse the node set C v Nodes in Will be with the node The data partition P where all nodes connected by the edge are located α , added to a set of adjacent data partitions A, which contains a series of data partitions connected by edges;

[0082] The component type set, the component quantity, and the adjacent partition set are integrated to form a prediction index table, which is expressed as θ = {N, T, A}.

[0083] S3: Determine the initial data partition of the building operation and maintenance data based on the prediction index table, and introduce the random forest model to predict the next data partition where the collision will occur;

[0084] The steps to determine the initial data partition of building operation and maintenance data based on the prediction index table are:

[0085] Traverse the prediction index table and set the component type set involved in the collision detection request as T C For each partition and its corresponding prediction index table, check whether each component type in the component type set T belongs to T C ;

[0086] Define a matching function F T (P i ):

[0087]

[0088] in, Indicates the first Type of component, when F T (P i )>0, the corresponding data partition P i As the initial data partition, an initial data partition set S = {P i1 , P i2 , .., P in};

[0089] For each partition P i and its corresponding prediction index table, by checking the component type set in the data partition and the specific component type set T C The relationship between the components determines the initial data partition. This component type matching-based method can quickly screen out the initial data partition from the perspective of data structure.

[0090] For example, suppose there is BIM operation and maintenance data of a large commercial building. The building is divided into multiple data partitions, including shopping malls, office areas, etc. We want to check the fire protection system related components in the building to ensure the normal operation of the fire protection system. In the fire protection system inspection, the specific component type set C T Including: fire hose, fire sprinkler, fire hydrant, fire alarm, in the data partition P1 includes the first floor of the shopping mall, the corresponding component type set T1 includes: floor tiles, lighting fixtures, fire hose, shop counter, the number of component types m = 4, in the data partition P2 includes the second floor of the office, the component types include: office desks and chairs, network lines, air conditioning ducts, the number of component types m = 3, through the matching function F T (P i ), substitute P1 and P2 into it, and we get F T (P1)>1,F T (P2)=0, so the corresponding data partition P1 is used as the initial partition. Through the above calculation, the initial partition set will include the data partition mall layer, because these data partitions contain component types related to the fire protection system (fire water pipes and fire hydrants). In this way, when conducting a fire protection system inspection, priority is given to the initial data partition without the need for a comprehensive inspection of the entire system first.

[0091] The implementation process of predicting the next data partition where a collision occurs is:

[0092] A random forest model is introduced, and the collision data in the historical operation and maintenance data is used as training data for model training;

[0093] The random forest consists of multiple decision trees. Let the number of decision trees be i (the same as the number of data partitions). For each decision tree, samples are randomly extracted from the training data set with replacement to form a training subset of the decision tree. At each node, a part of the features is randomly selected from all the features, and then the best features are selected from these features for splitting. The decision tree is continuously grown until the stopping condition is met (such as the depth of the tree reaches a preset value), and finally a random forest model is obtained.

[0094] Starting from the initial data partition, for the new data partition, the component type and the components in the adjacent partitions are input into the trained random forest model. Extract the characteristic values ​​of the initial data partition, that is, the component type in the partition and the component type in the adjacent partition, and obtain a series of characteristic data The feature data Input into the trained random forest model, each decision tree makes a prediction based on the input features, obtains a prediction result Y, and outputs the final collision probability based on the prediction result Among them, the final collision probability prediction result is the average value of all decision tree prediction results. It represents the predicted probability value of the collision between component type t (initial data partition) and components in adjacent data partition P. After traversing the collision probability after all initial data partitions are input, a collision threshold σ is set. When , the adjacent data partitions of the initial data partition in the input data that are larger than the collision threshold σ are marked as the next data partition S that will collide next .

[0095] S4: Use the optimized collision detection algorithm to perform collision detection on the next data partition where the collision occurs, determine whether the component space bounding boxes intersect, and obtain the collision detection results. Based on the collision detection results, a real-time warning is issued for the detected collision event on the building operation and maintenance supervision interface;

[0096] The steps to obtain the collision detection results are:

[0097] According to the prediction results, get the next collision data partition set S next , which is predicted by the random forest model mentioned above, and the data partitions that may collide have been marked.

[0098] Define a data partition P S ∈S next Partition, data partition P S The component set in C S =(c S1 , c S2 , .., c Sq ), where q represents the data partition P S The number of components within;

[0099] By C S Any two components c Sα , c Sβ , α∈q, β∈q and α≠β, let component c Sα The bounding box is B Sα =(x maxα , x minα ,y maxα ,y minα , z maxα , z minα), where for component c Sα , its spatial bounding box B Sα Defined by six parameters, x maxα and x minα Respectively represent the minimum and maximum coordinate values ​​of the component in the axial direction; y maxα and minα Respectively represent the minimum and maximum coordinate values ​​of the component in the axial direction; z maxα and z minα Respectively represent the minimum and maximum coordinate values ​​of the component in the axial direction;

[0100] Component c Sβ The bounding box is B Sβ =(x maxβ , x minβ ,y maxβ ,y minβ , z maxβ , z minβ ), where x maxβ and x minβ Respectively represent the minimum and maximum coordinate values ​​of the component in the axial direction; y maxβ and minβ Respectively represent the minimum and maximum coordinate values ​​of the component in the axial direction; z maxβ and z minαβ Respectively represent the minimum and maximum coordinate values ​​of the component in the axial direction;

[0101] The formula used to determine whether two spatial bounding boxes intersect is:

[0102]

[0103] When all conditions are met, the judgment result is that the spatial bounding boxes of the two components intersect, that is, the two components collide, otherwise there is no collision;

[0104] The judgment basis is to find a constructed space bounding box so that the space bounding box satisfies all the formula conditions, because the space bounding box is a circumscribed cuboid of the component. If the space bounding boxes of two components intersect, it means that the two components have overlapping areas in the three-dimensional space, and then the two components collide;

[0105] Specifically, in the x-axis direction, the maximum value of one bounding box must be greater than or equal to the minimum value of the other bounding box, and its minimum value must be less than or equal to the maximum value of the other bounding box, so that the two bounding boxes have overlapping parts in the axial direction. Similarly, the conditions must also be met in the y-axis and z-axis directions. Only when the conditions are met in all three directions, the two bounding boxes intersect.

[0106] Specifically, after knowing that two components have collided, a real-time warning of the detected collision event is issued through the building operation and maintenance supervision interface, and the relevant operation and maintenance personnel are informed through system interface prompts and notifications. The main idea of ​​this solution is to manage the BIM model data in data partitions and adopt an optimized collision detection algorithm. By dividing the building model into multiple data partitions according to functional areas or floors, when performing collision detection, the data partitions that may collide are first screened, and then the collision detection algorithm is executed in the screened data partitions, thereby reducing unnecessary calculations and improving collision detection efficiency.

[0107] Embodiment 2

[0108] like Figure 1 to Figure 2 As shown, based on the first embodiment,

[0109] Data partitioning module: obtains building operation and maintenance data and divides the building operation and maintenance data into n data partitions according to the layout and structural system of the building;

[0110] Data index establishment module: establishes corresponding data partition index for each data partition, and outputs the prediction index table using the building component graph model;

[0111] Data partition prediction module: Determine the initial data partition of building operation and maintenance data based on the prediction index table, and introduce the random forest model to predict the next data partition where collision will occur;

[0112] Data detection module: Use the optimized collision detection algorithm to perform collision detection on the next data partition where a collision occurs, determine whether the component space bounding boxes intersect, and obtain the collision detection results. Based on the collision detection results, a real-time warning is issued for the detected collision event on the building operation and maintenance supervision interface.

[0113] In the application, the several formulas involved are all calculated by taking their numerical values ​​after removing the dimensions, and the formulas are established by collecting a large amount of data and performing software simulation to obtain a formula for the most recent actual situation. Some coefficients or weights in the formulas are set by technical personnel in this field according to actual conditions, so they will not be elaborated here.

[0114] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented using software, the above embodiments may be implemented in whole or in part in the form of a computer program product. A person of ordinary skill in the art may appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein may be implemented in electronic hardware, or in a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution.

[0115] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A BIM-based building operation and maintenance supervision method, characterized in that: The steps of this method are as follows: S1: Obtain building operation and maintenance data, and divide the building operation and maintenance data into n data partitions according to the layout and structural system of the building; S2: Establish a corresponding data partition index for each data partition, and use the building component graph model to output a prediction index table; S3: Determine the initial data partition of the building operation and maintenance data based on the prediction index table, and introduce the random forest model to predict the next data partition where the collision will occur; S4: Use the optimized collision detection algorithm to perform collision detection on the next data partition where the collision occurs, determine whether the component space bounding boxes intersect, and obtain the collision detection results. Based on the collision detection results, a real-time warning is issued for the detected collision event on the building operation and maintenance supervision interface.

2. A BIM-based building operation and maintenance supervision method according to claim 1, characterized in that: The process of dividing the building operation and maintenance data into multiple data partitions is as follows: Obtain information data covering the entire life cycle of the building from the building operation and maintenance database, including geometric information, physical information and material information. Let the acquired building operation and maintenance data be D; According to the functional layout and spatial layout of the building, the data set is divided into n data partitions, denoted as D = (P1...P2, .., P n ); The data set is divided into n data partitions by functional areas and spatial layout.

3. The BIM-based building operation and maintenance supervision method according to claim 1 is characterized in that: The prediction index table includes a data partition component set, a component quantity, and an adjacent data partition set.

4. A BIM-based building operation and maintenance supervision method according to claim 3, characterized in that: The implementation process of outputting the corresponding prediction index table through the building component graph model is as follows: Construct architectural component diagram models; The construction process of the building component graph model is as follows: Extract all the relevant information of components from the building operation and maintenance data, including the geometric information, physical information, spatial location information and the connection relationship between components. Suppose the component set is C = (c1, c2, .., c N ), where N is the number of components; Treat each component as a node in a graph model, and the attributes of the component as the attributes of the node; Assume that the basic framework of the building component graph model is G = (C, E), where E is the edge set, expressed as E = {e jk │c j , c k }, specific, e jk It consists of two components c j , c k The connection relationship between them is determined, and 1≤j≤N, 1≤k≤N, j≠k; The output process of the prediction index table is: Traverse all nodes in the building component graph model and obtain the component type t of each node; Through a data partition P n , n∈n, traverse the data partition P n All nodes of the node are counted, and the component types of all nodes are counted to obtain a component type set T = (t1, t2, .., t N ); For a data partition P v , let the corresponding graph model node set be C v , traverse the node set C v Node c in v , will be associated with node c v The data partition P where all nodes connected by the edge are located α , added to a set of adjacent data partitions A, where the set A contains data partitions connected by edges; The component type set, component quantity, and adjacent partition set are integrated to form a prediction index table, expressed as θ = {N, T, A}.

5. A BIM-based building operation and maintenance supervision method according to claim 4, characterized in that: The step of determining the initial data partition of the building operation and maintenance data based on the prediction index table is: Traverse the prediction index table and set the component type set involved in the collision detection request as T C For each partition and its corresponding prediction index table, check whether each component type in the component type set T belongs to T C ; Define a matching function F T (P i ): in, Indicates the first Type of component, when F T (P i )>0, the corresponding data partition P i As the initial data partition, an initial data partition set S = {P i1 , P i2 , .., P in }.

6. A BIM-based building operation and maintenance supervision method according to claim 5, characterized in that: The implementation process of predicting the next data partition where a collision occurs is as follows: A random forest model is introduced, and the collision data in the historical operation and maintenance data is used as training data for model training; Starting from the initial data partition, for the new data partition, the component type and the components in the adjacent partitions are input into the trained random forest model. in , P in ∈S, extract the characteristic value of the initial data partition, that is, the component type in the partition and the component type in the adjacent partition, and obtain the characteristic data The feature data Input into the trained random forest model to obtain the prediction result Y; Output the final collision probability based on the prediction results Among them, the final collision probability prediction result is the average value of all decision tree prediction results. It represents the predicted probability value of the collision between component type t and components in adjacent data partition P. After traversing the collision probability after all initial data partition inputs, a collision threshold σ is set. When , the adjacent data partitions of the initial data partition in the input data that are larger than the collision threshold σ are marked as the next data partition S that will collide next .

7. A BIM-based building operation and maintenance supervision method according to claim 6, characterized in that: The steps for obtaining the collision detection result are: Define a data partition P S ∈S next Partition, data partition P S The component set in C S =(c S1 , c S2 , .., c Sq ), where q represents the data partition P S The number of components within; By C S Any two components c Sα , c Sβ , α∈q, β∈q and α≠β, let component c Sα The bounding box is B Sα =(x maxα , x minα ,y maxα ,y minα , z maxα , z minα ), where for component c Sα , its spatial bounding box B Sα Defined by six parameters, x maxα and x minα Respectively represent the minimum and maximum coordinate values ​​of the component in the axial direction; y maxα and minα Respectively represent the minimum and maximum coordinate values ​​of the component in the axial direction; z maxα and z minα Respectively represent the minimum and maximum coordinate values ​​of the component in the axial direction; Component c Sβ The bounding box is B Sβ =(x maxβ , x minβ ,y maxβ ,y minβ , z maxβ , z minβ ), where x maxβ and x minβ Respectively represent the minimum and maximum coordinate values ​​of the component in the axial direction; y maxβ and minβ Respectively represent the minimum and maximum coordinate values ​​of the component in the axial direction; z maxβ and z minαβ Respectively represent the minimum and maximum coordinate values ​​of the component in the axial direction; The following formula is used to determine whether two spatial bounding boxes intersect: When all conditions are met, the judgment result is that the spatial bounding boxes of the two components intersect, otherwise they do not intersect.

8. A BIM-based building operation and maintenance supervision system according to claim 1, using any one of the methods in claims 1 to 7, characterized in that: include: Data partitioning module: obtains building operation and maintenance data and divides the building operation and maintenance data into n data partitions according to the layout and structural system of the building; Data index establishment module: establishes corresponding data partition index for each data partition, and outputs a prediction index table using the building component graph model; Data partition prediction module: Determine the initial data partition of building operation and maintenance data based on the prediction index table, and introduce the random forest model to predict the next data partition where collision will occur; Data detection module: Use the optimized collision detection algorithm to perform collision detection on the next data partition where a collision occurs, determine whether the component space bounding boxes intersect, and obtain the collision detection results. Based on the collision detection results, a real-time warning is issued for the detected collision event on the building operation and maintenance supervision interface.

Images