Component system driven cost plan generation method and system

Abstract

This invention relates to the field of engineering cost technology, and discloses a method and system for generating cost schemes driven by a component system. The method includes: extracting annotation information from 2D CAD drawings and identifying the spatial positioning attributes of components; generating 3D bounding boxes for component positioning based on multi-view geometric projection rules; inferring spatial adjacency relationships of components and calculating adjacency geometric parameters through bounding box intersection detection; expanding the component system to add spatial adjacency trigger dependency definitions; parsing bill of quantities items to match adjacency records and calculating linked quota quantities; and generating a cost combination scheme including linked quotas. This invention solves the technical problem of not being able to automatically identify and measure linked quotas dependent on spatial adjacency relationships when a BIM model is lacking.

Description

Technical Field

[0001] This invention relates to the field of engineering cost technology, and more specifically, to a method and system for generating cost estimates driven by component systems. Background Technology

[0002] In bill of quantities pricing scenarios, the system establishes a component system based on the work content corresponding to the quota items, parsing the bill of quantities items into components and influencing factors before matching the quota items. Within the component system, some components have interrelationships, such as reinforcement bars at concrete column-beam joints, or tie bars between adjacent walls. The calculation of these interrelationship quotas depends on the spatial adjacency relationships between component instances. In many existing engineering projects, only two-dimensional CAD design drawings are available, without BIM models to extract component spatial location data.

[0003] The existing component system scheme matches quota items based on factors such as component type, material, and shape. The relationships between components are defined by type logic, but no way to obtain the spatial adjacency relationships between component instances has been established.

[0004] When a project lacks a BIM model, existing technologies cannot obtain the three-dimensional spatial location of components for adjacency detection, resulting in the inability to automatically identify and measure the linked quotas that rely on spatial adjacency relationships, affecting the completeness and accuracy of the cost plan. Summary of the Invention

[0005] This invention provides a method and system for generating cost estimates driven by component systems, which solves the technical problems in related technologies such as the inability to automatically identify the spatial adjacency relationship of components from two-dimensional drawings and the difficulty in realizing linkage quota measurement driven by adjacency relationship.

[0006] This invention discloses a method for generating cost estimates driven by a component system, comprising the following steps: acquiring engineering structural plan, elevation, and section data; performing annotation recognition on each drawing to extract the text content, coordinates, and view type of each annotation, generating a drawing annotation dataset; performing entity recognition on the annotation text in the drawing annotation dataset to extract component type, positioning axis number, elevation value, and cross-sectional dimensions, generating component spatial positioning attribute records; based on multi-view geometric projection rules, fusing the plan coordinates, elevation coordinates, and section coordinates into a three-dimensional spatial region estimate to generate a three-dimensional positioning bounding box for each component; and performing spatial intersection analysis on the three-dimensional positioning bounding boxes. The component performs adjacency relationship inference, determines the adjacency type based on component type combination rules, extracts cross-sectional parameters from associated annotation attributes to calculate adjacency geometric parameters, and generates a component spatial adjacency relationship table; it expands the component inter-component relationship data structure in the component system, adding dependency definitions for spatial adjacency trigger types; it obtains list item data, parses it into main components and influencing factors, and retrieves adjacency records of main component instances in the component spatial adjacency relationship table; it matches adjacency records with dependency definitions of spatial adjacency trigger types, and substitutes dependency definitions that meet trigger conditions into adjacency geometric parameters to calculate the linked quota engineering quantity; it merges linked component instances with main components, matches quota sub-items, and generates cost combination schemes.

[0007] Furthermore, the annotation recognition process for each drawing includes: using an object detection neural network to locate and recognize text annotations, dimension annotations, and leader line annotations in the drawing; the input of the object detection neural network is the drawing image tensor, which generates candidate regions through a convolutional feature extraction layer and a region proposal network, and outputs the annotation bounding box coordinates, annotation category, and confidence score through a classification layer and a regression layer.

[0008] Furthermore, the entity recognition process for the annotation text in the drawing annotation dataset includes: using a combination of rule template matching and named entity recognition model; rule template matching extracts attributes based on the fixed format features of the engineering drawing annotations; and the named entity recognition model is used to process non-standard format annotation text.

[0009] Furthermore, the named entity recognition model adopts a sequence labeling architecture combining a bidirectional long short-term memory neural network and a conditional random field. The input of the named entity recognition model is a character sequence of labeled text. Each character is converted into a vector representation through a word embedding layer, input to a bidirectional long short-term memory neural network layer for context encoding, and then optimized for global sequence labeling through a conditional random field layer. The output is a sequence of entity labels corresponding to the character sequence.

[0010] Furthermore, generating the three-dimensional positioning bounding box for each component includes: obtaining the coordinates of the axis intersection points by querying the axis coordinate table according to the positioning axis number in the plan view; determining the X and Y coordinate ranges of the component in the horizontal plane by combining the relative relationship between the component's marked position and the axis intersection points; extracting the bottom and top elevation values ​​of the component from the elevation markings in the elevation or section view to determine the Z coordinate range of the component in the vertical plane; and generating a three-dimensional axis-aligned bounding box that surrounds the component entity by expanding the corresponding dimension distance in each direction with the positioning reference point as the center according to the cross-sectional dimension parameters.

[0011] Furthermore, the generation of the three-dimensional positioning bounding box for each component also includes: cross-validating the positioning information from different views when the same component is marked in multiple drawings; when the horizontal positioning information of the same component appears in both the plan view and the section view, comparing the coordinate range obtained from the two analyses, if the deviation exceeds a preset tolerance threshold, manual verification and marking are performed, and if the deviation is within the preset tolerance threshold range, the intersection of the two is taken as the final coordinate range.

[0012] Furthermore, the adjacency relationship inference for component pairs with spatial intersection of 3D positioning bounding boxes includes: traversing all component pairs, performing intersection detection on the 3D positioning bounding boxes of each pair of components, determining whether the projection intervals of the two 3D positioning bounding boxes overlap in the X, Y, and Z directions, and filtering out candidate adjacent component pairs with spatial intersection; obtaining the component types of the two components in the candidate adjacent component pairs, querying a predefined component type combination rule table, and determining the adjacency type of the candidate adjacent component pairs; extracting relevant cross-sectional parameters from the spatial positioning attribute records of the adjacent component pairs according to the geometric calculation rules corresponding to the adjacency type, and calculating the contact area or intersection length; combining the component pair identifier, adjacency type, and adjacency geometric parameters into an adjacency record, and summarizing all adjacency records to generate a component spatial adjacency relationship table.

[0013] Furthermore, the adjacency relationship inference for component pairs with spatial intersection of three-dimensional positioning bounding boxes also includes: setting a minimum intersection volume threshold, and determining that the corresponding component pair has an adjacency relationship only when the intersection volume of two three-dimensional positioning bounding boxes is greater than the minimum intersection volume threshold; the intersection volume is calculated based on the product of the overlap lengths of the two three-dimensional positioning bounding boxes in the X-axis, Y-axis and Z-axis directions, and the overlap length in each direction is a non-negative value.

[0014] Furthermore, the dependency definition of the spatial adjacency trigger type includes a main component type field, an adjacent component type field, a linked component type field, an adjacent geometric parameter reference identifier field, and a linked quota quantity calculation formula field; the step of substituting the dependency definition that meets the trigger condition into the adjacent geometric parameter to calculate the linked quota quantity includes: querying the component system according to the main component type to obtain the dependency definitions of all spatial adjacency trigger types corresponding to the main component type; traversing the adjacency records of the main component instance, matching the component type combination in each adjacency record with the dependency definition of each spatial adjacency trigger type to determine the triggered dependency definition and the corresponding linked component type; obtaining the corresponding adjacent geometric parameter value from the matched adjacent record according to the adjacent geometric parameter reference identifier in the triggered dependency definition, substituting it into the linked quota quantity calculation formula to calculate the quantity value of the linked component; generating a linked component instance with a quantity, wherein the linked component instance includes the linked component type, quantity value, unit of measurement, and associated main component instance identifier.

[0015] This invention provides a component system-driven cost estimation scheme generation system, comprising: a drawing annotation extraction module, used to acquire engineering structural plan views, elevation views, and section views, perform annotation recognition on each drawing, extract the text content, drawing coordinates, and view type of each annotation, and generate a drawing annotation dataset; a spatial positioning attribute extraction module, used to perform entity recognition on the annotation text in the drawing annotation dataset, extract component type, positioning axis number, elevation value, and cross-sectional dimensions, and generate component spatial positioning attribute records; a 3D bounding box generation module, used to fuse plan view coordinates, elevation view coordinates, and section view coordinates into a 3D spatial region estimate based on multi-view geometric projection rules, and generate 3D positioning bounding boxes for each component; and an adjacency relationship inference module, used to store the 3D positioning bounding boxes. The system performs adjacency relationship inference on spatially intersecting component pairs, determines the adjacency type based on component type combination rules, calculates adjacency geometric parameters, and generates a component spatial adjacency relationship table. The component system extension module expands the data structure of relationships between components within the component system, adding dependency definitions for spatial adjacency trigger types. The bill of quantities parsing module acquires bill of quantities item data, parses it into main components and influencing factors, and retrieves adjacency records of main component instances from the component spatial adjacency relationship table. The linked engineering quantity calculation module matches adjacency records with dependency definitions for spatial adjacency trigger types, and substitutes the adjacency geometric parameters into dependency definitions that meet the trigger conditions to calculate the linked quota engineering quantity. The cost scheme generation module merges linked component instances with main components, matches quota items, and generates a cost combination scheme.

[0016] This invention extracts annotation information from 2D CAD drawings and reconstructs the 3D bounding box of components based on multi-view geometric projection rules, enabling the acquisition of component spatial positions in the absence of a BIM model. It establishes a method for obtaining spatial adjacency relationships between component instances by inferring spatial adjacency relationships and calculating adjacency geometric parameters through bounding box intersection detection and component type combination rules. Furthermore, by extending the component system to add spatial adjacency trigger dependency definitions and automatically calculating linked quota quantities by matching adjacency records, it achieves automatic identification and measurement of linked quotas dependent on spatial adjacency relationships. This invention solves the technical problem of being unable to automatically identify and measure linked quotas dependent on spatial adjacency relationships when a BIM model is lacking, improving the completeness and automation of cost estimation schemes. Attached Figure Description

[0017] Figure 1 This is a flowchart of a method for generating a cost estimate based on drawing annotations to generate spatial adjacency relationships of components, provided in an embodiment of the present invention. Figure 2 This is a sub-flowchart of generating a three-dimensional positioning bounding box for a component in the cost scheme generation method based on drawing annotations to generate spatial adjacency relationships of components provided in the embodiments of the present invention. Figure 3 This is a sub-flowchart of inferring the spatial adjacency relationship of components in the cost scheme generation method based on drawing annotations provided in this embodiment of the invention; Figure 4 This is a sub-flowchart of the method for generating cost schemes based on drawing annotations to generate spatial adjacency relationships of components, provided in an embodiment of the present invention, for calculating the linked quota engineering quantity; Figure 5 This is a distribution diagram of the number of CAD drawing annotations to be recognized provided in an embodiment of the present invention; Figure 6 This is a graph showing the accuracy of component spatial positioning attribute extraction provided in an embodiment of the present invention. Figure 7 This is a comparison diagram of the coordinate range of the three-dimensional bounding box of the components provided in the embodiments of the present invention; Figure 8 This is a flowchart of the component spatial adjacency detection process provided in the embodiments of the present invention; Figure 9 This is a scatter plot of the calculation parameters for the reinforcement quantity of beam-column joints provided in this embodiment of the invention; Figure 10 This is a quantity statistics chart of an example of a linkage component provided in an embodiment of the present invention; Figure 11 This is a cost composition analysis diagram of the cost combination scheme provided in the embodiment of the present invention. Detailed Implementation

[0018] In bill of quantities pricing scenarios, the system establishes a component system based on the work content corresponding to the quota items, parsing the bill of quantities items into components and influencing factors before matching the quota items. Within the component system, some components have interrelationships, such as reinforcement bars at concrete column-beam joints, or tie bars between adjacent walls. The calculation of these interrelationship quotas depends on the spatial adjacency relationships between component instances. In many existing engineering projects, only two-dimensional CAD design drawings are available, without BIM models to extract component spatial location data.

[0019] The existing component system scheme matches quota items based on factors such as component type, material, and shape. The relationships between components are defined by type logic, but no method has been established to obtain the spatial adjacency relationships between component instances. When a project lacks a BIM model, it is impossible to obtain the three-dimensional spatial position of components for adjacency detection. This results in the inability to automatically identify and measure linked quotas that rely on spatial adjacency relationships, affecting the integrity of the cost plan.

[0020] At least one embodiment of the present invention discloses a method for generating cost schemes driven by component systems, such as... Figure 1 As shown, it includes the following steps: Step 1: Obtain drawing data and extract annotation information; Acquire engineering structural plan, elevation, and section data, perform OCR recognition and annotation area detection on each drawing, extract the text content, drawing coordinates, and view type of each annotation, and generate a drawing annotation dataset.

[0021] It should be noted that drawing coordinates refer to the two-dimensional coordinate values ​​marked on the corresponding view drawing, which can be pixel coordinates or engineering coordinates in the drawing coordinate system. The view type refers to the projection type of the drawing on which the annotation is located, including horizontal projection plan views, orthographic projection elevation views, and longitudinal section views.

[0022] Furthermore, the annotation region detection employs an object detection neural network to locate and identify text annotations, dimension annotations, and leader line annotations in the drawing, outputting the bounding box coordinates and annotation category for each annotation. The input to the object detection neural network is the drawing image tensor, which generates candidate regions through a convolutional feature extraction layer and a region proposal network, and then outputs the annotation bounding box coordinates, annotation category, and confidence score through a classification layer and a regression layer.

[0023] Step 2: Extract the spatial positioning attributes of the components; Entity recognition is performed on the annotation text in the drawing annotation dataset to extract spatial positioning attributes such as component type, positioning axis number, elevation value and cross-sectional dimensions, and generate component spatial positioning attribute records.

[0024] It should be noted that the component type includes structural components such as columns, beams, walls, and slabs; the positioning axis number is used to determine the horizontal position of the component relative to the building axis grid in the plan view; the elevation value is used to determine the starting and ending elevations of the component in the vertical direction; and the cross-sectional dimensions are used to determine the geometric cross-sectional parameters of the component.

[0025] It should be noted that entity recognition can be performed using a combination of rule-based template matching and named entity recognition models. Rule-based template matching extracts attributes from the fixed format features of engineering drawing annotations; for example, column numbering is usually a combination of axis intersection numbers and cross-sectional dimensions. Named entity recognition models are used to handle non-standard format annotation text.

[0026] Furthermore, the named entity recognition model employs a sequence labeling architecture combining a bidirectional long short-term memory neural network (LSTM) and a conditional random field (CRF). The input to the named entity recognition model is a sequence of characters from the labeled text. Each character is converted into a vector representation through a word embedding layer, then input into the bidirectional LSTM neural network layer for context encoding, and finally optimized for global sequence labeling through a CRF layer. The output is a sequence of entity labels corresponding to the character sequence. The label types of the entity labels include entity categories such as component type, axis number, elevation value, and cross-sectional dimensions. The output entity label sequence of the named entity recognition model directly corresponds to the entity category of each character, allowing for the extraction of structured spatial localization attributes without additional decoding.

[0027] Step 3: Generate the 3D positioning bounding box of the component; Based on multi-view geometric projection rules, the coordinates of the plan view, elevation view, and section view are fused and converted into a three-dimensional spatial region estimate, generating a three-dimensional positioning bounding box for each component.

[0028] Furthermore, the multi-view geometric projection rule refers to the orthographic projection principle in architectural drawings. That is, the plan view provides the X-axis and Y-axis coordinate information of the component in the horizontal plane, while the elevation and section views provide the Z-axis coordinate information of the component in the vertical plane. By combining the two-dimensional projection coordinates of the same component in different views according to the corresponding axes, the position range of the component in three-dimensional space can be reconstructed.

[0029] Step 3 includes the following sub-steps: Step 3.1: Based on the positioning axis number in the plan view, look up the axis coordinate table to obtain the coordinates of the axis intersection point. Combine the relative relationship between the component's marked position and the axis intersection point to determine the X and Y coordinate ranges of the component in the horizontal plane.

[0030] Furthermore, the axis coordinate table is a coordinate mapping table established based on the axis numbers and axis spacing dimensions marked on the plan view. The axis coordinate table records the X-axis or Y-axis coordinate values ​​in the engineering coordinate system corresponding to each axis number. When determining the coordinate range of a component, the reference coordinates are first obtained from the axis coordinate table based on the axis number in the component's annotation, and then the minimum and maximum values ​​of the coordinate range are calculated based on the offset distance of the component's annotation position relative to the intersection point of the axes and the cross-sectional dimensions.

[0031] Step 3.2: Extract the bottom and top elevation values ​​of the component based on the elevation markings in the elevation or section view, and determine the vertical Z-coordinate range of the component.

[0032] Step 3.3: Based on the cross-sectional dimension parameters, expand the corresponding dimension distance in each direction with the positioning reference point as the center to generate a three-dimensional axis-aligned bounding box that surrounds the component entity.

[0033] It should be noted that the 3D positioning bounding box is represented using an axis-aligned bounding box, with the smallest vertex coordinates... and the coordinates of the largest vertex Define the spatial extent of the 3D positioning bounding box, where , , These represent the minimum coordinate values ​​of the 3D positioning bounding box in the X, Y, and Z axes, respectively. , , These represent the maximum coordinate values ​​of the 3D positioning bounding box in the X, Y, and Z axes, respectively.

[0034] In this embodiment of the application, in order to improve the accuracy of the 3D positioning bounding box, for cases where the same component is marked in multiple drawings, the positioning information from different views is cross-validated. When the horizontal positioning information of the same component appears in both the plan view and the section view, the range of the X or Y coordinates obtained from the two are compared. If the deviation exceeds a preset tolerance threshold, it is manually checked and marked. If the deviation is within the preset tolerance threshold range, the intersection of the two is taken as the final coordinate range.

[0035] Furthermore, the preset tolerance threshold is determined based on the scale of the engineering drawings and the accuracy of the annotation recognition. The typical range of the preset tolerance threshold is 50 mm to 200 mm, and the specific value of the preset tolerance threshold is set by the user according to the accuracy requirements of the project.

[0036] Step 4: Infer the spatial adjacency relationship of components; For component pairs with spatial intersections in the 3D positioning bounding boxes, adjacency relationship inference is performed. The adjacency type is determined according to the component type combination rules. The cross-sectional parameters are extracted from the associated annotation attributes to calculate the contact area and intersection length, and a component spatial adjacency relationship table is generated.

[0037] Step 4 includes the following sub-steps: Step 4.1: Traverse all component pairs, perform intersection detection on the 3D bounding boxes of each component pair, determine whether the projection intervals of the two 3D bounding boxes overlap in the three directions of X-axis, Y-axis and Z-axis, and filter out candidate adjacent component pairs with spatial intersection.

[0038] Furthermore, the overlap of the projection ranges of two 3D positioning bounding boxes on a certain axis means that the coordinate ranges of the two 3D positioning bounding boxes intersect on that axis. The judgment condition is that the maximum coordinate value of the first 3D positioning bounding box is greater than the minimum coordinate value of the second 3D positioning bounding box and the minimum coordinate value of the first 3D positioning bounding box is less than the maximum coordinate value of the second 3D positioning bounding box.

[0039] Step 4.2: Obtain the component types of the two components in the candidate adjacent component pair, query the predefined component type combination rule table, and determine the adjacency type of the candidate adjacent component pair.

[0040] Step 4.3: Based on the geometric calculation rules corresponding to the adjacency type, extract relevant cross-sectional parameters from the spatial positioning attribute records of the adjacent component pair, and calculate the adjacency geometric parameters such as contact area or intersection length.

[0041] Step 4.4: Combine the component pair identifier, adjacency type, and adjacency geometry parameters into an adjacency record, and summarize all adjacency records to generate a component spatial adjacency relationship table.

[0042] It should be noted that the component type combination rule table is a predefined set of rules that records the correspondence between different component type combinations and their adjacency types. For example, the combination of column components and beam components corresponds to the beam-column node adjacency type, the combination of wall components corresponds to the wall tie adjacency type, and the combination of column components and foundation components corresponds to the column base connection adjacency type.

[0043] It should be noted that the calculation method for the contact area depends on the adjacency type. For example, for beam-column joints, the contact area is the beam cross-sectional area, and the formula for calculating the contact area is... ,in For contact area, The width of the beam section. The beam section height is given. For wall tie-and-connection types, the intersection length is the vertical height at the intersection of the two walls, and the intersection length is taken as the length of the overlapping interval of the two walls in the Z-axis direction.

[0044] In this embodiment, to filter out false adjacencies caused by annotation recognition errors or cross-sectional dimension deviations, a minimum intersection volume threshold is set. Only when the intersection volume of two 3D bounding boxes exceeds the minimum intersection volume threshold are the corresponding component pairs determined to have an adjacency relationship. The intersection volume calculation formula is as follows:

[0045] in, For the intersection volume, , , These represent the overlap lengths of the two 3D bounding boxes along the X, Y, and Z axes, respectively. This represents the function that takes the maximum value.

[0046] Furthermore, the minimum intersection volume threshold is determined based on the component type and engineering accuracy requirements. The typical range of the minimum intersection volume threshold is 0.001 cubic meters to 0.01 cubic meters. The specific value of the minimum intersection volume threshold is set by the user according to the project characteristics.

[0047] Step 5: Expand the spatial adjacency triggering dependency definition of the component system; The component relationship data structure in the extended component system is expanded to include a dependency definition for spatial adjacency trigger types. The trigger condition for the dependency definition of spatial adjacency trigger types is that there are adjacency records with a specific combination of component types in the component spatial adjacency relationship table, and the reference identifier of the adjacency geometric parameters is associated.

[0048] It should be noted that the dependency definition of the spatial adjacency trigger type includes the following data fields: main component type field, adjacent component type field, linked component type field, adjacent geometric parameter reference identifier field, and linked quota quantity calculation formula field. The main component type field and adjacent component type field are used to match the adjacent records in the component spatial adjacency relationship table; the linked component type field is used to determine the component type corresponding to the linked quota generated after triggering; the adjacent geometric parameter reference identifier field is used to specify which adjacent geometric parameters are obtained from the adjacent records to participate in the quantity calculation; and the linked quota quantity calculation formula field is used to define the calculation logic of the quantity.

[0049] It should be noted that the dependency definition examples of spatial adjacent contact types include: the main component type is a concrete column, the adjacent component type is a concrete beam, the linked component type is a beam-column joint reinforcement, the adjacent geometric parameter reference identifier is the contact area, and the linked quota engineering quantity calculation formula is to calculate the weight of the steel bars based on the contact area and the joint reinforcement ratio.

[0050] Step 6: Parse the list items and identify adjacent records; Obtain the list item data, parse the main components and influencing factors according to the list name and item characteristics, and retrieve all adjacency records of the main component instance in the component spatial adjacency relationship table based on the spatial positioning attributes of the main component.

[0051] It should be noted that the parsing process for the list items includes: identifying the main component type based on the list name through text matching or classification models; extracting influencing factors such as material, specifications, and location based on the descriptive information in the item features; and determining the corresponding main component instance by associating the location description in the item features with the component's spatial location attribute record.

[0052] Step 7: Calculate the quantity of work in the linked quota; Traverse the dependency definitions of spatial adjacency trigger types of the main component in the component system, match the adjacency records of the main component instance with the dependency definitions of spatial adjacency trigger types, substitute the dependency definitions of spatial adjacency trigger types that meet the trigger conditions into the adjacency geometric parameters to calculate the linkage quota engineering quantity, and generate linkage component instances with engineering quantities.

[0053] Step 7 includes the following sub-steps: Step 7.1: Query the component system based on the main component type to obtain the dependency definitions of all spatial adjacency trigger types corresponding to the main component type.

[0054] Step 7.2: Traverse the adjacency records of the main component instance, match the component type combination in each adjacency record with the main component type and adjacent component type in the dependency definition of each spatial adjacency trigger type, and determine the dependency definition of the triggered spatial adjacency trigger type and the corresponding linkage component type.

[0055] Step 7.3: Based on the adjacency geometric parameter reference identifier in the dependency definition of the triggered spatial adjacency trigger type, obtain the corresponding adjacency geometric parameter value from the matched adjacency record, substitute it into the linkage quota engineering quantity calculation formula, and calculate the engineering quantity value of the linkage component.

[0056] Step 7.4: Generate a linked component instance with quantities. The linked component instance with quantities includes the linked component type, quantity value, unit of measurement, and the identifier of the associated main component instance.

[0057] Step 8: Generate a cost combination scheme; The linked component instances with engineering quantities are merged with the main component into a complete component combination. The corresponding quota items of each component are matched from the quota database based on the regional information of the project, and a cost combination scheme including the linked quota of spatial adjacency relationship is generated.

[0058] It should be noted that the matching process for quota sub-items includes: generating quota search conditions based on component type and combination of influencing factors; determining the applicable quota version based on the region to which the project is located; and performing a search in the quota database to obtain the matching quota sub-item number and comprehensive unit price.

[0059] It should be noted that the data structure of the cost combination scheme includes: item identifier in the bill of quantities, quota items and quantities of main components, quota items and quantities of linked components, comprehensive unit price and total price of each quota item, and total amount of the cost combination scheme.

[0060] In this embodiment of the application, in order to facilitate the review and adjustment by cost estimators, the triggering source information of each linkage quota is marked in the cost combination scheme. The triggering source information includes the main component identifier, adjacent component identifier and adjacent geometric parameter value that trigger the linkage quota.

[0061] The embodiments of the present invention have been described above. However, the embodiments are not limited to the specific implementation methods described above. The specific implementation methods described above are merely illustrative and not restrictive. Those skilled in the art can make more equivalent embodiments under the guidance of the present embodiments, and all of them are within the protection scope of the present embodiments.

[0062] A construction company undertook a renovation project for a 20XX-built office building with a frame structure. The project only retained the original 2D CAD construction drawings, including structural plans, elevations, and sections; a BIM model was not created. The cost department needs to price the structural reinforcement work for the three-story office area, which includes 6 concrete columns, 8 frame beams, and corresponding beam-column joint reinforcement. Due to the lack of 3D model data, the system needs to extract the spatial location information of the components from the CAD drawings, identify the adjacency relationships between beams and columns, automatically calculate the quantity of beam-column joint reinforcement, and match the corresponding cost items.

[0063] The system imports CAD files of the three-story structural plan, south elevation, and section 1-1, and performs OCR recognition and annotation area detection on the drawings. In the structural plan, it identifies axis numbers, column numbers, beam numbers, and section dimensions; in the elevation and section views, it identifies elevation and vertical dimension annotations. An object detection neural network locates each type of annotation and outputs the bounding box coordinates, annotation category, and confidence score.

[0064] Table 1. Three-layer planar diagram annotation dataset (partial):

[0065] Table 2. South elevation and section diagram annotation dataset (partial):

[0066] The system performs entity recognition on the labeled text. Using rule-based template matching, it identifies the component type as "column," the component number as "C-3," and the cross-sectional dimension as "600×600" in column number "C-3 (600×600)." It also identifies the component type as "beam," the component number as "KL-5 (300×650)," and the cross-sectional dimension as "300×650" in beam number "KL-5." The system extracts the positioning axis numbers "3 axis" and "C axis" from the axis label "3 / C axis." The named entity recognition model parses the elevation labels "+9.600" and "+13.200," extracting elevation values ​​of 9.600m and 13.200m, respectively.

[0067] Table 3. Partial Record of Spatial Positioning Attributes of Components:

[0068] The system, based on the axis coordinate table, determined that the X-coordinate corresponding to axis 3 is 8250mm, the X-coordinate corresponding to axis 4 is 13500mm, and the Y-coordinate corresponding to axis C is 12400mm. For column member COMP-C3-01, its positioning axis is the intersection of axis 3 and axis C. Combining this with the cross-sectional dimensions of 600×600mm, and extending 300mm outwards from the axis intersection point, the calculated X-coordinate range is 7950mm to 8550mm, and the Y-coordinate range is 12100mm to 12700mm. The bottom elevation extracted from the elevation and section views is 9.600m, and the top elevation is 13.200m, determining the Z-coordinate range to be 9600mm to 13200mm.

[0069] For beam member COMP-KL5-01, which spans between axes 3 and 4 and is located at axis C, the beam section has a width of 300mm and a height of 650mm. The bottom elevation of the beam is 12.950m, and the top elevation is 13.200m (the top of the beam is flush with the top of the column). Based on the beam's section width and positioning axis, the calculated X-coordinate range is 8250mm to 13500mm, the Y-coordinate range is 12250mm to 12550mm, and the Z-coordinate range is 12950mm to 13200mm.

[0070] Table 4. Parameters of the 3D Positioning Bounding Box for Components:

[0071] Taking COMP-C3-01 and COMP-KL5-01 as examples, the consistency of their coordinate ranges was verified. The column top elevation of 13200mm matches the beam top elevation of 13200mm, with a deviation of 0mm, which is less than the preset tolerance threshold of 150mm, and the verification is successful.

[0072] The system traverses the component pairs and performs intersection detection on COMP-C3-01 and COMP-KL5-01. In the X-axis direction, the column's range [7950, 8550] overlaps with the beam's range [8250, 13500] within a region [8250, 8550]. In the Y-axis direction, the column's range [12100, 12700] overlaps with the beam's range [12250, 12550]. In the Z-axis direction, the column's range [9600, 13200] overlaps with the beam's range [12950, ​​13200]. Since overlap exists in all three directions, they are identified as candidate adjacent component pairs.

[0073]

[0074] The intersection volume is 0.0225m³. 3 Greater than the minimum intersection volume threshold of 0.01m 3 This confirms the existence of an adjacency relationship.

[0075] The component type combination rule table shows that the combination of column and beam types corresponds to the "beam-column node adjacency" type. According to the geometric calculation rules for adjacency types, the contact area is taken as the beam cross-sectional area.

[0076] Table 5. Spatial Adjacency Relationships of Components (Partial):

[0077] For adjacency record ADJ-003, the cross section of beam COMP-KL6-01 is 300×700mm, and the contact area is:

[0078] The system adds a dependency definition for spatial adjacent contact types in the component system. For the linked quota scenario of beam-column joint reinforcement, the following rule is defined: when the main component type is "column" and the adjacent component type is "beam," the linked component type "beam-column joint reinforcement" is triggered. The adjacent geometric parameter reference identifier is set to "contact area," and the linked quota quantity calculation formula is defined as: calculating the weight of the reinforcement based on the contact area and the joint reinforcement ratio (taken as 0.8%), with the reinforcement density taken as 7850 kg / m³. 3 The average diameter of the reinforcing bars is taken as 12mm.

[0079] Table 6 Spatial Adjacency Trigger Type Dependency Definition Table:

[0080] The node thickness coefficient is taken as 0.15 times the beam height and is used to estimate the equivalent thickness of the node reinforcement area.

[0081] The cost estimator entered the bill of quantities item "Three-story frame column concrete pouring", bill of quantities number LIST-301, with the item description "C30 concrete, rectangular column 600×600, located at the intersection of axis 3 / C axis, elevation 9.600m to 13.200m". The system parsed the bill of quantities name and identified the main component type as "column". It extracted the positioning information "axis 3 / C axis" and the elevation range "9.600m to 13.200m" from the item characteristics, matched them with the component's spatial positioning attribute record, and associated it with the component instance COMP-C3-01.

[0082] The system retrieves all adjacent records of the main component with ID COMP-C3-01 from the component space adjacency table. The search result is adjacent record ADJ-001, which shows that COMP-C3-01 and beam component COMP-KL5-01 have a beam-column node adjacency relationship with each other, with a contact area of ​​0.195m². 2 .

[0083] Table 7: Association Table of List Item Resolution Results and Adjacency Records

[0084] The system queries the component system based on the main component type "column" to obtain the dependency definition DEP-001. It then traverses the adjacent records ADJ-001 of the main component COMP-C3-01. The component type combination (column-beam) of this record matches the main component type and adjacent component type of the dependency definition DEP-001, triggering the linkage component type "beam-column joint reinforcement".

[0085] The contact area of ​​0.195m² was obtained from the adjacency record ADJ-001. 2 Substitute these values ​​into the calculation formula for the linked quota project quantity. The section height of beam COMP-KL5-01 is 650mm, and the node thickness coefficient is:

[0086] Calculation of steel bar weight:

[0087] Generate a linked component instance LINK-C3-KL5-01 with a quantity of work, the linked component type is "beam-column joint reinforcement", the quantity is 11.91kg, the unit of measurement is kg, and the associated main component instance is COMP-C3-01.

[0088] Table 8 Examples of Linked Components with Quantities:

[0089] For the linked component example LINK-C4-KL6-01, the beam section height is 700mm, and the joint thickness coefficient is... mm, with a contact area of ​​0.210m² 2 The weight of the reinforcing steel is:

[0090] The system merges the main component COMP-C3-01 and the linked component instance LINK-C3-KL5-01 into a complete component assembly. Based on the project's location, "Province A," the applicable quota version is determined to be "Province A Construction Engineering Consumption Quota 20XX Edition." A search in the quota database reveals that the main component "Column Concrete Pouring C30" matches quota item number "5-128," with a comprehensive unit price of 385.60 yuan / m². 3 The linked component "beam-column joint reinforcement HRB400" matches the quota item number "5-267", with a comprehensive unit price of 6.85 yuan / kg.

[0091] Calculation of concrete quantity for main structural components:

[0092] Table 9: Detailed Cost Combination Scheme

[0093] The total amount of this cost combination plan is:

[0094] Throughout the implementation process, the data begins with the initial CAD drawing annotations, which are then extracted into a structured annotation dataset through OCR recognition and object detection (Tables 1 and 2). Entity recognition technology is used to parse the annotation text into spatial positioning attribute records for the components (Table 3), including key parameters such as axis numbers, elevations, and cross-sectional dimensions. Based on multi-view geometric projection rules, the two-dimensional annotation information is fused and converted into a three-dimensional positioning bounding box (Table 4), thus realizing the representation of the component's position in three-dimensional space.

[0095] By using bounding box intersection detection and intersection volume calculation, component pairs with spatial adjacency relationships are identified, and adjacency geometric parameters such as contact area are calculated to generate a component spatial adjacency relationship table (Table 5). The system extends the component hierarchy to define dependency rules for spatial adjacency triggering types (Table 6) and establishes a mapping relationship between adjacency relationships and linkage quotas.

[0096] During the item parsing phase, specific component instances are linked through the location information in the item characteristics, and all adjacent records of the component are retrieved (Table 7). Based on the triggering rules and adjacent geometric parameters, the quantities of the linked quota are calculated, generating linked component instances with quantities (Table 8). Finally, the main components and linked components are merged, and the quota items and comprehensive unit prices are matched to form a complete cost combination scheme (Table 9).

[0097] The entire data flow process realizes a complete closed loop from two-dimensional drawing annotation to three-dimensional spatial relationship inference, and then to automatic identification of linkage quota and engineering quantity calculation, solving the technical problem that linkage quotas that depend on spatial adjacency relationships cannot be automatically calculated in the absence of a BIM model.

[0098] Example 2: A method for generating cross-regional cost estimates based on a component system library In engineering cost estimation, there are significant differences in pricing quotas across the country in terms of the granularity of sub-items, the method of merging work content, and the rules for calculating resource consumption. The same construction work may be broken down into different numbers of quota sub-items in different regions, or it may be priced together, resulting in a lack of a unified correspondence between the tender bill of quantities items and the quota sub-items in various regions. The existing cost estimation methods mainly rely on cost estimators to manually match quotas based on their experience with quotas in specific regions. When the project involves cross-regional quota adaptation or when cost estimators are unfamiliar with the quota system of the target region, omissions or errors in quota application are prone to occur.

[0099] This embodiment provides a method for generating cross-regional cost estimates based on a component system library. By sorting out the differences in the sub-items and work content characteristics of the pricing quotas in various regions across the country, a unified component system library is constructed. The components and their influencing factors are used as an intermediate abstraction layer to realize the automatic parsing and matching of the bidding bill of quantities items to the quota sub-items in various regions.

[0100] The steps in this embodiment are as follows: Step S1: Standardized collection and preprocessing of quota data Acquire pricing quota data from all regions of the country, standardize the input of quota data according to a unified field structure, and generate a structured quota database.

[0101] It should be noted that the unified field structure includes a region identifier field, a professional category field, a quota chapter field, a sub-item code field, a sub-item name field, a work content field, a resource consumption field, and a unit of measurement field. The region identifier field is used to indicate the region to which the quota data belongs; the professional category field is used to indicate the engineering profession to which the quota belongs, including civil engineering, decoration, installation, and municipal engineering; the resource consumption field further includes three sub-fields: labor man-day consumption, material consumption, and machine-hour consumption.

[0102] Furthermore, the entered quota data undergoes standardized preprocessing, which includes the following three types of operations: The first type is terminology standardization, which converts regional expressions in quotas from different regions into industry standard terms, such as standardizing "concrete" to "concrete concrete", "pump pipe erection and dismantling" to "pump pipe installation and dismantling", and "cast-in-place concrete components" to "cast-in-place concrete components"; the second type is redundant information removal, which removes redundant modifiers in sub-item names and duplicate descriptions in work content; the third type is missing information completion, which automatically completes the missing information to the preset default value when a sub-item does not have a certain attribute information, in order to avoid interruption of the subsequent matching process.

[0103] It should be noted that the terminology unification operation is based on a pre-established terminology mapping table. The terminology mapping table records the mapping relationship between common non-standard terms in various regions and their corresponding industry standard terms. The data source for the terminology mapping table is the manual review of quota texts in various regions and industry standard specifications.

[0104] Step S2: Construct a component system based on the difference analysis of quota items Cross-regional difference analysis is performed on the quota items of various regions in the structured quota database. Based on the difference analysis results, the main components and sub-components are defined with the core construction components as the root nodes, the hierarchical relationship between components is established, and the component system is generated.

[0105] Step S2 includes the following sub-steps: Step S2.1: Using the same unit of measurement and the same core construction work as screening criteria, conduct cross-regional comparisons of quota items in the structured quota database to extract two types of difference characteristics. The first type of difference characteristic is the difference in item merging and splitting, that is, the same construction work is split into multiple independent pricing items in some regions, while it is merged into one item in other regions; the second type of difference characteristic is the difference in item subdivision dimensions, that is, the same construction work is subdivided according to different dimensions in different regions.

[0106] It should be noted that, taking concrete column pouring as an example, the first type of difference is manifested as follows: Beijing and Shanghai merge the concrete pumping work into the concrete column pouring sub-item, while Tianjin completely separates the concrete pumping fee into an independent quota sub-item; the second type of difference is manifested as follows: Tianjin subdivides the concrete pumping fee into three sub-items according to two dimensions: pump type and elevation: concrete pump, fixed pump below zero, and fixed pump above zero; while Shanghai subdivides it into two sub-items according to pump type: concrete pump truck and concrete pump.

[0107] Step S2.2: Define main components and sub-components based on differences in characteristics. A main component is defined as the core construction work to complete a certain building part or process, and each region has a corresponding core quota item, which is the pricing subject of that building part or process; a sub-component is defined as the auxiliary work required to complete the main component, and some regions split the sub-component into an independent pricing item while others merge the sub-component into the main component item, and the sub-component has an independent resource consumption.

[0108] It should be noted that the naming of main components adopts a unified naming format, which is a combination of construction process plus structural type or material plus core part plus other conditions. For example, in "cast-in-place concrete rectangular column", "cast-in-place" is the construction process, "concrete" is the material, "rectangular" is the structural type, and "column" is the core part. The purpose of the unified naming format is to eliminate the differences in the description of sub-items in different regions and establish standardized identification of components across regions.

[0109] Step S2.3: Establish the hierarchical relationship between main components and sub-components based on the difference analysis results. There are three types of hierarchical relationships: The first is an inclusion relationship, where the sub-component is necessary to complete the main component; without the work of this sub-component, the main component construction cannot be completed, and at least one region will include this sub-component in the main component's sub-item pricing. The second is an association relationship, where the sub-component is optional to complete the main component; its construction is selected based on the construction plan or design requirements, and each region treats this sub-component as an independent sub-item. The third is a mutual exclusion relationship, where different process sub-items under the same type of sub-component can only select one process during construction and cannot be implemented simultaneously.

[0110] It should be noted that the component system has only two levels of structure: the first level is the main component and the second level is the sub-component. No deeper levels are set to avoid redundancy in the component system.

[0111] It should be noted that the following is an example of a component system with cast-in-place concrete rectangular columns as the main component: The main component is a cast-in-place concrete rectangular column. The main component includes the concrete pumping cost of the sub-component. The concrete pumping cost of the sub-component includes mutually exclusive items such as pump, fixed pump below zero, fixed pump above zero, conveying pump truck, and conveying pump; the main component includes the installation, removal, and amortization of pump pipes for the sub-component. The installation, removal, and amortization of pump pipes for the sub-component includes mutually exclusive items such as vertical pump pipe installation and removal and horizontal pump pipe installation and removal; the main component includes the concrete pouring and vibration of the sub-component.

[0112] Step S3: Extract the influencing factors of components and establish an influencing factor database. For each component in the component system, key dimension conditions affecting quota application and pricing are extracted from the name of the quota sub-item, work content, and cross-regional split dimension. Standardized optional values ​​are determined for each dimension condition, and a component influencing factor library is generated.

[0113] Step S3 includes the following sub-steps: Step S3.1: For the main component, summarize the common characteristics of the basic sub-item names, work content, and measurement dimensions corresponding to all regions across the country. Extract the core dimensions that define the main component's attributes from these common characteristics, and use them as the influencing conditions for the main component. The influencing conditions for the main component are usually structural, material, shape, or part-related attributes, derived from the core modifiers in the sub-item names.

[0114] It should be noted that, taking cast-in-place concrete rectangular columns as an example, the common characteristics of the main component sub-items in various regions are cast-in-place process, concrete material, and rectangular shape. The main component influencing conditions extracted from this include: component type, with corresponding standardized options of cast-in-place components and precast components; component material, with corresponding standardized options of concrete and steel structure, etc.; and shape type, with corresponding standardized options of rectangular, circular, and irregular shapes, etc.

[0115] Step S3.2: For sub-components, compare the differences in sub-item subdivision dimensions and sub-item names across regions. Extract the subdivision dimensions for classifying the construction type of the sub-component from the regional sub-item subdivision criteria, and use these as influencing conditions for the sub-component. The influencing conditions for sub-components are usually construction process, construction parameter, or construction scope attributes, and originate from the key criteria for regional sub-item subdivision.

[0116] It should be noted that, taking the concrete pumping cost of sub-components as an example, the breakdown of sub-items varies by region: In Tianjin, it is divided into pump type (elephant pump and stationary pump) and elevation (below zero and above zero); in Shanghai, it is divided into pump truck and pump delivery. Therefore, the sub-component influencing conditions extracted include: pumping type, with corresponding standardized options of elephant pump, stationary pump, pump truck, and pump delivery; and pumping height, with corresponding standardized options of below zero and above zero.

[0117] Step S3.3: Perform integrity verification and standardization on all extracted influencing conditions. Integrity verification ensures that the optional values ​​for each influencing condition cover all sub-item types across the country without omission; standardization ensures that the optional values ​​under the same influencing condition are mutually exclusive and uniformly expressed, converting regional expressions into standardized terminology.

[0118] It should be noted that the specific operations of integrity verification include: verifying whether the optional values ​​of each influencing condition are mutually exclusive, that is, the same list item cannot select multiple optional values ​​at the same influencing condition; verifying whether the optional values ​​of each influencing condition cover all regional sub-items across the country, that is, each regional sub-item can find the corresponding optional value under the influencing condition; and verifying whether there are any ambiguous expressions, and unifying the ambiguous expressions into standardized terms.

[0119] Step S4: Component-based mapping and structured storage of quota sub-items The quota items of each region in the structured quota database are mapped and associated with the component system and the component influencing factor library. Each quota item is transformed into a structured representation of component identifier, influencing conditions and condition values, and stored in the component system library in a structured form.

[0120] Step S4 includes the following sub-steps: Step S4.1: For the cleaned quota sub-item data, map each quota sub-item to the corresponding component in the component system according to a three-level matching logic of keyword matching, work content verification, and rule priority determination. Keyword matching refers to extracting core keywords from the sub-item name and performing text matching with the component name in the component system; work content verification refers to comparing and verifying the work content description of the sub-item with the work content range of the candidate component; rule priority determination refers to determining the final matching result according to the rule that the main component takes precedence over the sub-component and exact matching takes precedence over fuzzy matching when a sub-item may match multiple components.

[0121] It should be noted that the matching operation outputs three types of results: the first type is the component level matching result, which determines whether the sub-item corresponds to the main component or the sub-component and associates the inclusion relationship between components; the second type is the influencing factor matching result, which matches all the influencing conditions corresponding to the sub-item; the third type is the optional value matching result, which matches a standardized optional value for each influencing condition.

[0122] Step S4.2: Based on the mapping results, the quotas for each region are decomposed and reconstructed according to the component dimension, breaking the original sub-item and chapter divisions of the region. The quota system of each region is reconstructed according to the hierarchy from main components to sub-components, generating a component-based quota mapping table. The data structure of the component-based quota mapping table includes a region identifier field, a main component identifier field, a sub-component identifier field, a corresponding sub-item code field, and an influencing factor attribute field.

[0123] It should be noted that, taking cast-in-place concrete rectangular columns as an example, some records in the component-based quota mapping table are as follows: In Beijing, the main component is a cast-in-place concrete rectangular column, and the sub-component is empty, meaning the sub-item is priced together. The corresponding sub-item code is 5-10, and the influencing factor attributes are: component type equals cast-in-place, component material equals concrete, and shape type equals rectangular. In Tianjin, the main component is a cast-in-place concrete rectangular column, and the sub-component is concrete pumping fee. The corresponding sub-item code is 14-3, and the influencing factor attribute is: pumping type equals elephant pump. In Shanghai, the main component is a cast-in-place concrete rectangular column, and the sub-component is concrete pumping fee. The corresponding sub-item code is 01-17-3-37, and the influencing factor attribute is: pumping type equals conveyor pump truck.

[0124] Step S4.3: Store the component system, component influencing factor library, and componentized quota mapping table in the component system library. The component system library uses a relational data structure for storage and includes four data tables: component definition table, influencing factor definition table, influencing factor optional value table, and quota mapping table. The data tables are linked through the component identifier field and the influencing factor identifier field.

[0125] It should be noted that the fields in the component definition table include component identifier, component name, component level, and parent component identifier. The component level can be either a main component or a sub-component, and the parent component identifier is used to record the main component to which the sub-component belongs. The fields in the influencing factor definition table include influencing factor identifier, influencing factor name, and associated component identifier. The fields in the influencing factor optional value table include optional value identifier, optional value name, and the identifier of the influencing factor to which it belongs. The fields in the quota mapping table include mapping record identifier, region identifier, component identifier, sub-item code, and combination of influencing factor attribute values.

[0126] Step S5: Analysis of Items in the Bill of Quantities for Tender Obtain the bill of quantities data for bidding projects, perform text parsing on each item in the bill of quantities based on the bill of quantities name and item characteristics, and parse each item into a structured result of the corresponding component and component influencing factors in the component system library, generating a bill of quantities parsing result set.

[0127] Step S5 includes the following sub-steps: Step S5.1: Construct a Bill of Quantities Parsing Knowledge Base. Extract keywords, material descriptions, and relationships between bill of quantities items from a large amount of historical bill of quantities sample data. Establish a mapping relationship between the extracted keywords and component names and optional influencing factor values ​​in the component system library to form a bill of quantities parsing knowledge base. The bill of quantities parsing knowledge base includes a keyword family tree table and a bill of quantities text feature pattern table. The keyword family tree table records the hierarchical mapping relationship between common keywords in the bill of quantities text and component identifiers and optional influencing factor values. The bill of quantities text feature pattern table records the text structure patterns of different types of bill of quantities items.

[0128] It should be noted that the keyword genealogy refers to organizing the keywords appearing in the list text into a tree-like hierarchical structure according to the component dimension and the attribute dimension. In the component dimension, the superior keyword corresponds to the main component, and the subordinate keyword corresponds to the child component; in the attribute dimension, the keyword corresponds to the optional values ​​of the influencing factors. For example, "column" is the superior keyword in the component dimension, and "rectangular column" and "circular column" are its subordinate keywords, which are mapped to the rectangular and circular optional values ​​of the shape type in the component influencing factors, respectively.

[0129] Step S5.2: Perform text parsing for each list item. The text parsing combines text semantic analysis with keyword matching from the list parsing knowledge base. First, extract keywords from the list name and match them with the keyword genealogy table in the list parsing knowledge base to identify the main component type corresponding to the list item. Then, perform semantic analysis on each descriptive information in the list item features to extract attribute information such as material, specifications, construction technology, and location. Match the extracted attribute information with the influencing conditions and optional values ​​in the component influencing factor library to determine the specific optional value corresponding to each influencing condition. Finally, based on the main component type, query the list of sub-components contained in the component system. Combine the descriptive information in the list item features to determine whether the work content of the sub-component is involved. If so, simultaneously identify the sub-component and its influencing factors.

[0130] It should be noted that, taking the list item "010502001001 Rectangular Column 1. Concrete Type: Ready-mixed 2. Concrete Strength Grade: C35 3. Includes Pumping 4. Location: Foundation Column" as an example, the text parsing process and results are as follows: The keyword "column" is matched from the list name "rectangular column" to determine that the main component is a column; "rectangular" is matched with the optional value "rectangular" for the shape type influencing condition; the optional value "ready-mixed" for the concrete type influencing condition is extracted from the first item of the item "Concrete Type: Ready-mixed"; the optional value "C35" for the concrete strength grade influencing condition is extracted from the second item of the item "Concrete Strength Grade: C35"; the concrete pumping cost of the sub-component involved in this list item is identified from the third item of the item "Includes Pumping"; and the optional value "foundation" for the contact position influencing condition is extracted from the fourth item of the item "Location: Foundation Column". The comprehensive analysis results are as follows: the main component is a column, the component type is cast-in-place, the component material is concrete, the shape type is rectangular, the concrete type is ready-mixed, the concrete strength grade is C35, and the contact position is the foundation; at the same time, the concrete pumping cost of the sub-component is identified.

[0131] Step S5.3: Perform a consistency check on the parsed results. The consistency check includes: verifying whether the parsed component identifier exists in the component definition table of the component system library; verifying whether the parsed optional values ​​of influencing factors belong to the legal optional values ​​in the corresponding influencing factor definition table; and verifying whether the subordinate relationship between the main component and the sub-component is consistent with the hierarchical relationship defined in the component system. For list items that fail the check, mark them as pending manual review.

[0132] Step S6: Matching the list items with regional quotas based on the component system library Based on the component identifier and influencing factor attribute value combination of each item in the list parsing result set, combined with the information of the current project's location and the pricing quota version adopted, the matching quota sub-item is retrieved in the quota mapping table of the component system library, and the matching result between the list item and the quota sub-item is generated.

[0133] Step S6 includes the following sub-steps: Step S6.1: Based on the regional identifier of the current project and the version of the pricing quota used, select a subset of quota mapping records for the target region from the quota mapping table of the component system library.

[0134] Step S6.2: For the parsing results of each list item, retrieve matching records in the target region quota mapping record subset using the main component identifier as the primary key. The matching process is as follows: compare the combination of main component influencing factor attribute values ​​parsed from the list item with the combination of influencing factor attribute values ​​in the quota mapping record condition by condition, select the quota mapping record with completely matching or most closely matching influencing factor attribute values, and obtain the corresponding main component quota sub-item code.

[0135] It should be noted that when multiple quota mapping records partially match the influencing factor attribute values ​​of the list item, an influencing factor weight scoring mechanism is used to determine the optimal matching result. This mechanism presets a weight value for each influencing condition, determined based on the degree of influence of that condition on quota selection. The weight values ​​of each matching condition are summed to obtain a matching score, and the quota mapping record with the highest matching score is selected as the final matching result.

[0136] Step S6.3: For the sub-components involved in the list item, retrieve the corresponding quota mapping record for the sub-component in the target region quota mapping record subset. When the target region splits the sub-component work into independent sub-items, match the independent quota sub-item code corresponding to the sub-component; when the target region merges the sub-component work into the main component sub-item, mark that the sub-component does not need to apply quotas independently in the target region.

[0137] Step S6.4: Summarize the matching results of the main component quota items and the sub-component quota items. For sub-components with mutual exclusion relationships, determine the selected mutual exclusion items based on the process description in the list item characteristics, exclude the unselected mutual exclusion items, and generate the complete quota matching results for each list item.

[0138] Step S7: Generate cost combination scheme Based on the matching results between the bill of quantities items and the quota sub-items, the comprehensive unit price and resource consumption data of each matching quota sub-item are obtained from the structured quota database. Combined with the engineering quantity of the bill of quantities items, the total price of each quota sub-item is calculated, and the cost combination scheme is generated.

[0139] It should be noted that the data structure of the cost combination scheme includes: the bill of quantities item code, the bill of quantities item name, the main component quota item code and item name, the comprehensive unit price, quantity and total price of the main component quota item, the quota item code and item name of each sub-component, the comprehensive unit price, quantity and total price of each sub-component quota item, and the total cost of the bill of quantities item.

[0140] In this embodiment of the application, in order to facilitate cost estimators in reviewing the rationality of the cost plan, the matching basis information of each quota item is marked in the cost combination plan. The matching basis information includes the component identifier on which the matching is based, the combination of influencing factor attribute values ​​and the matching score, so that cost estimators can intuitively understand the reason for the selection of each quota item and make manual adjustments when necessary.

[0141] In this embodiment of the application, when the same project needs to be adapted to the quotas of multiple regions, steps S6 and S7 are repeated for the same tender bill of quantities with different region identifiers as inputs, so as to generate cost combination schemes of the same bill of quantities under the quota system of different regions, thereby realizing the rapid generation and comparative analysis of cross-regional cost schemes.

Claims

1. A method for generating cost estimates driven by a component system, characterized in that, The process includes the following steps: acquiring engineering structural plan, elevation, and section data; performing annotation recognition on each drawing; extracting the text content, drawing coordinates, and view type of each annotation; and generating a drawing annotation dataset. Entity recognition is performed on the annotation text in the drawing annotation dataset to extract component type, positioning axis number, elevation value and cross-sectional size, and generate component spatial positioning attribute records; based on multi-view geometric projection rules, the coordinates of the plan view, elevation view and section view are fused and converted into three-dimensional spatial region estimation to generate three-dimensional positioning bounding boxes for each component; For component pairs with spatial intersections in the 3D positioning bounding boxes, adjacency relationship inference is performed. The adjacency type is determined according to the component type combination rules. The cross-sectional parameters are extracted from the associated annotation attributes to calculate the adjacency geometric parameters and generate a component spatial adjacency relationship table. Expand the data structure of inter-component relationships in the component system, and add dependency definitions for spatial adjacency trigger types; obtain list item data, parse the main components and influencing factors, and retrieve the adjacency records of main component instances in the component spatial adjacency relationship table; match the adjacency records with the dependency definitions of spatial adjacency trigger types, and substitute the adjacency geometric parameters into the dependency definitions that meet the trigger conditions to calculate the linked quota engineering quantity; merge the linked component instances with the main components, match the quota sub-items, and generate cost combination schemes.

2. The method according to claim 1, characterized in that, The annotation recognition process for each drawing includes: using an object detection neural network to locate and recognize text annotations, dimension annotations, and leader line annotations in the drawing; the input of the object detection neural network is the drawing image tensor, which generates candidate regions through a convolutional feature extraction layer and a region proposal network, and outputs the annotation bounding box coordinates, annotation category, and confidence score through a classification layer and a regression layer.

3. The method according to claim 1, characterized in that, The entity recognition process for annotation text in the drawing annotation dataset includes: performing the process using a combination of rule template matching and named entity recognition model; rule template matching extracts attributes based on the fixed format features of engineering drawing annotations; and the named entity recognition model is used to process annotation text in non-standard formats.

4. The method according to claim 3, characterized in that, The named entity recognition model adopts a sequence labeling architecture combining a bidirectional long short-term memory neural network and a conditional random field. The input of the named entity recognition model is a character sequence of labeled text. Each character is converted into a vector representation through a word embedding layer, input to a bidirectional long short-term memory neural network layer for context encoding, and then optimized for global sequence labeling through a conditional random field layer. The output is a sequence of entity labels corresponding to the character sequence.

5. The method according to claim 1, characterized in that, The process of generating the three-dimensional positioning bounding box for each component includes: obtaining the coordinates of the axis intersection points by querying the axis coordinate table based on the positioning axis number in the plan view; determining the X and Y coordinate ranges of the component in the horizontal plane by combining the relative relationship between the component's marked position and the axis intersection points; extracting the bottom and top elevation values ​​of the component based on the elevation markings in the elevation or section view to determine the Z coordinate range of the component in the vertical plane; and generating a three-dimensional axis-aligned bounding box that surrounds the component entity by expanding the corresponding dimension distance in each direction with the positioning reference point as the center based on the cross-sectional dimension parameters.

6. The method according to claim 5, characterized in that, The process of generating the three-dimensional positioning bounding box for each component also includes: cross-validating the positioning information from different views when the same component is marked in multiple drawings; when the horizontal positioning information of the same component appears in both the plan view and the section view, comparing the coordinate ranges obtained from the two analyses, and if the deviation exceeds a preset tolerance threshold, manual verification and marking are performed; if the deviation is within the preset tolerance threshold range, the intersection of the two is taken as the final coordinate range.

7. The method according to claim 1, characterized in that, The adjacency relationship inference for component pairs with spatial intersection of 3D bounding boxes includes: traversing all component pairs, performing intersection detection on the 3D bounding boxes of each pair, determining whether the projection intervals of the two 3D bounding boxes overlap in the X, Y, and Z directions, and filtering out candidate adjacent component pairs with spatial intersection; obtaining the component types of the two components in the candidate adjacent component pairs, querying a predefined component type combination rule table, and determining the adjacency type of the candidate adjacent component pairs; extracting relevant cross-sectional parameters from the spatial positioning attribute records of the adjacent component pairs according to the geometric calculation rules corresponding to the adjacency type, and calculating the contact area or intersection length; combining the component pair identifier, adjacency type, and adjacency geometric parameters into an adjacency record, and summarizing all adjacency records to generate a component spatial adjacency relationship table.

8. The method according to claim 7, characterized in that, The adjacency relationship inference for component pairs with spatial intersection of three-dimensional positioning bounding boxes further includes: setting a minimum intersection volume threshold, and determining that the corresponding component pair has an adjacency relationship only when the intersection volume of two three-dimensional positioning bounding boxes is greater than the minimum intersection volume threshold; the intersection volume is calculated based on the product of the overlap lengths of the two three-dimensional positioning bounding boxes in the X-axis, Y-axis and Z-axis directions, and the overlap lengths in each direction are non-negative values; The dependency definition of the spatial adjacency trigger type includes a main component type field, an adjacent component type field, a linked component type field, an adjacent geometric parameter reference identifier field, and a linked quota quantity calculation formula field. The step of substituting the dependency definition that meets the trigger condition into the adjacent geometric parameter to calculate the linked quota quantity includes: querying the component system based on the main component type to obtain the dependency definitions of all spatial adjacency trigger types corresponding to the main component type; traversing the adjacent records of the main component instance, matching the component type combination in each adjacent record with the dependency definitions of each spatial adjacency trigger type to determine the triggered dependency definition and the corresponding linked component type; obtaining the corresponding adjacent geometric parameter value from the matched adjacent record based on the adjacent geometric parameter reference identifier in the triggered dependency definition, substituting it into the linked quota quantity calculation formula to calculate the quantity value of the linked component; and generating a linked component instance with a quantity, wherein the linked component instance includes the linked component type, quantity value, unit of measurement, and associated main component instance identifier.

9. The method according to claim 1, characterized in that, Also includes: Step S1, Standardized collection and preprocessing of quota data: Input the unified field structure of pricing quotas for all regions of the country, and perform three types of preprocessing: standardization of terminology, redundancy removal, and missing data completion. Step S2: Construct a component system based on the difference analysis of quota items: compare the differences in merging / splitting of items across regions and the differences in subdivision dimensions, define the main components and sub-components, and establish three hierarchical relationships: inclusion, association, and mutual exclusion. Step S3, extract component influencing factors and establish an influencing factor library: extract influencing conditions from sub-item names and split dimensions, and determine standardized optional values; Step S4, Component mapping and structured storage of quota sub-items: Map quota sub-items of each region into structured representations and store them in the component system library; Step S5, Analysis of Tender Bill of Quantities Items: Construct a keyword genealogy knowledge base, and analyze the bill of quantities items into components and influencing factors through text semantic analysis and keyword matching; Step S6, matching the list and quota based on the component system library: using component identifier and influencing factor attribute values ​​as a bridge, retrieve the matching quota sub-item for the target region in the quota mapping table, including partial matching with a weighted scoring mechanism; Step S7, Generate Cost Combination Scheme: Summarize the matching results to generate a complete cost scheme, supporting cross-regional comparison.

10. A component system-driven cost estimation scheme generation system, used to execute the method according to any one of claims 1 to 9, characterized in that, include: The drawing annotation extraction module is used to acquire engineering structural plan, elevation and section data, perform annotation recognition on each drawing, extract the text content, drawing coordinates and view type of each annotation, and generate a drawing annotation dataset; The spatial positioning attribute extraction module is used to perform entity recognition on the annotation text in the drawing annotation dataset, extract component type, positioning axis number, elevation value and cross-sectional size, and generate component spatial positioning attribute records; the 3D bounding box generation module is used to fuse the coordinates of the plan view, elevation view and section view into a 3D spatial region estimate based on the multi-view geometric projection rules, and generate the 3D positioning bounding box of each component. The adjacency relationship inference module is used to perform adjacency relationship inference on component pairs with spatial intersection of 3D positioning bounding boxes. It determines the adjacency type according to the component type combination rules, calculates the adjacency geometric parameters, and generates a component spatial adjacency relationship table. The component system extension module is used to extend the data structure of relationships between components in the component system and add dependency definitions for spatial adjacency trigger types. The list parsing module is used to obtain list item data, parse it into main components and influencing factors, and retrieve the adjacency records of main component instances in the component space adjacency relationship table. The linkage engineering quantity calculation module is used to match the adjacency record with the dependency definition of the spatial adjacency trigger type, and substitute the dependency definition that meets the trigger condition into the adjacency geometric parameters to calculate the linkage quota engineering quantity. The cost estimate generation module is used to merge linked component instances with main components, match quota items, and generate cost combination schemes.

Images