Data-driven automated generation of building curtain wall node detail drawings

Abstract

This application provides a data-driven method for automatically generating detailed drawings of building curtain wall nodes, relating to the field of building curtain wall engineering technology. First, it acquires the curtain wall skin segment data and main structure configuration data of the target curtain wall project. Then, it performs geometric mapping and deviation analysis on the two data points to generate a spatial deviation vector field. Next, it calls a node detail generation network to process this spatial deviation vector field and node construction type markers to generate a node construction feature tensor. Finally, it performs geometric parametric decoding based on the node construction feature tensor to generate a set of node detail drawing instructions. This invention can efficiently and accurately generate curtain wall node details that meet the actual project requirements, improving the design and construction quality of building curtain wall engineering.

Description

Technical Field

[0001] This application relates to the field of building curtain wall engineering technology, and more specifically, to a data-driven method for automatically generating detailed drawings of building curtain wall nodes. Background Technology

[0002] In the field of building curtain wall engineering, curtain wall detail drawings meticulously depict the assembly relationships and connection structures between curtain wall components, playing a crucial role in ensuring the quality, safety, and aesthetics of the curtain wall. Traditionally, the generation of curtain wall detail drawings relied primarily on manual drawing by designers. Designers needed to complete the detail drawings step-by-step based on the specific requirements of the curtain wall project, combined with their own experience and professional knowledge. This method is not only inefficient but also easily affected by the designer's individual abilities and subjective factors, resulting in inconsistent quality of the detail drawings.

[0003] With the development of computer-aided design (CAD) technology, although some rule- and template-based tools for generating curtain wall detail drawings have emerged, these tools typically require the pre-definition of a large number of rules and templates and lack the ability to flexibly handle complex geometric relationships in actual projects. When faced with curtain wall skin segments and main structures of different shapes, sizes, and spatial locations, it is often difficult to accurately generate detail drawings that meet actual needs, thus failing to satisfy the diverse and personalized design requirements of modern building curtain wall projects. Summary of the Invention

[0004] In view of this, the purpose of this application is to provide a data-driven method for automatically generating detailed drawings of building curtain wall nodes.

[0005] According to a first aspect of this application, a data-driven method for automatically generating detailed drawings of building curtain walls is provided, the method comprising: Acquire curtain wall skin segment data and main structure configuration data of the target curtain wall project. The curtain wall skin segment data includes the boundary polyline sequence and segment identifier of each curtain wall skin segment, and the main structure configuration data includes the spatial coordinates of the main structure control points. Based on the curtain wall skin segment data and the main structure configuration data, geometric mapping and deviation analysis are performed on the curtain wall skin segments and the main structure to generate a spatial deviation vector field between the curtain wall skin segments and the main structure. The spatial deviation vector field includes the normal deviation vector and tangential deviation vector at each curtain wall skin segment boundary polyline sampling point relative to the main structure control point. The pre-built node detail diagram is called to generate a network that processes the spatial deviation vector field and the preset node construction type label to generate a node construction feature tensor at the connection node of the curtain wall skin segment. The node construction feature tensor includes node connector layout features, sealing construction features and adjustment space features. Based on the node construction feature tensor, geometric parameterization decoding is performed to generate a set of node detail drawing instructions at the connection nodes of the curtain wall skin segments. The set of node detail drawing instructions includes node area cutting plane indication, assembly baseline annotation, and component outline drawing instructions.

[0006] According to a second aspect of this application, a data-driven automated generation system for building curtain wall detail drawings is provided. The data-driven automated generation system for building curtain wall detail drawings includes a machine-readable storage medium and a processor. The machine-readable storage medium stores machine-executable instructions. When the processor executes the machine-executable instructions, the data-driven automated generation system for building curtain wall detail drawings implements the aforementioned data-driven automated generation method for building curtain wall detail drawings.

[0007] Based on any of the above aspects, the technical effect of this application is as follows: First, the system acquires the curtain wall skin segment data and main structure configuration data of the target curtain wall project. Through geometric mapping and deviation analysis of the curtain wall skin segments and the main structure, a spatial deviation vector field is generated. This accurately quantifies the geometric differences between the curtain wall skin segments and the main structure. A pre-built node detail drawing generation network is then used to process the spatial deviation vector field and preset node construction type labels, generating a node construction feature tensor. This tensor comprehensively considers various features such as node connector layout, sealing construction, and adjustment space, ensuring that the generated node details meet structural safety requirements while possessing good sealing performance and adjustability. Finally, based on the node construction feature tensor, geometric parametric decoding is performed to generate a set of node detail drawing instructions. This directly guides the drawing software in drawing node details, significantly improving the efficiency and quality of node detail drawing generation and reducing manual intervention and errors. Attached Figure Description

[0008] Figure 1 A flowchart illustrating the data-driven automated generation method for building curtain wall detail drawings provided in this application embodiment is shown. Figure 2 A schematic diagram of the component structure of the automated generation system for building curtain wall detail drawings provided in this application embodiment is shown. Detailed Implementation

[0009] Figure 1 This paper illustrates a flowchart of a data-driven automated generation method for building curtain wall detail drawings provided in an embodiment of this application. The detailed steps include: This embodiment uses the automated generation of detailed drawings for aluminum profile frame curtain wall nodes in a commercial complex building curtain wall project as an application scenario to provide a detailed description of the technical solution of the present invention. In this embodiment, the curtain wall skin segment data of the target curtain wall project comes from the curtain wall skin model in the Building Information Modeling (BIM), and the main structure configuration data comes from the coordinate set of structural control points extracted by registration from the 3D laser scan point cloud after the main structure construction is completed. The use of all design data has been authorized in writing by the project owner and the design party. The 3D scan data does not involve personal privacy, and the curtain wall component processing data is only used for the production organization of this project.

[0010] Step S110: Obtain the curtain wall skin segment data and main structure configuration data of the target curtain wall project. The curtain wall skin segment data includes the boundary polyline sequence and segment identifier of each curtain wall skin segment. The main structure configuration data includes the spatial coordinates of the main structure control points.

[0011] The curtain wall facade panel data is read from the curtain wall facade model via the Building Information Modeling (BIM) data exchange interface. The curtain wall facade panel data is a list of panel objects. Each panel object in the list contains a panel identifier field and a boundary polyline sequence field. The panel identifier field is a string type, with values ​​such as Panel001, Panel002, etc., representing globally unique identifiers. The boundary polyline sequence field is an ordered list of points, where each point represents a 3D spatial coordinate. Adjacent points are connected by straight line segments to form closed boundary polylines. The vertices of all boundary polyline sequences are arranged in the same direction of rotation.

[0012] The main structure configuration data is read through a 3D laser scanning point cloud data processing interface. The main structure configuration data is a list of control points, each containing a control point identifier field and a spatial coordinate field. The spatial coordinate field contains three real components: X, Y, and Z coordinates, measured in millimeters. The main structure control points cover the outer surfaces of the main structure beams and columns and the edge lines of the floor slabs within the curtain wall installation area.

[0013] Step S120: Based on the curtain wall skin segment data and the main structure configuration data, perform geometric mapping and deviation analysis on the curtain wall skin segments and the main structure to generate a spatial deviation vector field between the curtain wall skin segments and the main structure. The spatial deviation vector field includes the normal deviation vector and tangential deviation vector at the boundary polyline sampling point of each curtain wall skin segment relative to the control point of the main structure.

[0014] After acquiring the curtain wall skin segment data and the main structure configuration data, the spatial deviation distribution of each curtain wall skin segment relative to the main structure surface is calculated through geometric mapping.

[0015] Step S121: Spatially offset the boundary polyline sequence in the curtain wall skin segment data along the normal of the curtain wall skin segment to obtain the offset polyline sequence of the corresponding curtain wall skin segment. Each offset polyline segment of the offset polyline sequence forms a normal mapping correspondence with the original boundary polyline segment. The offset distance of the spatial offset processing is determined based on the theoretical spacing parameters between the curtain wall skin and the main structure preset in the curtain wall construction design.

[0016] For each boundary polyline sequence of the curtain wall skin segment in the curtain wall skin segment data, the normal direction of the segment is first calculated. The normal direction is calculated as follows: three non-collinear vertices are taken from the boundary polyline sequence to form two intersecting vectors. The cross product of these two vectors is then normalized to obtain the unit normal vector N. The theoretical spacing value Dtheory corresponding to the curtain wall skin segment is read from the curtain wall construction design parameter table. The theoretical spacing value is the design net distance from the inner surface of the curtain wall skin to the outer surface of the main structure. Each vertex Pi in the boundary polyline sequence is shifted along the normal direction N by a distance Dtheory to obtain the offset vertex Poffset_i = Pi + N * Dtheory. All offset vertices are connected in their original order to form an offset polyline sequence. The offset polyline sequence and the original boundary polyline sequence form a one-to-one correspondence of normal mappings; each pair of corresponding polyline segments constitutes a normal mapping pair.

[0017] Step S122: Perform spatial triangulation on the main structure control points in the main structure configuration data to generate a set of main structure surface triangular mesh units. Each main structure surface triangular mesh unit is composed of three main structure control points connected together. The spatial triangulation process constructs spatial topological connection relationships based on the Delaunay triangulation criteria between the main structure control points.

[0018] Using all control points in the main structure configuration data as the input point set, a 3D Delaunay triangulation is constructed. The Delaunay triangulation criterion is that for any triangular mesh element, its circumsphere does not contain any other control points. An incremental construction method based on the Bowyer-Watson algorithm is employed, sequentially inserting each control point into the current triangulation structure. For each newly inserted point, its corresponding triangular mesh element is located, and the element is split. The edges of the local region are then flipped according to the circumsphere criterion to restore the Delaunay property. After all control points are inserted, external triangular mesh elements crossing the main structure boundary are removed, retaining only the set of main structure surface triangular mesh elements covering the curtain wall installation area. Each main structure surface triangular mesh element consists of three control point indices and one element outward normal vector.

[0019] Step S123: Perform distance field calculation processing on each polyline sampling point on the offset polyline sequence and the set of triangular mesh elements on the main structure surface to determine the nearest projection point of each polyline sampling point on the triangular mesh element on the main structure surface. The distance field calculation processing is achieved by searching for the surface point with the smallest Euclidean distance to the polyline sampling point in the set of triangular mesh elements on the main structure surface.

[0020] The offset polyline sequence is sampled with equal arc lengths. The sampling interval is determined based on the total arc length of the polyline at the boundary of the curtain wall skin segments and the preset number of sampling points per meter. The total number of sampling points is denoted as M. Each sampling point is denoted as Qk, where k ranges from 1 to M. For each sampling point Qk, each triangular mesh element in the set of triangular mesh elements on the main structure surface is traversed, and the shortest distance from Qk to the surface of that triangular mesh element and the shortest distance point are calculated. The shortest distance point from Qk to a triangular mesh element is calculated as follows: Qk is projected onto the plane where the triangular mesh element is located to obtain the projection point Qproj. If Qproj is located inside the triangular mesh element, the shortest distance point is Qproj, and the shortest distance is the perpendicular distance from Qk to the plane. If Qproj is located outside the triangular mesh element, the shortest distance points from Qk to the three sides of the triangular mesh element are calculated respectively, and the smallest distance is selected. The shortest distance point with the smallest shortest distance is selected as the nearest projection point Rk of Qk among all triangular mesh elements.

[0021] Step S124: Calculate the normal deviation vector based on the spatial position difference between the polyline sampling point and the corresponding nearest projection point. The direction of the normal deviation vector is along the normal direction of the curtain wall skin segment, and the modulus is the distance from the polyline sampling point to the nearest projection point along the normal direction. The normal deviation vector represents the degree of spatial deviation of the curtain wall skin segment relative to the curved surface of the main structure in the direction perpendicular to the skin.

[0022] For each sampling point Qk and its nearest projected point Rk, calculate the spatial position difference vector V = Qk - Rk. The normal deviation vector Vnorm = (V·N)*N, where N is the unit normal vector of the curtain wall skin segment calculated in step S121, and · represents the vector dot product. The direction of the normal deviation vector Vnorm is parallel to N, and its magnitude |Vnorm| = |V·N| is in millimeters. A positive value indicates that the curtain wall skin segment deviates outward relative to the main structure, and a negative value indicates that it deviates inward.

[0023] Step S125: Calculate the tangential deviation vector based on the spatial position difference between the polyline sampling point and the corresponding nearest projection point. The direction of the tangential deviation vector is perpendicular to the normal deviation vector and is located in the tangential plane of the curtain wall skin segment. The modulus is the distance from the polyline sampling point to the nearest projection point along the tangential direction. The tangential deviation vector represents the degree of spatial slippage of the curtain wall skin segment relative to the curved surface of the main structure in the direction parallel to the skin.

[0024] Let V be the spatial position difference vector, and Vtangent deviation vector Vtang = V - Vnorm = V - (V·N)*N. The tangent deviation vector Vtang lies in the tangential plane of the curtain wall skin segment, and its direction points to the sliding direction of the curtain wall skin segment relative to the curved surface of the main structure. The modulus |Vtang| = sqrt(|V|^2 - (V·N)^2), with the dimension in millimeters.

[0025] Step S126: Assign the normal deviation vector and the tangential deviation vector to the corresponding polyline sampling points to form a spatial deviation vector tuple at each polyline sampling point of the curtain wall skin segment boundary. The spatial deviation vector tuple stores the normal deviation vector and the tangential deviation vector as deviation attributes of the polyline sampling points in the same data structure.

[0026] For each sampling point Qk, a spatial deviation vector tuple data structure is created. This structure contains four fields: the sampling point index field stores k, the normal deviation vector field stores the three-dimensional components of Vnorm, the tangential deviation vector field stores the three-dimensional components of Vtang, and the segment identifier field stores the segment identifier of the curtain wall skin segment to which the sampling point belongs. All spatial deviation vector tuples of sampling points are arranged in the order of their sampling point indices to form a spatial deviation vector tuple sequence.

[0027] Step S127: Perform continuity constraint processing on the spatial deviation vector tuple along the boundary polyline sequence to ensure that the normal deviation vector and tangential deviation vector of adjacent polyline sampling points satisfy the local smooth transition relationship, and generate a smooth deviation vector sequence. The continuity constraint processing uses a weighted moving average operator between the deviation vectors of adjacent sampling points to eliminate local noise disturbances.

[0028] Following the sampling point index order of the spatial bias vector tuple sequence, a weighted moving average is performed on the normal and tangential bias vectors of each sampling point. Centered on the k-th sampling point, 2r+1 sampling points are taken within a neighborhood window of radius r. The weights of each point within the window are allocated according to a Gaussian kernel function, with the standard deviation parameter σsmooth controlling the smoothing degree. The smoothed normal bias vector of the k-th sampling point is the Gaussian weighted average of the normal bias vectors of all points within the window; the tangential bias vector is processed similarly. After performing a moving average on all sampling points, a smoothed bias vector sequence is generated. Each element in the sequence remains a spatial bias vector tuple data structure, but the normal and tangential bias vectors have been replaced with smoothed values.

[0029] Step S128: Perform spatial indexing on the smoothed deviation vector sequence combined with the segment identifiers of the curtain wall skin segments, and construct a joint distribution set of deviation vectors and perform deviation clustering analysis.

[0030] Step S1281: Perform spatial indexing on the smooth deviation vector sequence and the segment identifiers of the curtain wall skin segments to generate a spatial deviation vector field with the curtain wall skin segments as index units. Each spatial deviation vector in the spatial deviation vector field corresponds to a polyline sampling point and contains a normal deviation vector and a tangential deviation vector. The spatial indexing process constructs a multi-level mapping relationship from the segment identifiers to the smooth deviation vector sequence.

[0031] A two-level index mapping structure is constructed. The key of the first-level index is the segment identifier of the curtain wall skin segment, and the value is a pointer to the storage address of the smoothed deviation vector sequence corresponding to that segment. The key of the second-level index is the sampling point index, and the value is the storage offset of the spatial deviation vector tuple corresponding to that sampling point. Through the two-level index, the corresponding spatial deviation vector can be quickly retrieved from the segment identifier and the sampling point index. The spatial deviation vectors of all segments are uniformly managed by the two-level index to form a spatial deviation vector field.

[0032] Step S1282: Extract the normal and tangential deviation vectors of all curtain wall skin segment boundary polyline sampling points in the spatial deviation vector field, construct a joint distribution set of deviation vectors for the entire project, and store all spatial deviation vectors of each curtain wall skin segment in the joint distribution set of deviation vectors according to the segment identifier.

[0033] Traverse the first-level index of the spatial deviation vector field, extracting all spatial deviation vectors under each segment identifier in turn. Create a grouped storage unit for each segment identifier, storing the normal deviation vector magnitude and tangential deviation vector magnitude of all spatial deviation vectors under that segment as two feature dimensions. The set of all grouped storage units constitutes the joint distribution set of deviation vectors.

[0034] Step S1283: Perform deviation clustering analysis on the joint distribution set of deviation vectors, divide the spatial deviation vectors into several deviation pattern categories according to the joint distribution characteristics of the normal deviation vector magnitude and the tangential deviation vector magnitude, and generate a set of node deviation pattern classification labels.

[0035] Deviation clustering analysis employs a Gaussian mixture model clustering algorithm. A two-dimensional feature vector is constructed by combining the normal and tangential magnitudes of all spatial deviation vectors in the deviation vector distribution set. The number of clusters, K_mode, is set, and the mean vector, covariance matrix, and mixture coefficient parameters of the Gaussian mixture model are iteratively optimized using an expectation-maximization algorithm until the log-likelihood function converges. Each cluster corresponds to a deviation pattern category, and the mean vector of that cluster is recorded as the deviation pattern prototype. Each spatial deviation vector is assigned a deviation pattern category label, and all category labels constitute a node deviation pattern classification label set.

[0036] Step S130: Call the pre-built node detail map to generate the network processing space deviation vector field and the preset node construction type label, and generate the node construction feature tensor at the connection node of the curtain wall skin segment. The node construction feature tensor includes the node connector layout features, sealing construction features and adjustment space features.

[0037] The pre-built node detail map generation network is a pre-trained deep neural network model. Its inputs are a spatial bias vector field and node construction type labels, and its output is a node construction feature tensor.

[0038] Step S131: Input the spatial deviation vector field into the spatial encoding module of the pre-constructed node detail map generation network for feature extraction processing. Construct the node neighborhood spatial deviation tensor through the normal deviation vector and tangential deviation vector in the spatial deviation vector field. The node neighborhood spatial deviation tensor aggregates the deviation vector information of multiple polyline sampling points within the neighborhood with the polyline sampling point where the node is located as the center.

[0039] The sampling points of the polyline where the connecting nodes are located are determined by the geometric feature points of the polyline of the curtain wall skin segment boundary, including the vertex, midpoint, and curvature extrema of the boundary polyline. For each connecting node, taking its own sampling point as the center, all sampling points within a preset neighborhood length are taken along the boundary polyline before and after it, and the total number of neighborhood sampling points is denoted as Snode. For each sampling point in the neighborhood, the three scalar components of its normal deviation vector and the three scalar components of its tangential deviation vector are extracted, for a total of six scalar channels, which are arranged in the order of the sampling points to form a node neighborhood spatial deviation matrix of shape Snode×6. This matrix is ​​then reshaped into a tensor of shape Snode×2×3 and batch processing dimension channels are added to obtain the node neighborhood spatial deviation tensor.

[0040] Step S132: Perform multi-offset rate hole depth separable convolution processing on the node neighborhood spatial deviation tensor to extract the multi-receptive field deviation feature map at the connection node of the curtain wall skin. The multi-offset rate hole depth separable convolution processing scans the node neighborhood spatial deviation tensor in parallel with depth separable convolution kernels with different hole ratios to capture the deviation distribution pattern in different spatial ranges.

[0041] The multi-offset dilated depth-separable convolution processing comprises three parallel branches with dilation ratios r1, r2, and r3, respectively. These dilation ratios are coprime and gradually increase. Each branch's depth-separable convolution processing is executed in two steps: first, a one-dimensional dilated convolution is performed independently on each feature channel of the input tensor. The kernel length is along the sampling point sequence direction, and the dilation ratio controls the sampling interval of each element in the kernel on the input sequence. The padding method uses the same size padding to maintain the sequence length. Then, the features output from each channel's dilated convolution are concatenated along the channel direction, and inter-channel information is fused using 1×1 pointwise convolution. The number of output channels for pointwise convolution is preset. The output features of the three parallel branches are concatenated along the channel direction to form a multi-receptive field bias feature map with a shape of Snode×Cmulti, where Cmulti is the sum of the number of output channels from the three branches.

[0042] Step S133: The preset node construction type markers are processed by the type embedding module of the node detail diagram generation network to generate node construction type embedding vectors. The vector embedding process maps the discrete node construction type markers to dense vector representations in a continuous vector space. The node construction type markers include beam connection node type, column connection node type and corner connection node type.

[0043] Node construction type tags are pre-annotated by the domain designer, and their values ​​are derived from an enumeration set: beam connection node type tag is coded as 1, column connection node type tag is coded as 2, and corner connection node type tag is coded as 3. The type embedding module is a trainable embedding lookup table, with the number of rows equal to the total number of enumerated values ​​for the node construction type tags, and the number of columns equal to the embedding dimension Demb. Using the node construction type tag code as the row index, the Demb-dimensional vector of the corresponding row is retrieved from the embedding lookup table as the node construction type embedding vector.

[0044] Step S134: Call the cross-attention fusion module of the node detail map generation network to perform feature fusion processing on the multi-receptive field bias feature map and the node construction type embedding vector. Use the node construction type embedding vector as the attention query tensor to perform weighted response extraction on the multi-receptive field bias feature map to generate a type-aware bias feature map. The weighted response extraction is achieved by calculating the similarity weight between the node construction type embedding vector and the spatial location features of the multi-receptive field bias feature map and performing a weighted aggregation operation.

[0045] The cross-attention fusion module employs a multi-head cross-attention mechanism. The node-constructed type embedding vector is mapped to a query vector Q via a linear transformation matrix WQ. The multi-receptive-field bias feature map is mapped to a key sequence K via a linear transformation matrix WK, and to a value sequence V via a linear transformation matrix WV. Attention weights are calculated by scaling and Softmax normalization through the dot product of Q and K, with the scaling factor being the square root of the embedding dimension Demb. Weighted aggregation multiplies the attention weights by V to obtain the context vector. This process is executed in parallel H times, with no shared parameters for the linear transformation matrices. The H output context vectors are concatenated and mapped to the final output vector via the output linear transformation matrix WO. This final output vector is then repeatedly broadcast along the spatial dimension of the multi-receptive-field bias feature map and concatenated with the original feature map along the channels to obtain the type-aware bias feature map.

[0046] Step S135: Using the node detail diagram to generate the connector sub-network, the sealing structure sub-network, and the adjustment space sub-network, respectively generate the node connector layout features, sealing structure features, and adjustment space features, and perform feature fusion and compression processing.

[0047] Step S1351: The connector generation sub-network of the node detail map generation network is used to process the type-aware bias feature map to generate connector layout patterns. The predicted connector geometric layout descriptor is used as the node connector layout feature. The connector geometric layout descriptor contains the connector position distribution heat map and the connector direction angle distribution information. The connector generation sub-network is implemented by a temporal layout decoding structure composed of multiple causal dilated convolutional layers stacked together.

[0048] The connector generation subnetwork comprises three cascaded causal dilated convolutional layers, with the dilation ratio of each layer increasing exponentially. The causal dilated convolution ensures that the output at the current position depends only on past sequence positions, and the convolution direction aligns with the forward direction of the sampling point along the boundary polygon. The output of the third causal dilated convolutional layer is mapped to a connector location heatmap using a sigmoid activation function. The heatmap outputs a probability value between 0 and 1 at each of the Snode sampling points, indicating the confidence level of a connector's presence at that location. Simultaneously, a fully connected layer outputs connector orientation angle distribution information, expressed in radians and distributed between 0 and π. The connector location heatmap and connector orientation angle distribution information together constitute the node connector layout features.

[0049] Step S1352: The sealing structure generation sub-network of the node detail map generation network is used to process the sealing structure parameters of the type-aware bias feature map, and the sealing structure parameter descriptor is predicted as the sealing structure feature. The sealing structure parameter descriptor includes the distribution of the sealing material filling thickness and the sealing installation method mark. The sealing structure generation sub-network gradually restores the spatial resolution of the sealing structure parameters through the deconvolution upsampling layer.

[0050] The sealing structure generation subnetwork comprises two one-dimensional deconvolutional upsampling layers. The first deconvolutional layer expands the spatial resolution of the feature sequence by an upsampling factor, and the second deconvolutional layer further expands it to the target resolution. The output channels of the two deconvolutional layers correspond to the sealing material filling thickness distribution and the sealing installation method label, respectively. The sealing material filling thickness distribution output is mapped to a positive real number sequence with millimeters as the unit of measurement using the Softplus activation function. The sealing installation method label output is mapped to a method probability distribution using the Softmax function, with three enumerated methods: adhesive strip sealing, adhesive strip plus sealant sealing, and structural adhesive sealing. The output of the sealing structure generation subnetwork constitutes the sealing structure features.

[0051] Step S1353: The adjustment space generation sub-network of the node detail map generation network is used to process the type-aware deviation feature map to generate adjustment space reserved parameters. The predicted adjustment space parameter descriptor is used as the adjustment space feature. The adjustment space parameter descriptor contains the three-dimensional adjustable space range limit and the adjustment screw arrangement spacing information. The adjustment space generation sub-network adopts the channel attention weighting mechanism to assign differentiated parameter generation weights to features in different deviation directions.

[0052] The adjustment space generation subnetwork first compresses the type-aware bias feature map into channel description vectors using a global average pooling layer. Then, a two-layer fully connected bottleneck structure generates channel attention weight vectors. The number of channels in the middle layer of the bottleneck structure is compressed to one-quarter of the original number of channels. The output layer restores the original number of channels and is activated by a sigmoid function to obtain the weight vector. This weight vector is multiplied channel-by-channel with the original type-aware bias feature map to obtain a weighted feature map. The weighted feature map is then passed through a fully connected layer to output the three-dimensional adjustable space range limits and the adjustment screw arrangement spacing. The three-dimensional adjustable space range limits are six scalars: minimum, maximum, minimum, maximum, minimum, and maximum values ​​in the X, Y, Z directions, and maximum values, with units of millimeters. The adjustment screw arrangement spacing is a single scalar with units of millimeters. The outputs of the adjustment space generation subnetwork constitute the adjustment space features.

[0053] Step S1354: Tensor splicing of node connector layout features, sealing structure features and adjustment space features along the feature channel direction to generate node structure feature tensor.

[0054] The node connector layout feature tensor, the sealing structure feature tensor, and the adjustment space feature tensor are concatenated end to end along the last dimension. The total number of channels of the concatenated tensor is equal to the sum of the number of channels of the three sub-feature tensors. The shape of the concatenated tensor remains unchanged, thus generating the node structure feature tensor.

[0055] Step S1355: Extract the statistical distribution characteristics of each channel feature component in the node construction feature tensor, calculate the mean and standard deviation of each channel feature component as channel normalization parameters, perform channel-by-channel normalization processing on the node construction feature tensor based on the channel normalization parameters to generate a normalized node construction feature tensor, perform nonlinear feature compression processing on the normalized node construction feature tensor, and map the normalized node construction feature tensor to a low-dimensional compact representation space through a fully connected bottleneck layer to generate a compact node construction feature representation.

[0056] For each feature channel of the node-constructed feature tensor, calculate the mean μc and standard deviation σc of all spatial location feature values ​​for that channel. The normalized feature value Fnorm = (F - μc) / (σc + ε), where ε is a smoothing constant. The fully connected bottleneck layer maps the normalized node-constructed feature tensor point-by-point to a low-dimensional compact representation space according to its spatial location. The input dimension of the bottleneck layer is equal to the total number of channels Ctotal of the node-constructed feature tensor, and the output dimension is the compressed dimension Ccompact = Ctotal / rcompress, where rcompress is the compression ratio. The compressed compact node-constructed feature representation is used to replace the original node-constructed feature tensor in the input of step S140.

[0057] Step S140: Perform geometric parameterization decoding based on the node construction feature tensor to generate a set of node detail drawing instructions at the connection nodes of the curtain wall skin segments. The set of node detail drawing instructions includes node area cutting plane indication, assembly baseline annotation and component outline drawing instructions.

[0058] The node construction feature tensor generated in step S130 is input into the geometric parameterization decoder. The geometric parameters of the node detail map are parsed out through three decoding branches and combined into drawing instructions.

[0059] Step S141: Input the node construction feature tensor into the connector decoding branch of the geometric parameterization decoder for parameter parsing processing, and decode the node connector layout features into a connector geometric parameter set. The connector geometric parameter set includes the connector cross-sectional contour coordinates and connector length dimensions. The connector decoding branch consists of a sequentially connected fully connected mapping layer and a coordinate regression output layer.

[0060] The fully connected mapping layer of the connector decoding branch contains two hidden layers. The first hidden layer has half the number of neurons as the input feature dimension, and the second hidden layer has half the number of neurons as the first hidden layer. The activation function is a modified linear unit with leakage. The coordinate regression output layer is a linear fully connected layer. The output dimension is the number of connector cross-sectional contour coordinates multiplied by the dimension of each coordinate point, plus the scalar dimension of the connector length. The connector cross-sectional contour coordinates are a sequence of vertices of a two-dimensional contour polygon on the assembly datum plane. The number of vertices in the contour polygon is preset, and each vertex contains two components: U-coordinate and V-coordinate. The connector length is a scalar of the stretched length along the assembly datum direction. The connector cross-sectional contour coordinates and the connector length together constitute the set of connector geometric parameters.

[0061] Step S142: Input the node construction feature tensor into the sealing construction decoding branch of the geometric parameterization decoder for parameter parsing processing, and decode the sealing construction feature into a set of sealing construction geometric parameters. The set of sealing construction geometric parameters includes the width of the sealant joint and the embedding depth of the sealing material. The sealing construction decoding branch performs stepwise parameter expansion of the sealing construction feature through a sequence of gated cyclic units.

[0062] The gated recurrent unit sequence of the sealing structure decoding branch contains multiple unfolding steps. Each unfolding step outputs a one-dimensional sealing structure parameter, including the sealant gap width and the sealant embedding depth. The initial hidden state of the gated recurrent unit is obtained by mapping the node construction feature tensor through a fully connected layer. The input of each unfolding step is a vector of the output of the previous unfolding step after linear transformation. The update gate and reset gate of this gated recurrent unit control the information retention and updating during the unfolding process. The sealing structure parameters output by all unfolding steps are combined to form a set of sealing structure geometric parameters, with the sealant gap width and sealant embedding depth measured in millimeters.

[0063] Step S143: Input the node construction feature tensor into the adjustment space decoding branch of the geometric parameterization decoder for parameter parsing processing, and decode the adjustment space features into a set of adjustment space geometric parameters. The set of adjustment space geometric parameters includes the range of the reserved gap in the adjustment space and the coordinates of the positioning point of the adjustment bolt. The adjustment space decoding branch uses a multi-head self-attention mechanism to capture the interdependencies between the parameters of each dimension of the adjustment space.

[0064] The multi-head self-attention mechanism in the adjustment space decoding branch maps the channel segments corresponding to the adjustment space features in the node's constructed feature tensor through query, key, and value linear transformation matrices, calculates the scaling dot product attention weights, and concatenates the results after multi-head parallel execution. The self-attention output is then output through a fully connected layer, showing the adjustment space reserved gap range and the coordinates of the adjustment bolt positioning points. The adjustment space reserved gap range includes the minimum and maximum gap values, measured in millimeters. The coordinates of the adjustment bolt positioning points are the three-dimensional coordinates of the adjustment bolts in the node's local coordinate system, with a preset number, and each coordinate point contains three scalar components.

[0065] Step S144: Extract the cross-sectional contour coordinates of the connector from the set of geometric parameters of the connector, and generate the connector contour drawing instruction. The connector contour drawing instruction includes a description of the outer contour of the connector with reference to the assembly datum line. The description of the outer contour of the connector is expressed by a closed polygon path formed by connecting the cross-sectional contour coordinates of the connector in sequence.

[0066] The UV 2D coordinate sequence of the connector section outline coordinates is connected end-to-end in sequence to form a closed polygon path description string. The path description adopts the standard path format of computer graphics, with the M command marking the path start point and the L command marking the connection of straight line segments. The connector outline drawing command includes a path description string, a linetype identifier field, and a linewidth field. The linetype identifier field is set to solid line, and the linewidth field value corresponds to the standard outline linewidth at the drawing output scale.

[0067] Step S145: Extract the width of the sealant joint and the embedding depth of the sealing material from the set of geometric parameters of the sealing structure, and generate a drawing instruction for the sealing structure contour. The drawing instruction for the sealing structure contour includes a graphic description of the sealant joint boundary contour and a graphic description of the sealing material filling area. The graphic description of the sealant joint boundary contour is defined by three geometric elements: the coordinates of the sealant joint start point, the coordinates of the sealant joint end point, and the width of the sealant joint.

[0068] The graphic description of the sealant joint boundary outline is a rectangular area outline formed by extending outwards from the centerline of a straight line segment determined by the coordinates of the start and end points of the sealant joint, with the width of the sealant joint as the offset width. The graphic description of the sealant filling area is based on the aforementioned rectangular area outline, with the sealant embedding depth as an additional attribute for the filling depth annotation. The sealant construction outline drawing command includes a rectangular area outline path description, a fill pattern identifier field, and a fill scale field.

[0069] Step S146: Extract the spatial location of the polyline sampling points in the connected node region in the spatial deviation vector field to construct the node cutting plane equation and generate a cutting plane indication command. The cutting plane indication command is used to determine the cutting direction and cutting position of the node detail view. The node cutting plane equation takes the mean direction of the normal deviation vector of the polyline sampling points in the connected node region as the plane normal vector and the geometric center of the polyline sampling points as the point through which the plane passes.

[0070] For each of the Snode polyline sampling points within the neighborhood of the connecting node, the vector mean of all normal deviation vectors is calculated to obtain the principal normal vector. The geometric center point is obtained by calculating the arithmetic mean of the spatial coordinates of all polyline sampling points. The equation of the cutting plane is a spatial plane with the geometric center point as the point through which the plane passes and the principal normal vector as the plane's normal vector. The cutting plane indication command includes two parameters: the coordinates of the point through which the cutting plane passes and the direction of the cutting plane's normal vector.

[0071] Step S147: Using the mean direction of the normal deviation vector of multiple polyline sampling points within the connected node area as the direction of the assembly datum line in the node detail drawing, and using the geometric center of the polyline sampling points within the connected node area as the origin of the assembly datum line, generate an assembly datum line annotation instruction. The assembly datum line annotation instruction includes the coordinates of the datum line start point, the datum line direction vector, and the datum line symbol.

[0072] The coordinates of the baseline starting point are the coordinates of the geometric center point in step S146. The baseline direction vector is the principal normal vector in step S146. The baseline identifier is a string, formatted as node identifier plus baseline suffix. The assembly baseline annotation command includes the above three parameters, as well as the annotation text height field and annotation line type field.

[0073] Step S1481: Integrate the connector outline drawing instruction, the sealing structure outline drawing instruction, the section plane indication instruction, and the assembly baseline annotation instruction to generate a set of node detail drawing instructions. The instructions in the node detail drawing instruction set are arranged into a structured instruction sequence according to the view generation order.

[0074] The connector outline drawing instructions generated in step S144, the sealing structure outline drawing instructions generated in step S145, the section plane indication instructions generated in step S146, and the assembly datum line annotation instructions generated in step S147 are aggregated into an instruction list container. The initial order of the instructions in the instruction list container is: section plane indication instructions first, assembly datum line annotation instructions second, sealing structure outline drawing instructions third, and connector outline drawing instructions last. This initial order corresponds to the actual drawing layer order of the node detail drawing: first determine the viewing angle and reference datum, then draw the sealing fill area, and finally draw the main outline of the connector. The instruction list container is the set of node detail drawing instructions.

[0075] Step S1482: Analyze the coordinate system of each drawing instruction in the node detail drawing instruction set, and transform the local coordinates in the connector contour drawing instruction and the sealing structure contour drawing instruction to the node local coordinate system with the assembly datum line as the reference, and generate the node detail drawing instruction set of coordinate system one.

[0076] The connector profile drawing command's connector section contour coordinates are originally defined in the connector's local plane coordinate system, with the geometric center of the connector section as the origin. The sealing structure profile drawing command's sealant joint start and end coordinates are originally defined in the sealing structure's local coordinate system, with the midpoint of the sealant joint as the origin. The assembly datum line annotation command defines a node local coordinate system with the assembly datum line origin as the origin, the assembly datum line direction vector as the Z-axis, and the cutting plane normal vector as the X-axis.

[0077] The coordinate system transformation process is as follows: First, calculate the homogeneous transformation matrix Tcon from the connector local coordinate system to the node local coordinate system. Multiply the connector section contour coordinates by Tcon to obtain the coordinates in the node local coordinate system. Then, calculate the homogeneous transformation matrix Tseal from the sealing structure local coordinate system to the node local coordinate system. Multiply the starting and ending coordinates of the sealing joint by Tseal to obtain the coordinates in the node local coordinate system. The transformed connector contour drawing instructions and sealing structure contour drawing instructions, together with the original assembly baseline annotation instructions and cutting plane indication instructions, constitute the node detail drawing instruction set for coordinate system one.

[0078] Step S1483: Perform drawing priority sorting on the set of node detail drawing instructions for coordinate system one, and arrange the instruction execution sequence according to the order of cutting plane indication instruction, assembly datum line annotation instruction, sealing structure contour drawing instruction and connector contour drawing instruction to generate an ordered set of node detail drawing instructions.

[0079] The drawing priority sorting process assigns an integer priority value to each instruction in the nodal detail drawing instruction set of coordinate system one. The cutting plane indication instruction has the highest priority, the assembly baseline annotation instruction has the second highest priority, the sealing structure contour drawing instruction has the middle priority, and the connector contour drawing instruction has the lowest priority. Instructions with higher priority values ​​appear earlier in the instruction execution sequence. When multiple instructions have the same priority value, they are arranged in ascending order according to their internal number within their instruction type. The resulting instruction list after sorting is the ordered nodal detail drawing instruction set. The order of instructions in the ordered nodal detail drawing instruction set ensures the correct overlay relationship during drawing output: reference information precedes graphic information, and the background layer precedes the foreground layer.

[0080] Step S210: Extract the adjustment space features of all curtain wall skin segment connection nodes from the node construction feature tensor, perform statistical analysis on the adjustment space features of all connection nodes, determine the distribution range of adjustable gap values ​​in the adjustment space features, and determine the distribution boundary by calculating the mean and extreme values ​​of the adjustable gap values ​​of all connection nodes.

[0081] Extract the three-dimensional adjustable space range boundary from the adjustment space features of each connection node generated in step S1353. For each connection node, take the difference between the maximum and minimum values ​​in the Z direction as the adjustable gap value of that connection node. Statistically calculate the adjustable gap values ​​of all connection nodes in the entire project, and calculate the mean μgap and standard deviation σgap. The lower bound of the adjustable gap value distribution range is the maximum value between μgap minus a preset multiple of the standard deviation σgap and zero, and the upper bound is μgap plus a preset multiple of the standard deviation σgap. Record this distribution range boundary value as the adjustable gap value range for the entire project.

[0082] Step S220: Generate an adaptive node grouping scheme based on the adjustable gap value distribution range. The adaptive node grouping scheme is used to guide the graded processing of components when prefabricating node components in the factory. The adaptive node grouping scheme divides several processing batch groups according to different intervals of the adjustable gap value distribution range.

[0083] The adjustable gap range of the entire project is divided into Ngap intervals based on the principle of equal width, with Ngap being the preset number of groups. Each interval corresponds to a processing batch group, and the start and end values ​​of the interval are recorded as the adjustable gap range of that batch group. All connection nodes in the entire project are traversed, and each connection node is assigned to its corresponding processing batch group based on the interval its adjustable gap value falls into. Each processing batch group includes a group number, an adjustable gap range, and a list of curtain wall skin segment identifiers for the connection nodes within that group. All processing batch groups constitute an adaptive node grouping scheme.

[0084] Step S230: Output the group node prefabrication identification file based on the adaptive node grouping scheme. The group node prefabrication identification file is used to associate the preparation of node components with the same processing parameters in batches. The group node prefabrication identification file includes the group number, the adjustable gap range, and the list of curtain wall skin segment identifications.

[0085] The adaptive node grouping scheme is serialized into a structured text file. Each line in the file corresponds to a processing batch group and contains the following fields: the group number field is a string type, the adjustable gap range field contains the minimum and maximum values ​​separated by a delimiter, and the list of curtain wall skin segment identifiers is separated by a delimiter. This structured text file is written to a storage device as the prefabrication processing identifier file for the grouped nodes.

[0086] Step S240: Extract the node connector layout features of each connected node in the node construction feature tensor, summarize the connector geometric layout descriptors of each connected node corresponding to the prefabrication identification file of the same group of nodes into the group connector layout summary information, generate the group connector layout scheme based on the group connector layout summary information, and arrange the cross-sectional contours of each connector processed in the same batch on the raw material plate to optimize material utilization.

[0087] Extract the connector cross-sectional contour coordinates of each connection node from the node connector layout features generated in step S1351. According to the grouping of nodes in the prefabrication identification file, summarize the connector cross-sectional contour coordinates of all connection nodes within the same group into a grouped connector layout summary information. The grouped connector layout scheme adopts a two-dimensional irregular layout algorithm based on critical polygons. All connector cross-sectional contours within the same group are considered as parts to be laid out, and the raw material sheet is defined as an infinitely long strip of fixed width. The objective function is to minimize the length of the sheet used. The layout algorithm determines the densest non-overlapping arrangement position between parts by calculating the critical polygons between each part, and uses a left-lower-corner priority placement strategy to place parts sequentially on the sheet until all parts are placed. The generated grouped connector layout scheme includes the placement coordinates and rotation angle of each part on the sheet.

[0088] Step S250: Generate CNC stamping blanking program instructions based on the grouped connector layout scheme. The CNC stamping blanking program instructions are used to drive the sheet metal stamping equipment to perform batch blanking operations of connectors according to the layout path.

[0089] The placement coordinates and rotation angles of each connector in the grouped connector layout scheme are converted into a toolpath coordinate sequence for the CNC stamping equipment. The toolpath includes rapid traverse commands for the idle stroke segment and cutting feed commands for the blanking stroke segment. The cutting sequence is optimized using a greedy algorithm based on the shortest path principle. Each toolpath segment is output using standard CNC machine tool instruction codes, including G-codes, coordinate words, and feed rate words. The entire program begins with a program number and ends with a program end command. A complete CNC stamping blanking program instruction text file is generated.

[0090] Step S260: Extract the connector outline drawing instructions corresponding to each connection node from the node detail drawing instruction set, and convert the connector outline drawing instructions into connector material codes and quantity statistics records.

[0091] Extract the connector cross-sectional contour coordinates from the connector contour drawing instruction generated in step S144, and calculate the width and height of the bounding rectangle of the cross-section as the specification description. Simultaneously, extract the connector length dimension. Retrieve the matching material code from the preset connector material code table based on the specification description and length dimension. The material code table uses the specification range and length range as a joint lookup key, returning the material code string. For each connector node, generate a connector material code and quantity statistics record, with the record fields including the connector node identifier, material code, and quantity value.

[0092] Step S270: Aggregate the material codes and quantity statistics of connectors by group according to the prefabrication identification file of the group node, generate batch material purchase orders, generate component processing schedule table according to the batch material purchase orders and the prefabrication identification file of the group node, and arrange the batch processing tasks according to the expected delivery cycle in the component processing schedule table.

[0093] The connector material codes and quantity statistics records generated in step S260 are grouped and aggregated according to group number. The quantities of the same material code within the same group are accumulated to generate a batch material purchase order. Each record in the order contains a material code, a total quantity value, and the corresponding processing batch group number.

[0094] The component processing schedule is arranged in ascending order by batch group number, with each task assigned an estimated start and end time. The time interval between adjacent batches is determined by the processing time quota in the estimated delivery cycle and the mold change preparation time between batches. Each row in the schedule includes the batch group number, task description, estimated start time, estimated end time, and the preceding batch group number.

[0095] Step S280: Link the component processing schedule table with the group node prefabrication identification file to generate a factory production instruction package containing processing batch identification, bill of materials and delivery time nodes.

[0096] The group numbers of the prefabrication identification files for group nodes are associated one-to-one with the batch group numbers in the component processing schedule table, and then merged to generate a factory production instruction package data structure. This factory production instruction package data structure contains three top-level fields: the processing batch identifier field, which is the group number; the bill of materials field, which is a subset of records for that group in the batched material purchase order; and the delivery time node field, which is the estimated completion time for that group in the component processing schedule table. All group factory production instruction package data structures are serialized into a unified structured file for output.

[0097] Step S310: Extract the assembly baseline annotation instructions for all connected nodes in the node detail drawing instruction set, and construct the assembly baseline network topology diagram for the entire project. The assembly baseline network topology diagram for the entire project forms a spatial grid structure with the assembly baselines of each connected node as topological edges and the connected nodes as topological vertices.

[0098] Extract the baseline start coordinates and baseline direction vector of each connected node from the assembly baseline annotation instruction generated in step S147. Each connected node is used as a vertex in the topology graph, and each vertex contains a vertex identifier and vertex coordinate attributes. When the spatial distance between the baseline start coordinates of two connected nodes is less than a preset adjacency distance threshold, and the angle between the baseline direction vectors of the two nodes is less than a preset adjacency angle threshold, a topology edge is established between the two vertices. The direction of the topology edge is consistent with the assembly propagation direction of the two connected nodes. All vertices and edges constitute the assembly baseline network topology graph for the entire project.

[0099] Step S320: Perform assembly dimension analysis on the assembly baseline network topology diagram of the entire project to identify key assembly control nodes and assembly propagation paths in the assembly baseline network. The assembly propagation path represents the cumulative propagation direction of assembly errors along the baseline network.

[0100] Assembly dimension analysis is performed on the assembly baseline network topology of the entire project. The degree centrality and eigenvector centrality of each vertex are calculated. Degree centrality is the number of topological edges directly connected to the vertex. Eigenvector centrality is obtained by recursively weighting the eigenvectors connected to vertices with high centrality. Vertices with both degree centrality and eigenvector centrality exceeding a preset threshold are marked as critical assembly control nodes. Starting from the boundary nodes of the assembly baseline network topology, error propagation paths are traced layer by layer along the topological edges, identifying multiple assembly propagation paths from the starting boundary node to the ending boundary node. Each assembly propagation path is an ordered sequence of vertices, and the direction of the topological edges between adjacent vertices represents the error propagation direction.

[0101] Step S330: Generate a phased assembly measurement control scheme based on the assembly propagation path. The phased assembly measurement control scheme sets up on-site measurement control points at key nodes of the assembly propagation path to control the range of error accumulation.

[0102] Along each assembly propagation path, key assembly control nodes are used as the locations for measurement control points. Starting from the beginning of the assembly propagation path, a field measurement control point is set at every predetermined number of connection nodes. When a key assembly control node is encountered, a measurement control point is forcibly set at that node. Each field measurement control point records its spatial coordinates and the corresponding connection node identifier. The union of all field measurement control points along the assembly propagation path constitutes a phased assembly measurement control scheme. This phased assembly measurement control scheme divides the assembly process of all connection nodes in the project into multiple assembly flow segments with field measurement control points as stage boundaries.

[0103] Step S340: Based on the phased assembly measurement control scheme, output the on-site assembly measurement control point layout drawing, and mark the spatial coordinates and allowable deviation range of each measurement control point on the on-site assembly measurement control point layout drawing.

[0104] Using the unfolded facade drawing of the curtain wall building as the base map, the projected position of each on-site measurement control point is marked on the base map with measurement control point symbols. A label line is drawn next to each measurement control point symbol, containing the measurement control point number, three-dimensional spatial coordinates, and allowable deviation range. The allowable deviation range is a preset scale of the adjustable gap value of the corresponding connection node for that measurement control point. All labeling information is combined with the base map to output an on-site assembly measurement control point layout drawing.

[0105] Step S350: Extract the section plane indication instructions of each connecting node from the node detail drawing instruction set, determine the node detail drawing observation view direction of each connecting node, and perform joint analysis with the assembly baseline direction of the corresponding connecting node to generate a three-dimensional visualization assembly simulation scene description file. The three-dimensional visualization assembly simulation scene description file is used to display the assembly process of each connecting node in the actual assembly sequence in the visualization terminal.

[0106] The cutting plane normal vector is extracted from the cutting plane indication command generated in step S146 as the viewing angle direction of the node detail drawing. The viewing angle direction of the node detail drawing for each connected node, the starting coordinates and direction vector of the assembly baseline generated in step S147, the connector contour drawing command generated in step S144, and the sealing structure contour drawing command generated in step S145 are jointly encapsulated into an assembly scene description record. All assembly scene description records are arranged according to the assembly propagation path determined in step S330 to generate a 3D visualization assembly simulation scene description file. This 3D visualization assembly simulation scene description file adopts an exchangeable 3D scene description format and includes the camera position, viewing angle direction, 3D geometric information, and assembly order index for each scene record.

[0107] Step S360: Drive the display device to play the assembly simulation screen according to the 3D visualization assembly simulation scene description file, generate the assembly animation key frame sequence, and synchronize the assembly animation key frame sequence with the phased assembly measurement control scheme in time to generate an assembly simulation demonstration file with measurement node annotations.

[0108] Using the assembly scene description record sequence as the timeline, linear interpolation-generated transition frames are inserted between every two adjacent scenes to form a keyframe sequence for the assembly animation. Within this keyframe sequence, the locations of the on-site measurement control points in the phased assembly measurement control scheme are corresponding to the positions in the assembly scene description record. A measurement control point annotation layer is overlaid on this keyframe image. The annotation layer highlights the locations of the measurement control points with a highlighted border and labels them with their numbers and allowable deviation values. The keyframe sequence with overlaid annotations is then packaged together with the original 3D scene into an assembly simulation demonstration file.

[0109] Step S370: Extract the connection node identifiers corresponding to each keyframe in the assembly simulation demonstration file with measurement node annotations, generate an assembly process card sequence, sort the assembly process card sequence according to the order of the assembly propagation path, and output the on-site construction guidance manual. In the on-site construction guidance manual, each assembly process card corresponds to the installation instructions and measurement requirements of a connection node.

[0110] Traverse the keyframe sequence of the assembly simulation demonstration file with measurement node annotations, extracting the corresponding connection node identifier for each frame. Generate an assembly process card for each connection node, including: connection node identifier number, node construction type mark, connector outline graphic, seal construction outline graphic, assembly baseline direction indicator arrow, adjustable gap value, measurement control point mark, and installation precautions text. Sort all assembly process cards according to the assembly propagation path sequence, add a cover and table of contents, and output a portable document format on-site construction guidance manual.

[0111] Step S410: Extract the three-dimensional adjustable space range limit and adjustment screw arrangement spacing information from the node construction feature tensor of all curtain wall skin segment connection nodes, and construct a full project adjustment space database. The full project adjustment space database stores the adjustable gap value and screw arrangement parameters of each connection node with the connection node identifier as the primary key.

[0112] Traverse all curtain wall skin segment connection nodes, and extract the three-dimensional adjustable space range boundary and adjustment screw arrangement spacing of each connection node from the adjustable space features generated in step S1353. Calculate the difference between the maximum and minimum Z-direction values ​​within the three-dimensional adjustable space range boundary as the adjustable gap value for that connection node. Using the connection node identifier as the primary key and the adjustable gap value and adjustment screw arrangement spacing as attribute columns, construct a relational storage structure as the entire project's adjustable space database.

[0113] Step S420: Sort the adjustable gap values ​​of each connection node to construct a full project node gap distribution sequence. The full project node gap distribution sequence arranges the adjustable gap records of all connection nodes in ascending order of adjustable gap values.

[0114] The adjustable gap values ​​of all connection nodes are extracted from the project's overall adjustment space database to generate a gap value list. This gap value list is then sorted in ascending order, resulting in a sequence of connection node identifiers arranged from smallest to largest adjustable gap value. Each element in this sequence contains both the connection node identifier and the adjustable gap value. This ordered sequence represents the overall project node gap distribution sequence.

[0115] Step S430: Generate an on-site installation sequence plan based on the distribution sequence of node gaps throughout the project. The on-site installation sequence plan arranges the assembly order of the connecting nodes in descending order of gap margin, with the connecting nodes with decreasing gap margins being installed on-site before the connecting nodes with increasing gap margins.

[0116] The project's node gap distribution sequence is arranged from smallest to largest adjustable gap. Smaller adjustable gaps indicate less installation adjustment margin and more stringent precision requirements, and should be installed first to utilize the gap compensation capability of subsequent nodes. Therefore, the on-site installation sequence directly adopts the order of the project's node gap distribution sequence as the assembly order of the connecting nodes, i.e., a progressively decreasing gap margin order. The assembly order of each connecting node is its position index in the sequence plus 1.

[0117] Step S440: Output an installation sequence distribution diagram based on the on-site installation sequence plan. The installation sequence distribution diagram is marked with color gradients on the curtain wall skin segment unfolded diagram to indicate the assembly sequence number of each connection node.

[0118] Using the unfolded diagram of the curtain wall facade as the base map, each connection node is marked with a colored dot on the projection position on the base map. The color of the dot is determined by a coding mapping based on the assembly sequence of that node. The earlier the assembly sequence, the more the color leans towards the cooler end; the later the sequence, the more the color leans towards the warmer end. The color gradation uses a preset gradient linear mapping. The number next to the dot is the assembly sequence number. The base map, dots, and numerical annotations together constitute a schematic diagram of the installation sequence distribution.

[0119] Step S450: Extract the segment identifiers of the curtain wall skin segments corresponding to each connection node, group the assembly sequence numbers of each connection node in the installation sequence distribution diagram according to the segment identifiers, generate a segment assembly sequence summary table, generate a curtain wall skin segment hoisting sequence plan based on the segment assembly sequence summary table, and arrange the segment hoisting sequence according to the earliest assembled connection node sequence number in each segment.

[0120] Traverse all connection nodes and group them according to the segment identifiers of the curtain wall skin sections. Within each group, the connection nodes are arranged in ascending order of their assembly sequence numbers. Each row of the section assembly sequence summary table records the segment identifier of a curtain wall skin section, the number of connection nodes within that section, the assembly sequence number of the earliest assembled connection node within that section, and the assembly sequence number of the latest assembled connection node within that section. Sort the section assembly sequence summary table in ascending order by the earliest assembly sequence number; the sorted result is the curtain wall skin section hoisting sequence plan. This curtain wall skin section hoisting sequence plan specifies the hoisting order of the curtain wall skin sections, ensuring that the sections requiring connection node installation first are hoisted into place with priority.

[0121] Step S460: Generate a comprehensive installation schedule network plan based on the curtain wall skin segment hoisting sequence plan and the on-site installation sequence plan. The comprehensive installation schedule network plan expresses the temporal dependency relationship between segment hoisting operations and node assembly operations as a directed acyclic graph structure.

[0122] A set of schedule activity nodes is constructed, with each curtain wall cladding segment hoisting operation as a hoisting activity node and each connecting node assembly operation as an assembly activity node. The rules for constructing directed edges are as follows: hoisting activity nodes point to the assembly activity node with the smallest assembly sequence number within that segment; within the same segment, assembly activity nodes with smaller assembly sequence numbers point to assembly activity nodes with larger sequence numbers; between different segments, if an assembly activity node in a later hoisting segment depends on the completion of an assembly activity node in an earlier hoisting segment, then a directed edge is created. All activity nodes and directed edges constitute a directed acyclic graph of the integrated installation schedule network plan.

[0123] Step S470: Perform critical path analysis on the integrated installation schedule network plan, identify critical process chains and non-critical process buffer periods in the installation schedule, generate a critical path schedule control baseline, associate the critical path schedule control baseline with the adjustment screw layout parameters in the project adjustment space database, generate detailed adjustment operation instructions for the connection nodes on the critical process chain, and output an integrated on-site installation execution plan that includes the integrated installation schedule network plan, the critical path schedule control baseline, and the adjustment operation instructions.

[0124] For each activity node in the directed acyclic graph, a time estimate is set: the time for hoisting activity nodes is the segmented hoisting time quota, and the time for assembly activity nodes is the node assembly time quota. The total time for each path is calculated from the start node to the end node; the total time equals the sum of the times of all activity nodes on the path. The path with the longest total time is the critical path, and the activity nodes on the critical path constitute the critical process chain. For activity nodes on non-critical paths, their buffer periods are calculated based on the difference between the latest start time and the earliest start time.

[0125] The critical path schedule control baseline records the planned start and finish times of each activity node on the critical path. For each connection node in the critical process chain, the spacing information of its adjusting screws is extracted from the project-wide adjustment space database to generate adjustment operation instructions. The adjustment operation instructions include the screw tightening sequence, the reference range for the number of turns for each screw, and the target adjustment clearance value.

[0126] The graphical structure data of the integrated installation schedule network plan, the tabular data of the critical path schedule control baseline, and the textual data of the adjustment operation instructions are integrated into an integrated field installation execution plan document. The graphical structure data is stored in the form of node tables and edge tables, and the tabular data is stored in column-separated text format.

[0127] For example, the method may also include: step S510: extracting the baseline starting coordinates and baseline direction vectors contained in the assembly baseline annotation instructions of all connected nodes in the node detail drawing instruction set, constructing a full project assembly baseline network topology diagram, and forming a spatial grid structure in the full project assembly baseline network topology diagram with the assembly baselines of each connected node as topology edges and the baseline starting coordinates of each connected node as topology vertices.

[0128] The construction method of the full project assembly baseline network topology diagram is the same as step S310.

[0129] Step S520: Perform assembly dimension propagation path analysis on the assembly baseline network topology diagram of the entire project. Trace the error propagation path from the assembly start baseline to the end baseline layer by layer along the topological edge direction of the assembly baseline network topology diagram. Identify the set of assembly propagation paths in the assembly baseline network of the entire project. The set of assembly propagation paths includes the sequence of topological vertices and the sequence of topological edges traversed by each assembly propagation path.

[0130] The assembly dimension propagation path analysis and processing is consistent with the assembly propagation path identification method in step S320. Each assembly propagation path is an ordered list of topological vertices, with adjacent topological vertices connected by topological edges. The set of assembly propagation paths contains all possible paths from all starting baselines to all end baselines.

[0131] Step S530: Extract the adjustment space features from the node construction feature tensor of the connected nodes corresponding to the topological vertices traversed by each assembly propagation path. Perform cumulative tolerance analysis on the three-dimensional adjustable space range limit in the adjustment space features along the assembly propagation path direction. Calculate the cumulative adjustable gap margin at each topological vertex of the assembly propagation path. The cumulative adjustable gap margin is obtained by sequentially superimposing the adjustable gap values ​​of each connected node from the starting topological vertex to the current topological vertex of the assembly propagation path.

[0132] For each assembly propagation path, traverse all topological vertices along the path starting from the initial topological vertex. For the q-th topological vertex on the path, sum the adjustable gap values ​​of the corresponding connected nodes of the preceding q-1 topological vertices to obtain the cumulative consumed gap value at that vertex. The cumulative adjustable gap margin is the initial total adjustable gap value of the starting topological vertex minus the cumulative consumed gap value. If the cumulative adjustable gap margin is less than zero, it is set to zero.

[0133] Step S540: Generate a phased assembly measurement and control scheme based on the cumulative adjustable clearance margin and the assembly propagation path set. Set up field measurement control points at the topology vertices of the assembly propagation path where the cumulative adjustable clearance margin is lower than the preset propagation margin threshold. The phased assembly measurement and control scheme divides the assembly process of all project connection nodes into assembly flow segments with the field measurement control points as the stage boundaries.

[0134] Each assembly propagation path is traversed, and the cumulative adjustable clearance margin of each topological vertex on the path is checked. If the cumulative adjustable clearance margin is lower than the preset propagation margin threshold, a field measurement control point is set at the connection node corresponding to that topological vertex. All connection nodes with set field measurement control points are deduplicated to form a set of measurement control points. The connection nodes between two adjacent field measurement control points constitute an assembly flow segment. The starting node of the assembly flow segment is the connection node where the previous measurement control point is located, and the ending node is the connection node where the next measurement control point is located. All assembly flow segments and their corresponding measurement control points constitute a phased assembly measurement control scheme.

[0135] Step S550: Based on the phased assembly measurement and control scheme, output a schematic diagram of the assembly flow section division for the entire project. The schematic diagram of the assembly flow section division for the entire project marks the start and end boundaries of each assembly flow section and the corresponding spatial coordinates of the on-site measurement control points on the unfolded diagram of the curtain wall skin.

[0136] The base map is the unfolded diagram of the curtain wall facade of the building. Each assembly section is distinguished on the base map by a different background color or border line type. The start and end boundaries of the assembly section are drawn with thick lines, and the assembly section number is marked at the boundary. The location of each on-site measurement control point is marked with an asterisk, and the measurement control point number and three-dimensional spatial coordinates are marked next to it.

[0137] Step S560: Extract the length dimensions and cross-sectional contour coordinates of the connectors from the geometric parameter set of the connectors in each assembly flow section, summarize them into a packing list of components for each assembly flow section, and output the construction organization information in an integrated package.

[0138] Step S561: Extract the length dimension and cross-sectional contour coordinates of the connectors from the geometric parameter set of the connectors in each assembly flow section, and generate a component packing list for each assembly flow section by unit. The component packing list for each assembly flow section classifies and collects all the connectors required in the same assembly flow section according to their specifications and models.

[0139] For each assembly flow segment, all connection nodes within that segment are traversed. The cross-sectional contour coordinates and length of each connection node are extracted from the geometric parameter set generated in step S141. The cross-sectional contour coordinates are used to calculate seven invariant moment values ​​as the cross-sectional shape feature vector of the connection. Hierarchical clustering analysis is performed on the cross-sectional shape feature vectors of all connections within the flow segment. A preset clustering distance threshold is used, and connections within the same cluster are considered to be of the same specification. Connections of the same specification are further subgrouped according to their length, with length differences within a preset tolerance range grouped into the same subgroup. Each subgroup constitutes a single line in the component packing list for the flow segment, including the specification number, cross-sectional contour description, length range, quantity, and the assembly flow segment number.

[0140] Step S562: Link the component packing list of the assembly line section with the phased assembly measurement and control scheme to the assembly line section, and generate an on-site logistics and distribution plan organized according to the assembly line section. The on-site logistics and distribution plan specifies the shipment batch, arrival time window and unloading and stacking area number of each component in the assembly line section.

[0141] Using the assembly line section number as the association key, each record in the component packing list for that section is linked to the construction time plan for that section in the phased assembly measurement and control scheme. Each assembly line section corresponds to a shipping batch, and the loading time of the shipping batch is set to the start time of construction for that section minus the transportation time and unloading buffer time. The arrival time window is the interval from the loading time plus the transportation time to the loading time plus the transportation time plus the preset receiving window time. The unloading and stacking area number is taken from the material stacking area identifier assigned to that assembly line section on the construction site layout plan. The above information is summarized into an on-site logistics distribution plan table according to the shipping batches.

[0142] Step S563: Perform trend analysis on the cumulative adjustable clearance margin sequence of each assembly propagation path in the assembly propagation path set, extract the change gradient of the cumulative adjustable clearance margin along the direction of the assembly propagation path, identify the path segments in the assembly propagation path that are prone to accumulating errors based on the change gradient, and generate on-site pre-assembly suggestion instructions for the path segments that are prone to accumulating errors.

[0143] Trend analysis processes the cumulative adjustable clearance margin of each topological vertex on each assembly propagation path using first-order difference calculation. The first-order difference is the cumulative value of the next vertex minus the cumulative value of the previous vertex. The absolute value of the first-order difference is the gradient of that segment. The absolute value of the gradient is compared with a preset gradient threshold. If it exceeds the threshold, the adjacent vertex interval of that segment is marked as a path segment prone to error accumulation. For each path segment prone to error accumulation, a field pre-assembly suggestion instruction is generated. The suggestion instruction includes the start and end identifiers of the pre-assembly vertices, the suggested location of the pre-assembly site, and a description of the pre-assembly measurement requirements.

[0144] Step S564: Package the phased assembly measurement and control plan, the schematic diagram of the assembly flow section division of the whole project, the on-site logistics and distribution plan organized according to the assembly flow section, and the on-site pre-assembly suggestion instructions into an integrated assembly construction organization document.

[0145] The phase division table of the phased assembly measurement and control scheme, the vector file of the full-project assembly flow section division diagram generated in step S550, the table file of the on-site logistics and distribution plan organized by assembly flow section, and the text file of each on-site pre-assembly suggestion instruction are archived into an integrated assembly construction organization file package. The file package is named after the assembly project name plus the generation timestamp. The file package contains a list file, which records the name and file type description of each sub-file.

[0146] Step S565: Output the integrated assembly construction organization document to the on-site construction management terminal. The integrated assembly construction organization document organizes all construction organization information with the assembly flow section as the index unit.

[0147] The integrated assembly construction organization file package is sent to the designated storage path of the on-site construction management terminal via the file transfer interface of the construction management system. After reading the file package, the construction organization management software on the on-site construction management terminal parses the contents of each sub-file according to the assembly flow section index, and displays the construction organization information in the order of the assembly flow section on the construction management interface. Construction personnel can view the corresponding measurement control requirements, logistics and distribution plan, component packing list, and pre-assembly suggestions by selecting the assembly flow section number.

[0148] Step S610: Extract the adjustable gap values ​​and three-dimensional adjustable space range boundaries from the node construction feature tensor at all curtain wall skin segment connection nodes, and construct a full-project adjustable space database with the connection node identifier as the primary key. The full-project adjustable space database stores the adjustable gap values, the spatial coordinate boundaries of the three-dimensional adjustable space range boundaries, and the corresponding curtain wall skin segment identifiers for each connection node.

[0149] The adjustment space feature portion is extracted from the node construction feature tensor of each connected node generated in step S130. For each connected node, six scalar values ​​of the three-dimensional adjustable space range boundary are read from the adjustment space feature, and the difference between the maximum and minimum values ​​in the Z direction is calculated as the adjustable gap value of that connected node. Using the connected node identifier as the primary key, the adjustable gap value as the first attribute column, the six scalar values ​​of the three-dimensional adjustable space range boundary as the second to seventh attribute columns, and the curtain wall skin segment identifier as the eighth attribute column, a relational table structure for the entire project's adjustment space database is constructed. The entire project's adjustment space database uses the connected node identifier as the index key, supporting fast retrieval by primary key.

[0150] Step S620: Sort the adjustable gap values ​​of each connection node across the entire project. Arrange all the adjustable gap records of the connection nodes in ascending order of adjustable gap values ​​to construct a project-wide node gap distribution sequence. Each sequence element of the project-wide node gap distribution sequence includes the connection node identifier, the adjustable gap value, and the identifier of the curtain wall skin segment to which it belongs.

[0151] The adjustable gap values ​​of all connection nodes are extracted from the project's overall adjustment space database, generating a list of triples. Each triple contains a connection node identifier, an adjustable gap value, and a curtain wall skin segment identifier. This list is then sorted in ascending order of adjustable gap values ​​using a merge sort algorithm. Elements with the same adjustable gap value are sorted by the lexicographical order of their connection node identifiers. The sorted list of triples represents the project-wide node gap distribution sequence. A smaller position index in the project-wide node gap distribution sequence indicates a smaller adjustable gap value for that connection node, and a tighter installation adjustment margin.

[0152] Step S630: Generate a field installation sequence scheme based on the total project node gap distribution sequence, wherein the connection nodes that are ranked earlier in the total project node gap distribution sequence are assigned an earlier assembly order, and the connection nodes that are ranked later are assigned a later assembly order, thus forming a scheme that defines the assembly order of connection nodes based on the arrangement order of the total project node gap distribution sequence.

[0153] The entire project's node gap distribution sequence has been arranged in ascending order of adjustable gap values. The on-site installation sequence directly adopts this sequence, assigning the assembly sequence number of the first connection node to 1, the second to 2, and so on up to the last connection node. A smaller assembly sequence number indicates that the connection node will be installed on-site earlier. This arrangement strategy prioritizes the installation and positioning of connection nodes with smaller gap margins, allowing subsequent connection nodes with larger gap margins to absorb accumulated errors from previous installations, thereby improving overall installation accuracy.

[0154] Step S640: Output an installation sequence distribution diagram based on the on-site installation sequence plan. The installation sequence distribution diagram is marked with color gradients on the curtain wall skin segmented spatial unfolding diagram, and the color gradients corresponding to the assembly sequence numbers are gradually transitioned along the assembly sequence direction.

[0155] Using the unfolded spatial diagram of the curtain wall facade as a base map, each connection node is projected onto its corresponding position on the unfolded diagram according to its spatial coordinates. A filled circle is drawn at each projection position. The color of the filled circle is determined by the assembly sequence number of the connection node using a color mapping function. The color mapping function uses a hue gradient: the connection node with the smallest assembly sequence number corresponds to a cool hue value, the connection node with the largest assembly sequence number corresponds to a warm hue value, and connection nodes with intermediate sequence numbers transition between cool and warm colors through linear interpolation. Each filled circle is labeled with its assembly sequence number. The base map, filled circles, and numerical labels together constitute a schematic diagram of the installation sequence distribution.

[0156] Step S650: Extract the segment identifier of the curtain wall skin segment corresponding to each connection node, group and collect the assembly sequence number of each connection node in the installation sequence distribution diagram according to the segment identifier, and generate a segment assembly sequence summary table. The segment assembly sequence summary table records the assembly sequence number of each connection node in each curtain wall skin segment and the corresponding adjustable gap value.

[0157] All connected nodes are grouped using the shard identifier as the grouping key. For each group corresponding to the shard identifier, the connected nodes within the group are sorted in ascending order by their assembly sequence number, generating a sub-table for the assembly sequence of that shard. Each row in the shard assembly sequence summary table corresponds to a shard identifier and contains the following fields: shard identifier, total number of connected nodes in the shard, earliest assembly sequence number in the shard, latest assembly sequence number in the shard, and a nested list field of assembly sequence numbers and adjustable gap values ​​for each connected node in the shard. The shard assembly sequence summary table is indexed using the shard identifier as the primary key.

[0158] Step S660: Generate a curtain wall skin segment hoisting sequence plan based on the segment assembly sequence summary table, generate a comprehensive installation schedule network plan and perform critical path analysis, and output an integrated installation execution plan.

[0159] Step S661: Generate the curtain wall skin segment hoisting sequence plan based on the segment assembly sequence summary table. Select the connection node with the smallest assembly sequence number (i.e., the earliest assembled node) in each curtain wall skin segment as the segment hoisting sequence anchor point. Arrange the segment hoisting sequence in ascending order according to the adjustable gap value of the connection node of each segment anchor point.

[0160] For each segment identifier in the summary table of segment assembly sequence, the adjustable gap value of the connection node corresponding to the earliest assembly sequence number within that segment is extracted. This adjustable gap value is used as the hoisting priority anchor point value for that segment. All segments are arranged in ascending order according to their hoisting priority anchor point values, and the resulting arrangement constitutes the curtain wall skin segment hoisting sequence scheme. This scheme specifies the order in which each curtain wall skin segment is hoisted to its corresponding position on the main building structure, ensuring that the segments with the most urgent installation needs are hoisted into place first.

[0161] Step S662: Generate a comprehensive installation schedule network plan based on the curtain wall skin segment hoisting sequence plan and the on-site installation sequence plan. Take the segment hoisting operation and node assembly operation as schedule activity nodes, and take the process sequence dependency relationship between segment hoisting and node assembly as directed edges to construct a directed acyclic graph structure that expresses the temporal logical relationship of all installation operations.

[0162] Create a set of schedule activity nodes. Create one hoisting activity node for each curtain wall cladding segment hoisting operation. Node attributes include activity name, activity type tag, and estimated duration value, which is the hoisting time quota for that segment. Create one assembly activity node for each connection node assembly operation. Node attributes include activity name, activity type tag, segment identifier, and estimated duration value, which is the assembly time quota for that connection node.

[0163] Create a set of directed edges. The establishment of directed edges follows these process constraints: A directed edge is drawn from the hoisting activity node corresponding to the same segment to the assembly activity node with the smallest assembly sequence number within that segment; a directed edge is drawn from the assembly activity node with a smaller assembly sequence number to the assembly activity node with a larger assembly sequence number within the same segment; if there is a geometric association interface between adjacent assembly activity nodes in different segments, then a directed edge is drawn from the assembly activity node of the previous segment to the assembly activity node of the next segment. All activity nodes and directed edges constitute a directed acyclic graph structure representing the sequential logical relationship of installation operations, which is the integrated installation schedule network plan.

[0164] Step S663: Perform critical path search processing on the directed acyclic graph structure of the integrated installation schedule network plan. Traverse all possible paths from the starting hoisting node to the end assembly node and calculate the cumulative duration of each path. Identify the path with the longest cumulative duration as the critical path of the installation schedule. The connecting nodes on the critical path constitute the critical process node chain.

[0165] Perform topology sorting on the directed acyclic graph of the integrated installation schedule network plan to obtain the topological order of the active nodes. Traverse each active node from front to back according to the topological order, calculating the earliest start time for each active node. For the initial hoisting node, the earliest start time is set as the project baseline time. For other active nodes, the earliest start time is equal to the maximum of the earliest finish times of all its predecessor active nodes, and the earliest finish time is equal to the earliest start time plus the estimated duration of that active node. The maximum of the earliest finish times among all final assembly active nodes is the total project duration.

[0166] Starting from the end assembly activity node and working backwards, identify the activity nodes on the critical path according to the rule that the latest completion time equals the earliest completion time. During backwards, for the current activity node, select the predecessor activity node whose earliest completion time equals the earliest start time of the current activity node and add it to the critical path. Tracing back from the end activity node to the starting hoisting node, the resulting sequence of critical path activity nodes constitutes the installation schedule critical path. The connecting nodes on the critical path form the critical process node chain. The installation progress of the connecting nodes on the critical process node chain directly affects the overall project duration; any delay will lead to an overall project delay.

[0167] Step S664: Extract the spacing information of the adjusting screws of each connecting node on the critical path from the full project adjustment space database, generate adjustment operation instructions for each connecting node on the critical process node chain, and output an integrated on-site installation execution plan that includes the comprehensive installation schedule network plan, the critical path of the installation schedule, the critical process node chain, and the adjustment operation instructions. The adjustment operation instructions include the screw turning sequence, the range of turning turns, and the target adjustment gap value.

[0168] For each connection node in the key process node chain, the corresponding adjusting screw arrangement spacing information is retrieved from the full project adjustment space database using the connection node identifier as the query key. The adjusting screw arrangement spacing information is the scalar value of the adjusting screw arrangement spacing output from the adjustment space feature in step S130. Based on the adjusting screw arrangement spacing value and the six scalar values ​​of the three-dimensional adjustable space range limit, the range of the number of turns of the adjusting screw is calculated. The calculation method for the number of turns range is to divide the adjustable clearance value by the screw pitch parameter to obtain the reference value of the number of turns, and multiply the reference value of the number of turns by the adjustment coefficient to obtain the upper limit and lower limit of the number of turns. The adjustment target clearance value is a preset ratio value of the adjustable clearance value.

[0169] The adjustment operation instructions are text instruction paragraphs generated for each individual connection node, including the connection node identifier and spatial location description, the number of the adjustment screw and its corresponding tightening sequence, the lower and upper limits of the number of turns for each screw, and the numerical value and dimensions of the target clearance value. The directed acyclic graph structure data of the integrated installation schedule network plan, the list of active nodes on the critical path, the chain sequence of critical process nodes, and the adjustment operation instructions for each node are integrated into a single integrated on-site installation execution plan document. The graph structure data is stored in node table and edge table formats, the critical path is stored in an ordered list format of node identifiers, and the adjustment operation instructions are stored in a numbered paragraph format. This integrated on-site installation execution plan serves as the primary technical basis for on-site construction management.

[0170] Step S710: Extract the connector position distribution heat map and connector direction angle distribution information contained in the connector geometric layout descriptor of all curtain wall skin segment connection nodes in the node construction feature tensor, and construct a full project connector layout database. The full project connector layout database stores the pixel coordinate sequence of the connector position distribution heat map and the connector direction angle of each connection node with the connection node identifier as the primary key.

[0171] From the node connector layout features generated in step S1351, extract the connector position distribution heatmap and connector orientation angle distribution information. The connector position distribution heatmap is a one-dimensional probability distribution sequence of sampling points along a boundary polyline, with each sampling point corresponding to a hotspot value. A pixel coordinate sequence is constructed using the pixel coordinates of all sampling points whose hotspot values ​​exceed a preset hotspot threshold. The connector orientation angle is stored as a radian value, stored as a scalar. A full-project connector layout database is constructed using the connector node identifier as the primary key and the pixel coordinate sequence and orientation angle as attribute values.

[0172] Step S720: Extract the heat map of the connector position distribution of all connection nodes under the same curtain wall skin segment identifier from the full project connector layout database. Project the heat map of the connector position distribution of each connection node in the same curtain wall skin segment onto the two-dimensional unfolded plane of the curtain wall skin segment to generate a general layout diagram of the segment connectors. The general layout diagram of the segment connectors indicates the connector assembly position of each connection node within the boundary range of the curtain wall skin segment.

[0173] Using the segment identifier of the curtain wall skin segment as the grouping key, the pixel coordinate sequence of all connection nodes within that segment is extracted from the overall project connector layout database. The 2D unfolded plane of the curtain wall skin segment is a rectangular planar area generated by flattening the segment boundary polyline using an equidistant unfolding algorithm. The pixel coordinate sequence of each connection node is projected onto this 2D unfolded plane through a spatial mapping transformation, preserving the relative distance ratio between the pixel coordinate sequences. The connector assembly position of each connection node is drawn on the 2D unfolded plane using marker symbols, generating a general layout diagram of the segment connectors.

[0174] Step S730: Perform spatial density analysis on the overall layout of the segmented connectors. Using the two-dimensional unfolded plane of the curtain wall skin segments as the analysis domain, calculate the distribution density value of the connector assembly positions per unit area, and generate a heat map of the density distribution of the segmented connectors. The heat map of the density distribution of the segmented connectors uses color depth to express the density of connectors in different areas of the curtain wall skin segments.

[0175] An evenly spaced grid is overlaid on the two-dimensional unfolded plane of the overall layout of the segmented connectors, with the grid cell side length being a preset density analysis grid size. For each grid cell, the number of connector assembly positions falling within that cell is counted, and this number is divided by the grid cell area to obtain the connector distribution density value per unit area for that grid cell. The density values ​​of all grid cells are then converted into fill colors using a color mapping table, with higher density values ​​resulting in darker colors, generating a heatmap of the segmented connector density distribution.

[0176] Step S740: Identify densely distributed areas of connectors whose assembly location density exceeds a preset density threshold based on the heat map of the density distribution of the segmented connectors. Construct a centralized assembly block of connectors based on the boundary contour of the densely distributed area of ​​connectors. Group the connector nodes in the same centralized assembly block into a centralized assembly group of connectors.

[0177] Extract the set of grid cells with density values ​​exceeding a preset density threshold from the heatmap of the density distribution of the patch connectors. Use an image morphological dilation and erosion closure algorithm to connect adjacent high-density grid cells into connected regions. Extract the coordinate sequence of the boundary grid cells of each connected region as the boundary contour of the densely distributed connector area. The area enclosed by a boundary contour is a connector assembly block. Merge the connector nodes contained within each connector assembly block into the same connector assembly group.

[0178] Step S750: Generate a pre-assembly scheme for the centralized assembly area based on the connector centralized assembly group. Connectors belonging to the same connector centralized assembly group are pre-assembled into connector component units during the factory prefabrication stage. The pre-assembly scheme for the centralized assembly area includes a three-dimensional assembly positioning diagram of the connector component unit and a component hoisting center of gravity mark.

[0179] For each connector assembly group, the cross-sectional contour coordinates and orientation angles of each connector node within the group are simulated and pre-assembled using a 3D spatial assembly algorithm. The 3D assembly positioning diagram of the connector component unit is stored in a 3D model file format, containing the 3D solid model of each connector and their assembly positioning relationships. The component's hoisting center of gravity is determined by calculating the weighted centroid of each connector within the connector component unit, with the centroid coordinates being the mass-weighted arithmetic mean of the centroids of all connectors within the group.

[0180] Step S760: Generate a component unit hoisting attitude planning instruction based on the pre-assembly scheme of the centralized assembly area and the directional angle distribution information of each connection node in the centralized assembly group of connectors. The component unit hoisting attitude planning instruction specifies the spatial attitude change sequence of the connector component unit throughout the entire process from ground hoisting to installation.

[0181] The initial posture of the connector component unit when it is laid flat on the ground is taken as the lifting posture, and the target posture when the component unit is finally installed in place on the main building structure is taken as the termination posture. The spatial posture change path between the lifting posture and the termination posture is generated into an intermediate posture sequence through spherical linear interpolation. Each intermediate posture contains the spatial position coordinates and orientation angle quaternion representation of the component. The component unit lifting posture planning instruction is a posture sequence file composed of the intermediate postures arranged in chronological order.

[0182] Step S770: Extract the adjustment space characteristics of each connection node in the connector assembly group, sort the adjustable gap values ​​of each connection node in the connector assembly group within the group, and generate the installation order of the connection nodes in the connector assembly group in ascending order of adjustable gap.

[0183] Adjustable clearance values ​​for each connection node within the centralized assembly group of this connector are extracted from the project's overall adjustment space database and sorted in ascending order to obtain the installation sequence within the group. This installation sequence specifies the order in which each connector within the connector assembly unit is installed after arrival at the installation site.

[0184] Step S780: Output the component unit hoisting attitude planning instructions and the installation sequence of the connection nodes in the centralized assembly group of the connector as a centralized assembly area construction instruction package. Output an integrated centralized assembly construction guidance document containing the centralized assembly area pre-assembly scheme, component unit hoisting attitude planning instructions and centralized assembly area construction instruction package. The centralized assembly area construction instruction package is used to guide the factory pre-assembly and on-site overall hoisting construction of the same connector centralized assembly area.

[0185] The pre-assembly scheme for the centralized assembly area generated in step S750, the component unit hoisting attitude planning instructions generated in step S760, and the intra-group installation sequence generated in step S770 are packaged into a centralized assembly area construction instruction package. The centralized assembly area construction instruction package is a compressed file containing three sub-files. The correspondence between the centralized assembly area construction instruction package and the identifiers of the connector centralized assembly groups is recorded in the index section of the integrated centralized assembly construction guidance document. The guidance document contains an index table of all connector centralized assembly areas for the entire project and references to the storage paths of each construction instruction package.

[0186] Figure 2 This application illustrates a data-driven automated generation system 100 for architectural curtain wall detail drawings, comprising a processor 1001 and a memory 1003. The processor 1001 and memory 1003 are connected, for example, via a bus 1002. Optionally, the data-driven automated generation system 100 may further include a transceiver 1004, which can be used for data interaction between this data-driven automated generation system and other data-driven automated generation systems for architectural curtain wall detail drawings, such as sending and / or receiving data. It should be noted that in actual scheduling, the transceiver 1004 is not limited to one, and the structure of this data-driven automated generation system 100 does not constitute a limitation on the embodiments of this application.

[0187] The memory 1003 is used to store program code for executing the embodiments of this application, and its execution is controlled by the processor 1001. The processor 1001 is used to execute the program code stored in the memory 1003 to implement the steps shown in the foregoing method embodiments.

[0188] The above description is only an optional implementation method for some implementation scenarios of this application. It should be noted that for those skilled in the art, other similar implementation methods based on the technical concept of this application, without departing from the technical concept of this application, also fall within the protection scope of the embodiments of this application.

Claims

1. A data-driven method for automatically generating detailed drawings of building curtain wall nodes, characterized in that, The detailed drawings of the building curtain wall nodes are used to express the assembly relationship and connection structure of the curtain wall components. The method includes: Acquire curtain wall skin segment data and main structure configuration data of the target curtain wall project. The curtain wall skin segment data includes the boundary polyline sequence and segment identifier of each curtain wall skin segment, and the main structure configuration data includes the spatial coordinates of the main structure control points. Based on the curtain wall skin segment data and the main structure configuration data, geometric mapping and deviation analysis are performed on the curtain wall skin segments and the main structure to generate a spatial deviation vector field between the curtain wall skin segments and the main structure. The spatial deviation vector field includes the normal deviation vector and tangential deviation vector at each curtain wall skin segment boundary polyline sampling point relative to the main structure control point. The pre-built node detail diagram is called to generate a network that processes the spatial deviation vector field and the preset node construction type label to generate a node construction feature tensor at the connection node of the curtain wall skin segment. The node construction feature tensor includes node connector layout features, sealing construction features and adjustment space features. Based on the node construction feature tensor, geometric parameterization decoding is performed to generate a set of node detail drawing instructions at the connection nodes of the curtain wall skin segments. The set of node detail drawing instructions includes node area cutting plane indication, assembly baseline annotation, and component outline drawing instructions.

2. The data-driven automated generation method for detailed architectural curtain wall nodes according to claim 1, characterized in that, The step of performing geometric mapping and deviation analysis on the curtain wall skin segments and the main structure based on the curtain wall skin segment data and the main structure configuration data to generate a spatial deviation vector field between the curtain wall skin segments and the main structure includes: The boundary polyline sequence in the curtain wall skin segment data is spatially offset along the normal of the curtain wall skin segment to obtain the offset polyline sequence of the corresponding curtain wall skin segment. Each offset polyline segment of the offset polyline sequence forms a normal mapping correspondence with the original boundary polyline segment. The offset distance of the spatial offset processing is determined based on the theoretical spacing parameter between the curtain wall skin and the main structure preset in the curtain wall construction design. Spatial triangulation processing is performed on the main structure control points in the main structure configuration data to generate a set of main structure surface triangular mesh units. Each main structure surface triangular mesh unit is composed of three main structure control points connected together. The spatial triangulation processing constructs spatial topological connection relationships based on the Delaunay triangulation criteria between the main structure control points. Each polyline sampling point on the offset polyline sequence is subjected to distance field calculation processing with the set of triangular mesh elements on the main structure surface to determine the nearest projection point of each polyline sampling point on the triangular mesh element on the main structure surface. The distance field calculation processing is achieved by searching for the surface point with the smallest Euclidean distance to the polyline sampling point in the set of triangular mesh elements on the main structure surface. The normal deviation vector is calculated based on the spatial position difference between the polygonal sampling point and the corresponding nearest projection point. The direction of the normal deviation vector is along the normal direction of the curtain wall skin segment, and the modulus is the distance from the polygonal sampling point to the nearest projection point along the normal direction. The normal deviation vector represents the degree of spatial deviation of the curtain wall skin segment relative to the curved surface of the main structure in the direction perpendicular to the skin. The tangential deviation vector is calculated based on the spatial position difference between the polygonal sampling point and the corresponding nearest projection point. The direction of the tangential deviation vector is perpendicular to the normal deviation vector and is located in the cutting plane of the curtain wall skin segment. The modulus is the distance from the polygonal sampling point to the nearest projection point along the tangential direction. The tangential deviation vector represents the degree of spatial slippage of the curtain wall skin segment relative to the curved surface of the main structure in the direction parallel to the skin. The normal deviation vector and tangential deviation vector are assigned to the corresponding polyline sampling points to form a spatial deviation vector tuple at each polyline sampling point of the curtain wall skin segment boundary. The spatial deviation vector tuple stores the normal deviation vector and tangential deviation vector as deviation attributes of the polyline sampling points in the same data structure. The spatial deviation vector tuple is subjected to continuity constraint processing along the boundary polyline sequence so that the normal deviation vector and tangential deviation vector of adjacent polyline sampling points satisfy the local smooth transition relationship, generating a smooth deviation vector sequence. The continuity constraint processing uses a weighted moving average operator between the deviation vectors of adjacent sampling points to eliminate local noise disturbances. The smoothed deviation vector sequence is combined with the segment identifier of the curtain wall skin segment for spatial indexing and organization, and a joint distribution set of deviation vectors is constructed and deviation clustering analysis is performed.

3. The data-driven automated generation method for detailed architectural curtain wall nodes according to claim 2, characterized in that, The process of spatially indexing and organizing the smoothed deviation vector sequence in conjunction with the segment identifiers of the curtain wall skin segments, and constructing a joint distribution set of deviation vectors and performing deviation clustering analysis, includes: The smooth deviation vector sequence is combined with the segment identifier of the curtain wall skin segment to perform spatial indexing and organization processing to generate a spatial deviation vector field with the curtain wall skin segment as the index unit. Each spatial deviation vector in the spatial deviation vector field corresponds to a polyline sampling point and contains a normal deviation vector and a tangential deviation vector. The spatial indexing and organization processing constructs a multi-level mapping relationship from the segment identifier to the smooth deviation vector sequence. Extract the normal and tangential deviation vectors of all curtain wall skin segment boundary polyline sampling points in the spatial deviation vector field, construct a joint distribution set of deviation vectors for the entire project, and store all spatial deviation vectors of each curtain wall skin segment in groups according to segment identifiers; The joint distribution set of deviation vectors is subjected to deviation clustering analysis. The spatial deviation vectors are divided into several deviation pattern categories according to the joint distribution characteristics of the normal deviation vector magnitude and the tangential deviation vector magnitude, and a set of node deviation pattern classification labels is generated.

4. The data-driven automated generation method for detailed architectural curtain wall nodes according to claim 1, characterized in that, The process of calling the pre-built node detail map generation network to process the spatial deviation vector field and the preset node construction type label generates a node construction feature tensor at the connection node of the curtain wall skin segment, including: The spatial deviation vector field is input into the spatial coding module of the pre-constructed node detail map generation network for feature extraction. The node neighborhood spatial deviation tensor is constructed by the normal deviation vector and the tangential deviation vector in the spatial deviation vector field. The node neighborhood spatial deviation tensor aggregates the deviation vector information of multiple polyline sampling points in the neighborhood with the polyline sampling point where the node is located as the center. The node neighborhood spatial deviation tensor is subjected to multi-offset ratio hole depth separable convolution processing to extract the multi-receptive field deviation feature map at the connection node of the curtain wall skin. The multi-offset ratio hole depth separable convolution processing scans the node neighborhood spatial deviation tensor in parallel with depth separable convolution kernels with different hole ratios to capture the deviation distribution pattern in different spatial ranges. The preset node construction type markers are processed by the type embedding module of the node detail graph generation network to generate node construction type embedding vectors. The vector embedding process maps discrete node construction type markers to dense vector representations in a continuous vector space. The node construction type markers include beam connection node type, column connection node type and corner connection node type. The cross-attention fusion module of the node detail map generation network is invoked to perform feature fusion processing on the multi-receptive field bias feature map and the node construction type embedding vector. The node construction type embedding vector is used as the attention query tensor to perform weighted response extraction on the multi-receptive field bias feature map to generate a type-aware bias feature map. The weighted response extraction is achieved by calculating the similarity weight between the node construction type embedding vector and the spatial location features of the multi-receptive field bias feature map and performing a weighted aggregation operation. The node connector layout features, sealing structure features, and adjustment space features are generated by using the node detail diagram to generate the connector sub-network, sealing structure generation sub-network, and adjustment space generation sub-network, respectively, and then feature fusion and compression processing is performed.

5. The data-driven automated generation method for detailed architectural curtain wall nodes according to claim 4, characterized in that, The process of generating node connector layout features, sealing structure features, and adjustment space features using the node detail map generation network, the connector generation sub-network, the sealing structure generation sub-network, and the adjustment space generation sub-network respectively, and then performing feature fusion and compression processing, including: The connector generation subnetwork of the node detail map generation network is used to process the type-aware bias feature map to generate connector layout patterns. The connector geometric layout descriptor is predicted as the node connector layout feature. The connector geometric layout descriptor includes the connector position distribution heat map and the connector direction angle distribution information. The connector generation subnetwork is implemented by a temporal layout decoding structure composed of multiple causal dilated convolutional layers stacked together. The sealing structure generation sub-network of the node detail map generation network is used to process the sealing structure parameters of the type-aware bias feature map, and the predicted sealing structure parameter descriptor is used as the sealing structure feature. The sealing structure parameter descriptor includes the distribution of the filling thickness of the sealing material and the marking of the sealing installation method. The sealing structure generation sub-network gradually recovers the spatial resolution of the sealing structure parameters through the deconvolution upsampling layer. The adjustment space generation subnetwork of the node detail map generation network is used to process the type perception deviation feature map to generate adjustment space reserved parameters. The predicted adjustment space parameter descriptor is used as the adjustment space feature. The adjustment space parameter descriptor includes the three-dimensional adjustable space range limit and the adjustment screw arrangement spacing information. The adjustment space generation subnetwork adopts the channel attention weighting mechanism to assign differentiated parameter generation weights to features in different deviation directions. The node connector layout features, sealing structure features, and adjustment space features are spliced ​​along the feature channel direction to generate a node structure feature tensor. The statistical distribution characteristics of each channel feature component in the node construction feature tensor are extracted, and the mean and standard deviation of each channel feature component are calculated as channel normalization parameters. Based on the channel normalization parameters, the node construction feature tensor is normalized channel by channel to generate a normalized node construction feature tensor. The normalized node construction feature tensor is then subjected to nonlinear feature compression. The normalized node construction feature tensor is mapped to a low-dimensional compact representation space through a fully connected bottleneck layer to generate a compact node construction feature representation.

6. The data-driven automated generation method for detailed architectural curtain wall nodes according to claim 1, characterized in that, The geometric parameterization decoding process based on the feature tensor constructed from the nodes generates a set of node detail drawing instructions for the connection nodes of the curtain wall skin segments, including: The node construction feature tensor is input into the connector decoding branch of the geometric parameterization decoder for parameter parsing processing. The node connector layout feature is decoded into a connector geometric parameter set. The connector geometric parameter set includes the connector cross-sectional contour coordinates and the connector length dimension. The connector decoding branch is composed of a sequentially connected fully connected mapping layer and a coordinate regression output layer. The node construction feature tensor is input into the sealing construction decoding branch of the geometric parameterization decoder for parameter parsing processing. The sealing construction feature is decoded into a set of sealing construction geometric parameters, which includes the width of the sealant joint and the embedding depth of the sealing material. The sealing construction decoding branch performs stepwise parameter expansion of the sealing construction feature through a sequence of gated cyclic units. The node construction feature tensor is input into the adjustment space decoding branch of the geometric parameterization decoder for parameter parsing processing. The adjustment space feature is decoded into a set of adjustment space geometric parameters. The set of adjustment space geometric parameters includes the range of the reserved gap in the adjustment space and the coordinates of the positioning point of the adjustment bolt. The adjustment space decoding branch uses a multi-head self-attention mechanism to capture the interdependencies between the parameters of each dimension of the adjustment space. Extract the cross-sectional contour coordinates of the connector from the set of geometric parameters of the connector, and generate a connector contour drawing instruction. The connector contour drawing instruction includes a graphic description of the outer contour of the connector with reference to the assembly datum line. The graphic description of the outer contour of the connector is expressed by a closed polygon path formed by connecting the cross-sectional contour coordinates of the connector in sequence. Extract the sealant joint width and sealant embedding depth from the set of geometric parameters of the sealing structure, and generate a sealing structure contour drawing instruction. The sealing structure contour drawing instruction includes a graphic description of the sealant joint boundary contour and a graphic description of the sealant filling area. The graphic description of the sealant joint boundary contour is defined by three geometric elements: the sealant joint start coordinates, the sealant joint end coordinates, and the sealant joint width. Extract the spatial positions of the polyline sampling points in the connecting node region of the spatial deviation vector field to construct the node cutting plane equation, and generate a cutting plane indication command. The cutting plane indication command is used to determine the cutting direction and cutting position of the node detail view. The node cutting plane equation takes the mean direction of the normal deviation vector of the polyline sampling points in the connecting node region as the plane normal vector and the geometric center of the polyline sampling points as the point through which the plane passes. The assembly datum line direction of the node detail drawing is taken as the direction of the average normal deviation vector of multiple polyline sampling points in the node area, and the geometric center of the polyline sampling points in the node area is taken as the origin of the assembly datum line. An assembly datum line annotation instruction is generated, which includes the datum line starting coordinates, the datum line direction vector and the datum line symbol. The instructions for drawing the outline of the connector, the outline of the sealing structure, the cutting plane indication, and the assembly baseline annotation are integrated, and the resulting set of node detail drawing instructions is processed by coordinate system I and priority sorting.

7. The data-driven automated generation method for detailed architectural curtain wall nodes according to claim 6, characterized in that, The process involves fusing the connector contour drawing instructions, sealing structure contour drawing instructions, section plane indication instructions, and assembly baseline annotation instructions, and then performing coordinate system 1 and priority sorting on the fused node detail drawing instruction set, including: By integrating the connector outline drawing instruction, the sealing structure outline drawing instruction, the section plane indication instruction, and the assembly baseline annotation instruction, a set of node detail drawing instructions is generated. The instructions in the node detail drawing instruction set are arranged into a structured instruction sequence according to the view generation order. The coordinate system of each drawing instruction in the node detail drawing instruction set is analyzed, and the local coordinates in the connector contour drawing instruction and the sealing structure contour drawing instruction are uniformly transformed to the node local coordinate system with the assembly datum line as the reference, thus generating the node detail drawing instruction set of coordinate system one. The node detail drawing instruction set of the coordinate system one is sorted by drawing priority. The instruction execution sequence is arranged according to the order of cutting plane indication instruction, assembly datum line annotation instruction, sealing structure contour drawing instruction and connector contour drawing instruction, so as to generate an ordered node detail drawing instruction set.

8. The data-driven automated generation method for detailed architectural curtain wall nodes according to claim 1, characterized in that, The method further includes: The adjustment space features of all curtain wall skin segment connection nodes are extracted from the node construction feature tensor. The adjustment space features of all connection nodes are statistically analyzed to determine the distribution range of adjustable gap values ​​in the adjustment space features. The statistical analysis determines the distribution boundary by calculating the mean and extreme values ​​of the adjustable gap values ​​of all connection nodes. An adaptive node grouping scheme is generated based on the adjustable gap value distribution range. The adaptive node grouping scheme is used to guide the graded processing of components when prefabricating node components in the factory. The adaptive node grouping scheme divides several processing batch groups according to different intervals of the adjustable gap value distribution range. Based on the adaptive node grouping scheme, a prefabrication identification file for grouped nodes is output. The prefabrication identification file for grouped nodes is used to associate the preparation of node components in the same batch of processing parameters. The prefabrication identification file for grouped nodes includes a group number, an adjustable gap range, and a list of curtain wall skin segment identifications. Extract the node connector layout features of each connected node in the node construction feature tensor, summarize the connector geometric layout descriptors of each connected node corresponding to the prefabrication identification file of the same group of nodes into the group connector layout summary information, generate the group connector layout scheme according to the group connector layout summary information, and arrange the cross-sectional contours of each connector processed in the same batch on the raw material plate to optimize material utilization. Based on the grouped connector layout scheme, a CNC stamping blanking program instruction is generated. The CNC stamping blanking program instruction is used to drive the sheet metal stamping equipment to perform batch blanking operations of the connectors according to the layout path. Extract the connector outline drawing instructions corresponding to each connection node from the node detail drawing instruction set, and convert the connector outline drawing instructions into connector material codes and quantity statistics records; The material codes and quantity statistics of the connectors are aggregated by group according to the prefabrication identification file of the group nodes to generate batch material purchase orders. The component processing schedule is generated according to the batch material purchase orders and the prefabrication identification file of the group nodes. The component processing schedule arranges the batch processing tasks according to the expected delivery cycle on the time axis. The component processing schedule table is linked with the prefabrication identification file of the group node to generate a factory production instruction package containing processing batch identification, bill of materials and delivery time nodes.

9. The data-driven automated generation method for detailed architectural curtain wall nodes according to claim 1, characterized in that, The method further includes: Extract the assembly baseline annotation instructions of all connected nodes from the node detail drawing instruction set, and construct the assembly baseline network topology diagram of the whole project. The assembly baseline network topology diagram of the whole project forms a spatial grid structure with the assembly baseline of each connected node as the topology edge and the connected node as the topology vertex. The assembly dimension analysis is performed on the topology diagram of the assembly baseline network of the entire project to identify key assembly control nodes and assembly propagation paths in the assembly baseline network. The assembly propagation path represents the cumulative propagation direction of assembly error along the baseline network. A phased assembly measurement and control scheme is generated based on the assembly propagation path. The phased assembly measurement and control scheme sets up on-site measurement and control points at key nodes of the assembly propagation path to control the error accumulation range. Based on the phased assembly measurement and control scheme, output the on-site assembly measurement and control point layout drawing, which marks the spatial coordinates and allowable deviation range of each measurement and control point; Extract the section plane indication instructions of each connecting node from the node detail drawing instruction set, determine the node detail drawing observation view direction of each connecting node, and perform joint analysis of the node detail drawing observation view direction and the assembly baseline direction of the corresponding connecting node to generate a three-dimensional visualization assembly simulation scene description file. The three-dimensional visualization assembly simulation scene description file is used to display the assembly process of each connecting node in the actual assembly sequence in the visualization terminal. The assembly simulation scene description file is used to drive the display device to play the assembly simulation screen, generate the assembly animation key frame sequence, and synchronize the assembly animation key frame sequence with the phased assembly measurement control scheme in time to generate an assembly simulation demonstration file with measurement node annotations. Extract the connection node identifiers corresponding to each keyframe in the assembly simulation demonstration file with measurement node annotations, generate an assembly process card sequence, sort the assembly process card sequence according to the order of the assembly propagation path, and output a field construction guidance manual. In the field construction guidance manual, each assembly process card corresponds to the installation instructions and measurement requirements of a connection node.

10. The data-driven automated generation method for detailed architectural curtain wall nodes according to claim 1, characterized in that, The method further includes: Extract the three-dimensional adjustable space range limit and the information on the spacing of the adjustable screws from the node structure feature tensor of all curtain wall skin segment connection nodes, and construct a full project adjustable space database. The full project adjustable space database stores the adjustable gap value and screw arrangement parameters of each connection node with the connection node identifier as the primary key. The adjustable gap values ​​of each connection node are sorted to construct a full project node gap distribution sequence. The full project node gap distribution sequence arranges the adjustable gap records of all connection nodes in ascending order of adjustable gap values. Based on the total project node gap distribution sequence, a field installation sequence plan is generated. The field installation sequence plan arranges the assembly order of the connection nodes in a progressively decreasing gap margin order, with the connection nodes with decreasing gap margins being installed on-site before the connection nodes with increasing gap margins. Based on the above-mentioned on-site installation sequence scheme, an installation sequence distribution diagram is output. The installation sequence distribution diagram is marked with color gradients on the curtain wall skin panel unfolded diagram to indicate the assembly sequence number of each connection node. Extract the segment identifiers of the curtain wall skin segments corresponding to each connection node, group and collect the assembly sequence numbers of each connection node in the installation sequence distribution diagram according to the segment identifiers, generate a segment assembly sequence summary table, generate a curtain wall skin segment hoisting sequence scheme based on the segment assembly sequence summary table, and arrange the segment hoisting order according to the sequence number of the earliest assembled connection node in each segment. Based on the curtain wall skin segment hoisting sequence scheme and the on-site installation sequence scheme, a comprehensive installation schedule network plan is generated. The comprehensive installation schedule network plan expresses the temporal dependency relationship between segment hoisting operations and node assembly operations as a directed acyclic graph structure. The comprehensive installation schedule network plan is subjected to critical path analysis to identify critical process chains and non-critical process buffer periods in the installation schedule. A critical path schedule control baseline is generated, and the critical path schedule control baseline is associated with the adjustment screw layout parameters in the full project adjustment space database. Detailed adjustment operation instructions are generated for the connection nodes on the critical process chains, and an integrated on-site installation execution plan including the comprehensive installation schedule network plan, the critical path schedule control baseline, and the adjustment operation instructions is output.

Images