Intelligent pre-assembly method for silo slip-form template based on multi-source data fusion

Abstract

The present application relates to pre-assembly technical field, the present application discloses a kind of shallow round bin slip form template intelligent pre-assembly method based on multi-source data fusion, comprising: the point cloud dataset and BIM dataset are geometrically registered, generate the point cloud dataset after registration, physical characteristics, dynamic characteristics are associated with the point cloud dataset after registration, obtain joint feature vector set, then the point cloud dataset after registration is weighted, obtain weighted point cloud dataset.The present application is associated by neighbor node aggregation to capture construction action to support policy generation, then efficiency, precision, safety multi-objective reward function drives policy generation;Finally, through the amount of misalignment check, stress check and dynamic check progressive verification system is formed, from static compliance to dynamic stability Comprehensive guarantee scheme feasibility, break through the limitation of traditional dependence single model, realize the high precision, high safety and strong adaptability of construction pre-assembly.

Description

Technical Field

[0001] This invention relates to the field of pre-assembly technology, and more specifically, to an intelligent pre-assembly method for shallow circular silo slipform templates based on multi-source data fusion. Background Technology

[0002] Shallow circular silos are a type of building structure commonly used for storing bulk materials. They are characterized by a circular or polygonal base and relatively low height. During their construction, intelligent pre-assembly technology for building materials is widely used. This technology improves construction efficiency and reduces on-site work time by assembling parts of the structure in a factory beforehand. Simultaneously, movable formwork technology has also been introduced into the construction of shallow circular silos, allowing for flexible adjustment of the formwork position during concrete pouring, ensuring the accuracy and stability of the structure. These modern construction methods effectively improve the construction quality and efficiency of shallow circular silos.

[0003] However, due to insufficient multi-source data collaboration in existing technologies, traditional pre-assembly techniques rely on a single data source, leading to a disconnect between geometric information and physical characteristics, and making it difficult to integrate real-time disturbances in dynamic working conditions, thus resulting in decision-making biases. At the same time, existing geometric registration methods, such as traditional ICP algorithms, often ignore physical constraints and are unable to cope with high-noise point cloud data, leading to the accumulation of registration errors and affecting the reliability of subsequent analysis. Furthermore, traditional decision models, based on fixed rules or a single physical model, are difficult to adapt to complex scene changes and cannot dynamically balance the multi-objective optimization requirements of efficiency, accuracy, and safety. In addition, existing verification methods mostly rely on a single approach, lack the ability to predict risks across the entire chain, and easily overlook micro-dynamic effects. Summary of the Invention

[0004] To overcome the above-mentioned deficiencies of the prior art, embodiments of the present invention provide an intelligent pre-assembly method for shallow circular silo sliding formwork based on multi-source data fusion.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] A method for intelligent pre-assembly of slipform templates for shallow circular silos based on multi-source data fusion, the method comprising:

[0007] S101: Acquire multi-source data, parse and clean the multi-source data to obtain a standardized dataset, and extract geometric, physical and dynamic features from the standardized dataset; wherein, the standardized dataset includes point cloud dataset, BIM dataset and drawing geometry dataset;

[0008] S102: Perform geometric registration between the point cloud dataset and the BIM dataset to generate a registered point cloud dataset. Associate the physical features and dynamic features with the registered point cloud dataset to obtain a joint feature vector set. Then, weight the registered point cloud dataset to obtain a weighted point cloud dataset.

[0009] S103: Obtain material parameters and combine them with physical characteristics to form mechanical constraints. Construct a process cause-effect graph based on the joint feature vector set and construction sequence. On the basis of mechanical constraints, transform template parameters into a set of legal symbols, design a reward function, and obtain the initial assembly strategy.

[0010] S104: Combine mechanical constraints to perform multi-dimensional verification of the initial assembly strategy to obtain the final assembly strategy.

[0011] Furthermore, the specific steps for geometric registration of the point cloud dataset and the BIM dataset are as follows:

[0012] Extract normal vectors from the point cloud dataset, calculate curvature, and select high curvature points as point cloud feature points to obtain a point cloud feature point set;

[0013] Based on the BIM dataset, template plane equations are applied to extract BIM feature points, resulting in a set of BIM feature points.

[0014] Randomly select point cloud features, search for the corresponding nearest neighbor points in the BIM feature points, and obtain the coarsely registered point cloud dataset through the transformation matrix;

[0015] Based on the coarsely registered point cloud dataset, the BIM dataset is gradually processed to obtain an initial set of point pairs. Constraints and filters are applied to this initial set to generate a constrained and filtered set of point pairs. By minimizing the objective function, the registered point cloud dataset is obtained.

[0016]

[0017] In the formula: R is the rotation matrix, and t is the translation vector. For each iteration, the updated point pair set, For the points in the coarsely registered point cloud dataset, These are points in the BIM dataset.

[0018] Furthermore, the association of physical features, dynamic features, and the registered point cloud dataset includes:

[0019] The key points of the template are extracted from the registered point cloud dataset to obtain a key point set, which includes the three-dimensional coordinates, normal vectors and curvature values ​​of the key points.

[0020] Calculate the concrete lateral pressure at the key point based on its vertical height in the three-dimensional coordinates; calculate the wind load component based on the normal vector; and calculate the thermal deformation based on the thermal expansion coefficient and temperature difference of the formwork material.

[0021] The key point set, concrete lateral pressure, wind load component, and thermal deformation are combined to form a joint feature vector set.

[0022] Furthermore, the construction of the process causal graph based on the joint feature vector set and construction sequence includes:

[0023] Each construction action is treated as a node, and the physical attributes of the construction action and the key features in the joint feature vector set are bound to the node to generate a node set;

[0024] Based on the construction sequence, the sequential dependencies in the construction work are defined as edges, and the weights are calculated to obtain the edge set.

[0025] Connecting the set of nodes and the set of edges yields the process cause-effect graph.

[0026] Furthermore, the construction of the process causal graph based on the joint feature vector set and construction sequence also includes:

[0027] Initialize the node features in the process cause-effect graph, aggregate the neighbor nodes for each node, and collect the feature vectors of the neighbor nodes:

[0028]

[0029] In the formula: Let be the set of neighboring nodes of a node. Attention coefficient For parameter matrices, This is the activation function.

[0030] Furthermore, the design reward function includes:

[0031] The reward function is divided into an efficiency term, an accuracy term, and a safety term. The efficiency term is defined as the reciprocal of the assembly time, the accuracy term is the reciprocal of the deviation, and the safety term is defined as the difference between the allowable stress and the actual stress.

[0032] The reward value is obtained by weighting and summing the efficiency, accuracy, and security items separately.

[0033] Furthermore, the initial assembly strategy includes:

[0034] The assembly process is broken down into discrete and continuous actions;

[0035] For discrete actions, the assembly order is encoded as an integer; for continuous actions, fine-tuning displacements are set; and discrete and continuous actions are combined into an action vector.

[0036] An initial strategy is generated based on the current assembly state, and the relative benefit of a certain assembly action in the current assembly state is measured:

[0037]

[0038] In the formula: Given the expected reward of choosing action a under the current assembly state s, The baseline value for the current assembly state s;

[0039] By calculating relative returns, the initial strategy is updated using the maximization objective function to obtain the initial assembly strategy.

[0040] Furthermore, updating the initial policy by maximizing the objective function includes:

[0041]

[0042] In the formula: The probability of the action in the new strategy. This represents the probability of the action under the old strategy. For the clip parameter.

[0043] An electronic device includes a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the above-described intelligent pre-assembly method for shallow circular silo sliding formwork based on multi-source data fusion.

[0044] A computer-readable storage medium storing a computer program, which, when executed, implements the above-described intelligent pre-assembly method for shallow circular silo sliding formwork based on multi-source data fusion.

[0045] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0046] This application discloses an intelligent pre-assembly method for shallow circular silo sliding formwork templates based on multi-source data fusion, including: geometrically registering point cloud datasets and BIM datasets to generate registered point cloud datasets; associating physical features and dynamic features with the registered point cloud datasets to obtain a joint feature vector set; and then weighting the registered point cloud datasets to obtain a weighted point cloud dataset. This invention constructs a multi-source data-driven intelligent pre-assembly framework through the deep fusion of geometric features and physical features. Its advantages lie in its multi-dimensional technology integration and closed-loop optimization capabilities. First, geometric registration adopts a feature selection strategy driven by curvature and normal vectors, combined with an improved ICP algorithm and geometric constraint equations, effectively suppressing error propagation and improving robustness through dynamic weight adjustment while ensuring registration accuracy. Second, it couples concrete lateral pressure, wind load components, and thermal deformation with the geometric properties of key points to form a multi-physics joint feature vector, realizing cross-modal mapping from geometric shape to mechanical response and providing physical consistency constraints for decision-making. In addition, it transforms mechanical formulas into regularization terms and captures the correlation between construction actions through neighbor node aggregation to support strategy generation, and then drives strategy generation with a multi-objective reward function of efficiency, accuracy, and safety. Finally, it forms a progressive verification system through misalignment verification, stress verification, and dynamic verification, comprehensively ensuring the feasibility of the scheme from static compliance to dynamic stability, breaking through the limitations of traditional reliance on a single model, and achieving high precision, high safety, and strong adaptability in construction pre-assembly. Attached Figure Description

[0047] Figure 1 A flowchart of an intelligent pre-assembly method for shallow circular warehouse sliding formwork based on multi-source data fusion is provided for this invention;

[0048] Figure 2 A schematic diagram of the structure of an electronic device provided by the present invention;

[0049] Figure 3 A schematic diagram of the structure of a computer-readable storage medium provided by the present invention;

[0050] Figure 4 The architecture diagram of the intelligent pre-assembly method for shallow circular warehouse sliding formwork based on multi-source data fusion provided by the present invention is shown below.

[0051] Figure 5 The flowchart of S102 in the intelligent pre-assembly method of shallow circular warehouse sliding formwork template based on multi-source data fusion provided by the present invention. Detailed Implementation

[0052] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0053] Example 1

[0054] Please see Figure 1 and Figure 4 As shown in the figure, this embodiment discloses a method for intelligent pre-assembly of shallow circular silo slipform templates based on multi-source data fusion. The method includes:

[0055] S101: Acquire multi-source data, parse and clean the multi-source data to obtain a standardized dataset, and extract geometric, physical and dynamic features from the standardized dataset; wherein, the standardized dataset includes point cloud dataset, BIM dataset and drawing geometry dataset;

[0056] In this embodiment, the multi-source data includes lidar point clouds, BIM model building data, construction drawings, and sensor data. The parsing and cleaning of the multi-source data includes:

[0057] For LiDAR point clouds, Open3D is used to read the point cloud files and extract the three-dimensional coordinates and intensity values. For BIM model building data, the geometric entities of the template are extracted through the IFC parsing library, and the entity ID, geometric type and design coordinates are recorded.

[0058] The construction drawings are extracted from the template boundary lines and dimensions in the DWG using CAD. The sensor data includes stress, temperature and displacement, which are acquired by stress sensors, temperature sensors and inductive displacement sensors, respectively.

[0059] Unifying the coordinates and aligning the spatiotemporal data for all data includes:

[0060] To convert global coordinates to a local construction coordinate system, with the center of the shallow circular silo as the origin and the z-axis pointing vertically upwards, the conversion formula is:

[0061]

[0062] In the formula: Let O be the origin of the local coordinate system. The rotation angle of the coordinate system. Global coordinates.

[0063] At the same time, the data is interpolated based on the timestamp of the sensor data.

[0064] Extracting geometric, physical, and dynamic features from standardized datasets includes:

[0065] The construction area is divided into multiple grids, the boundary line segments of the template are detected, the boundary line set is obtained, the cylindrical surface is fitted by the least squares method, the center coordinates and radius are calculated, and the geometric features are obtained.

[0066] The lateral pressure on concrete is calculated based on geometric features and sensor data. The calculation formula is as follows:

[0067]

[0068] In the formula: Where h is the unit weight of concrete and h is the height of the formwork. This is the coefficient of earth pressure at rest.

[0069] Meanwhile, the stiffness of the template is calculated based on formulas from mechanics of materials:

[0070]

[0071] In the formula: B is the strain matrix, and D is the material elasticity matrix (E=210Gpa, v=0.3).

[0072] The physical characteristics are constituted by the lateral pressure and stiffness matrix of the concrete.

[0073] In addition, the dynamic characteristics are calculated from sensor data and meteorological data. The dynamic characteristics include wind load and deformation, and their calculation formulas are as follows:

[0074] Wind load:

[0075]

[0076] In the formula: Let A be the air density and A be the windward area of ​​the template. is the drag coefficient, with a value of 1.3, and v is the wind speed;

[0077] Deformation amount:

[0078]

[0079] In the formula: The coefficient of thermal expansion is... This is the initial length of the template. This refers to the temperature difference.

[0080] S102: As Figure 5As shown, the point cloud dataset and the BIM dataset are geometrically registered to generate a registered point cloud dataset. The physical features and dynamic features are associated with the registered point cloud dataset to obtain a joint feature vector set. Then, the registered point cloud dataset is weighted to obtain a weighted point cloud dataset.

[0081] As a specific implementation method, the geometric registration of the point cloud dataset and the BIM dataset involves the following steps:

[0082] Extract normal vectors from the point cloud dataset, calculate curvature, and select high curvature points as point cloud feature points to obtain a point cloud feature point set;

[0083] Applying template plane equations based on BIM dataset Extract BIM feature points to obtain a BIM feature point set;

[0084] In the formula: x, y, z represent the coordinates of a point in three-dimensional space, a, b, c are normal vectors, and d is a weighted constant term for the distance from the plane to the origin.

[0085] Randomly select point cloud features, search for the corresponding nearest neighbor points in the BIM feature points, and obtain the coarsely registered point cloud dataset through the transformation matrix;

[0086] The above steps aim to select easily identifiable key points from the point cloud dataset and BIM dataset to facilitate rapid registration in the future.

[0087] Based on the coarsely registered point cloud dataset, the BIM dataset is gradually processed to obtain an initial set of point pairs. Constraints and filters are applied to this initial set to generate a constrained and filtered set of point pairs. By minimizing the objective function, the registered point cloud dataset is obtained.

[0088]

[0089] In the formula: R is the rotation matrix, and t is the translation vector. For each iteration, the updated point pair set, For the points in the coarsely registered point cloud dataset, These are points in the BIM dataset.

[0090] Among them, the initial point pair set is constrained and filtered, that is, point pairs whose normal vectors exceed the constraint threshold and point pairs whose flatness error exceeds 1.5mm are removed. The constraint threshold can be set independently according to the registration accuracy, and this application does not limit it.

[0091] The process of associating physical features, dynamic features, and the registered point cloud dataset includes:

[0092] The key points of the template are extracted from the registered point cloud dataset to obtain a key point set, which includes the three-dimensional coordinates, normal vectors and curvature values ​​of the key points.

[0093] Key points for splicing include the corner points of the seams, the coordinates of the center of the circle, and the points with high curvature.

[0094] Calculate the concrete lateral pressure at the key point based on its vertical height in the three-dimensional coordinates; calculate the wind load component based on the normal vector; and calculate the thermal deformation based on the thermal expansion coefficient and temperature difference of the formwork material.

[0095] The key point set, concrete lateral pressure, wind load component, and thermal deformation are combined to form a joint feature vector set.

[0096] Furthermore, to avoid the accumulation of errors affecting the overall result, this embodiment embeds a geometric constraint equation: center offset. ,in accordance with By dynamically adjusting the registration weights, a weighted point cloud dataset is obtained.

[0097]

[0098] In the formula: This is the attenuation coefficient, with a value of 0.5.

[0099] S103: Obtain material parameters and combine them with physical characteristics to form mechanical constraints. Construct a process cause-effect graph based on the joint feature vector set and construction sequence. On the basis of mechanical constraints, transform template parameters into a set of legal symbols, design a reward function, and obtain the initial assembly strategy.

[0100] It should be added that obtaining material parameters and combining them with physical characteristics to form mechanical constraints specifically includes:

[0101] Transform the lateral pressure formula into a regularization term of the loss function:

[0102]

[0103] Transform the buckling critical load formula into a regularization term:

[0104]

[0105] In the formula: , These are the weighting coefficients. This is the predicted value of lateral pressure. This represents the actual load currently borne by the template. This is the formula for the critical buckling load. Let I be the template elastic modulus, I be the moment of inertia of the section, K be the effective length coefficient, and L be the actual length.

[0106] For example, constructing the process causal graph based on the joint feature vector set and construction sequence includes:

[0107] Each construction action is treated as a node, and the physical attributes of the construction action and the key features in the joint feature vector set are bound to the node to generate a node set;

[0108] Based on the construction sequence, the sequential dependencies in the construction work are defined as edges, and the weights are calculated to obtain the edge set.

[0109] The weights are calculated as follows:

[0110] Based on the construction sequence rule, if action i must precede action j, then the edge weight w ij =1;

[0111] Based on the force transmission path, if action i directly affects the force state of action j, then the edge weight w ij Calculated based on the load distribution ratio in historical data.

[0112] Connecting the set of nodes and the set of edges yields the process cause-effect graph.

[0113] Furthermore, the construction of the process causal graph based on the joint feature vector set and construction sequence also includes:

[0114] The node features in the process causal graph are initialized using a graph attention network. For each node, neighbor nodes are aggregated, and the feature vectors of these neighbor nodes are collected to capture the correlation between each construction action.

[0115]

[0116] In the formula: Let be the set of neighboring nodes of a node. Attention coefficient For parameter matrices, This is the activation function.

[0117] In this embodiment, the activation function Select the modified linear unit function:

[0118]

[0119] That is, if the input x is a positive number, the input equals the output; if the input x is a negative number or 0, the output is 0.

[0120] And parameter matrix The specific meaning is:

[0121] The input dimension d represents the dimension of the initial feature vector of each node. This represents the dimension of the new feature space after the linear transformation.

[0122] Among them, attention coefficient The calculation process is as follows:

[0123] Calculate the correlation between a node and its neighboring nodes:

[0124]

[0125] In the formula: a is the learnable attention vector. This is the activation function, specifically defined as follows:

[0126]

[0127] In the formula: x is the input value, The slope of the negative input region, with a value of 0.01;

[0128] If the input x is positive, the input equals the output; if the input x is negative, it is multiplied by 0.01 to improve robustness.

[0129] It should be added that the input layer of the graph attention network is the feature vector of each node in the process causal graph;

[0130] The graph attention layer performs the following operations: it calculates the attention coefficient between each node and its neighboring nodes, aggregates the features of the neighboring nodes using attention weighting, and introduces the LeakyReLU activation function for nonlinear transformation.

[0131] Output layer: Outputs the logical vector of each node. This is used to represent the logical characteristics between construction actions.

[0132] The training method used is supervised training, which is existing technology and will not be elaborated on here.

[0133] Then, normalization is performed to obtain the attention coefficient. .

[0134] Meanwhile, the design reward function includes:

[0135] The reward function is divided into an efficiency term, an accuracy term, and a safety term. The efficiency term is defined as the reciprocal of the assembly time, the accuracy term is the reciprocal of the deviation, and the safety term is defined as the difference between the allowable stress and the actual stress.

[0136] The reward value is obtained by weighting and summing the efficiency, accuracy, and security items separately, with weights of 0.4, 0.3, and 0.3 respectively. The calculation formula is as follows:

[0137]

[0138] In the formula: For assembly time, For deviation, To allow stress, For actual stress, 、 、 These are the weighting coefficients for the three items.

[0139] In this embodiment, the initial assembly strategy includes:

[0140] The assembly process is broken down into discrete and continuous actions;

[0141] For discrete actions, the assembly order is encoded as an integer; for continuous actions, fine-tuning displacements are set; and discrete and continuous actions are combined into an action vector.

[0142] An initial strategy is generated based on the current assembly state, and the relative benefit of a certain assembly action in the current assembly state is measured:

[0143]

[0144] In the formula: Given the expected reward of choosing action a under the current assembly state s, The baseline value for the current assembly state s;

[0145] By calculating relative returns, the initial strategy is updated using the maximization objective function to obtain the initial assembly strategy.

[0146] The step of updating the initial policy by maximizing the objective function includes:

[0147]

[0148] In the formula: The probability of the action in the new strategy. This represents the probability of the action under the old strategy. For the clip parameter.

[0149] In addition, the set of legal symbols is a set of formalized symbols, which includes action symbols, state symbols, relation symbols, logical connectors, and quantifier symbols.

[0150] Action symbols represent specific operational steps in the construction process, such as installing formwork, adjusting joints, and pouring concrete.

[0151] Status symbols: represent the status, attributes, or results during the construction process, used to describe the effect after an action is performed, or the current environmental state of the construction.

[0152] Relational symbols: These indicate the relationships between actions, between actions and states, and between states, such as relational conjunctions like cause, need, and conflict;

[0153] Logical connectors: These represent logical connectives, such as AND, OR, and NOT.

[0154] Quantifiers are used to express generalization or existence judgments, that is, whether a rule applies to all cases, or whether there are certain cases that satisfy the condition. For example, existence means that at least one action or state satisfies a certain condition.

[0155] S104: Combine mechanical constraints to perform multi-dimensional verification of the initial assembly strategy to obtain the final assembly strategy;

[0156] The multi-dimensional verification includes: misalignment verification, stress verification, and dynamic verification.

[0157] As a specific example, the specific steps for verifying the misalignment amount include:

[0158] Extract the coordinates of key points at the template joints, and calculate whether the spacing between adjacent templates meets the width and whether the height difference between adjacent template surfaces meets the constraints.

[0159] It should be noted that the width and height difference constraints are ±5mm and <2mm, respectively, and the basis for these values ​​can be found in the "Code for Acceptance of Construction Quality of Concrete Structures".

[0160] Stress verification includes: discretizing the template geometry into tetrahedral elements, defining material properties, and applying load boundary conditions;

[0161] Solve the linear statics equations to obtain the stress distribution results. Based on the stress distribution results, determine whether the material's yield strength is satisfied. The linear statics equations are:

[0162]

[0163] In the formula: Here is the stiffness matrix. Let F be the nodal displacement and F be the load vector.

[0164] Dynamic verification includes:

[0165] Simulation of collision and friction at the template contact surface using the discrete element method:

[0166]

[0167] In the formula: , For normal stiffness and damping coefficient, , For normal contact depth and velocity, The coefficient of friction, The tangential relative velocity, It is the absolute value of the normal force, i.e., the contact pressure.

[0168] The yield strength varies depending on the material, and this application does not impose any restrictions on it.

[0169] Example 2

[0170] Please see Figure 2 As shown, this embodiment discloses an electronic device, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the intelligent pre-assembly method for shallow circular silo sliding formwork template based on multi-source data fusion provided by the above methods.

[0171] Since the electronic device described in this embodiment is the one used to implement the intelligent pre-assembly method for shallow circular warehouse sliding formwork based on multi-source data fusion in this application embodiment, those skilled in the art can understand the specific implementation methods and various variations of the electronic device in this embodiment based on the intelligent pre-assembly method for shallow circular warehouse sliding formwork based on multi-source data fusion described in this application embodiment. Therefore, how the electronic device implements the method in this application embodiment will not be described in detail here. Any electronic device used by those skilled in the art to implement the intelligent pre-assembly method for shallow circular warehouse sliding formwork based on multi-source data fusion in this application embodiment falls within the scope of protection of this application.

[0172] Example 3

[0173] Please see Figure 3 As shown, this embodiment discloses a computer-readable storage medium, including a memory, a processor, and a computer program stored on the memory and running on the processor. When the processor executes the computer program, it implements the intelligent pre-assembly method for shallow circular silo sliding formwork template based on multi-source data fusion provided by the above methods.

[0174] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters, weights, and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0175] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via a wired or wireless network. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0176] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed in this invention can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0177] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0178] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0179] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

[0180] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A multi-source data fusion-based intelligent pre-assembly method for a silo slip-form template, characterized in that, The method comprises: S101: Obtain multi-source data, parse and clean the multi-source data to obtain a standardized data set, and extract geometric, physical and dynamic characteristics from the standardized data set; wherein the standardized data set comprises a point cloud data set, a BIM data set and a drawing geometry data set; S102: Geometrically register the point cloud data set and the BIM data set to generate a registered point cloud data set, associate physical characteristics, dynamic characteristics with the registered point cloud data set, obtain a joint feature vector set, and weight the registered point cloud data set to obtain a weighted point cloud data set; S103: Obtain material parameters combined with physical characteristics to form mechanical constraints, construct a process causal diagram according to the joint feature vector set and a construction sequence, convert template parameters into a legal symbol set based on the mechanical constraints, design a reward function, and obtain an initial assembly strategy; S104: Perform multi-dimensional verification on the initial assembly strategy in combination with the mechanical constraints to obtain a final assembly strategy; Wherein the association of the physical characteristics, the dynamic characteristics and the registered point cloud data set comprises: Extracting splicing key points of the template from the registered point cloud data set to obtain a key point set, the key point set comprising three-dimensional coordinates, normal vectors and curvature values of the key points; Calculating the concrete side pressure of the point according to the vertical height in the three-dimensional coordinates of the key points, calculating the wind load component according to the normal vector, and calculating the thermal deformation amount according to the thermal expansion coefficient and the temperature difference of the template material; Combining the key point set, the concrete side pressure, the wind load component and the thermal deformation amount to form a joint feature vector set; The process causal diagram is constructed according to the joint feature vector set and the construction sequence, comprising: Regarding each construction action as a node, binding the physical properties in the construction action and the key point features in the joint feature vector set to the node to generate a node set; Defining the precedence and dependency relationship in the construction work as an edge according to the construction sequence, and performing weight calculation to obtain an edge set; Connecting the node set and the edge set to obtain the process causal diagram, initializing the node features in the process causal diagram, aggregating the neighbor nodes for each node, and collecting the feature vectors of the neighbor nodes: wherein: is a set of neighbor nodes of the node, is an attention coefficient, is a parameter matrix, is an activation function.

2. The multi-source data fusion-based intelligent pre-assembly method for the silo slip-form template according to claim 1, characterized in that, The geometric registration of the point cloud data set and the BIM data set comprises the following specific steps: Extracting normal vectors from the point cloud data set, calculating curvature, selecting high-curvature points as point cloud feature points, and obtaining a point cloud feature point set; Applying a template plane equation based on the BIM data set to extract BIM feature points and obtain a BIM feature point set; Randomly selecting a point cloud feature and searching for the corresponding nearest neighbor point in the BIM feature point set to obtain a coarsely registered point cloud data set through a transformation matrix; Gradually filtering the BIM data set based on the coarsely registered point cloud data set to obtain an initial point pair set, performing constraint and filtering on the initial point pair set to generate a constraint and filtered point pair set, and obtaining a registered point cloud data set by minimizing an objective function: where R is a rotation matrix and t is a translation vector, is the set of updated point pairs for each iteration, is the point in the point cloud dataset after coarse registration, is the point in the BIM dataset.

3. The multi-source data fusion-based intelligent pre-assembly method for the silo slip-form template according to claim 1, characterized in that, The design of the reward function comprises: The reward function is divided into an efficiency term, an accuracy term and a safety term, the efficiency term is defined as the reciprocal of the assembly time, the accuracy term is the reciprocal of the deviation, and the safety term is defined as the difference between the allowable stress and the actual stress; The efficiency term, the accuracy term and the safety term are respectively weighted and summed to obtain a reward value.

4. The multi-source data fusion-based intelligent pre-assembly method for the silo slip-form template according to claim 1, characterized in that, The initial assembly strategy comprises: The assembly action is split to obtain discrete actions and continuous actions; The assembly sequence is encoded as an integer for the discrete actions, a fine adjustment displacement is set for the continuous actions, and the discrete actions and the continuous actions are combined into an action vector; An initial strategy is generated based on a current assembly state, and the relative income of a certain assembly action under the current assembly state is measured: wherein: is the expected return for a selected action a at the current assembly state s, is the baseline value for the current assembly state s; The initial strategy is updated by maximizing the objective function through calculating the relative income, and the initial assembly strategy is obtained.

5. The multi-source data fusion-based intelligent pre-assembly method for the silo slip-form template according to claim 4, characterized in that, The updating of the initial strategy by maximizing the objective function comprises: In the formula: is the action probability of the new policy, represents the action probability of the old policy, is the clip parameter.

6. An electronic device comprising a memory, a processor, and a computer program stored on the memory and running on the processor, characterized in that, The processor executes the computer program to realize the intelligent pre-assembly method of the shallow circular silo slip-form template based on multi-source data fusion in any one of claims 1 to 5.

7. A computer readable storage medium characterized in that, The computer program is stored on the computer readable storage medium, and the computer program is executed to realize the intelligent pre-assembly method of the shallow circular silo slip-form template based on multi-source data fusion in any one of claims 1 to 5.

Images