Precast beam yard layout optimization system based on BIM and parametric database integration

Abstract

The present invention discloses a precast beam yard layout optimization system based on BIM and parametric database integration, which relates to the technical field of computer-aided design. The system includes: a BIM data acquisition and processing unit, which is used to extract geometric feature tensors, material feature tensors and cross-sectional feature tensors of precast beams from the BIM parametric database, and construct a multi-scale feature tensor of precast beams; a model construction unit, which is used to establish a multi-physics field coupling model of a site based on multi-scale tensors and site parameters; based on the multi-physics field coupling model, calculate the dynamic bearing capacity distribution; calculate the dynamic feasibility index; a layout optimization unit, which is used to convert the optimization problem into a continuous spatiotemporal Markov decision process, find the instruction matrix of the crane corresponding to the optimal strategy function, as the final execution plan, to complete the optimization layout of the precast beam yard. The present invention ensures the safety and stability of the layout plan in a complex construction environment, maximizes the construction efficiency and reduces the risk.

Description

Technical Field

[0001] The present invention relates to the technical field of computer-aided design, and in particular to a prefabricated beam yard layout optimization system based on the integration of BIM and a parametric database. Background Art

[0002] In the construction of the precast beam yard layout, a reasonable, safe and efficient layout plan is crucial to the smooth implementation of the project. As an important part of large-scale infrastructure such as bridges and buildings, precast beams have the characteristics of heavy weight, complex geometric dimensions, and uneven stress distribution. The layout of the precast beam yard must not only consider the utilization of the on-site space, but also ensure the stability and safety of each precast beam in its placement position. At the same time, as the construction progresses, site conditions (such as bearing capacity, groundwater infiltration, temperature changes, etc.) will continue to change. How to maintain the rationality of the layout in a dynamic environment is a key issue that needs to be solved. Therefore, it is of great technical significance to build a system that can dynamically respond to environmental changes and achieve the optimal layout.

[0003] In the prior art, the layout optimization of the precast beam yard mainly relies on traditional empirical methods and heuristic rules, which usually use static planning and manual adjustment to determine the placement of precast beams. Although this static layout method can meet basic construction requirements in relatively simple and stable sites, its limitations are very obvious under complex and changeable site conditions. For example, the traditional precast beam layout usually relies on a preliminary estimate of the foundation bearing capacity, and the bearing capacity calculation is often based on simple soil properties, compaction and static load conditions. However, in actual construction, the site bearing capacity will change significantly due to factors such as hydrological conditions, groundwater level fluctuations, progressive compression of the soil, and mechanical vibration during construction. In this case, the traditional static bearing capacity estimation is difficult to reflect the changes in the foundation bearing capacity in real time, which can easily lead to insufficient bearing capacity, which in turn causes the precast beam to sink, tilt, and even serious foundation instability accidents. Especially in the rainy season or in areas with large water level fluctuations, the impact of groundwater infiltration and water pressure changes on the foundation bearing capacity is particularly prominent, and traditional methods are difficult to flexibly respond to these changes. Summary of the invention

[0004] The purpose of the present invention is to provide a precast beam yard layout optimization system based on the integration of BIM and parametric database, which ensures the safety and stability of the layout scheme in a complex construction environment, maximizes construction efficiency and reduces risks.

[0005] In order to solve the above technical problems, the present invention provides a precast beam site layout optimization system based on the integration of BIM and parameterized database, the system comprising: a BIM data acquisition and processing unit, used to extract the geometric feature tensor, material feature tensor and cross-sectional feature tensor of the precast beam from the BIM parameterized database, and respectively use the geometric feature tensor, material feature tensor and cross-sectional feature tensor as an element in a vector to construct a multi-scale feature tensor of the precast beam; a model construction unit, used to establish a site multi-physics field coupling model based on geomechanics, hydrology and heat conduction based on the multi-scale tensor and site parameters; The field coupling model is used to calculate the dynamic bearing capacity distribution; based on the multi-scale tensor and site parameters, the dynamic feasibility index is calculated for each precast beam and position combination; the layout optimization unit is used to transform the optimization problem into a continuous spatiotemporal Markov decision process, based on the dynamic bearing capacity distribution and dynamic feasibility index obtained by the model building unit, define the objective function of the optimization problem, use the deep deterministic policy gradient algorithm based on the Actor-Critic architecture to solve the optimization problem, find the command matrix of the crane corresponding to the optimal policy function, and obtain the final execution plan to complete the optimization layout of the precast beam yard.

[0006] Furthermore, when the layout optimization unit transforms the optimization problem into a continuous spatiotemporal Markov decision process, the state space is defined as:

[0007] ,

[0008] The action space is:

[0009] ;

[0010] The reward function is:

[0011] ;

[0012] in, express System status at the moment; is the decision function, is the gradient of the decision function, precast beam The decision function is , indicating precast beam In time Place in position probability; for Time location Dynamic bearing capacity distribution at precast beams exist Time location The dynamic bearing capacity distribution at ; for Time location Dynamic feasibility index of precast beams exist Time location The dynamic feasibility index at ; for The command matrix of the crane at time , each element in the command matrix represents the An execution action parameter of the crane at the moment; for System actions at all times; To perform actions Back from state Rewards received; is the gradient of the objective function of the optimization problem.

[0013] Furthermore, when the layout optimization unit uses a deep deterministic policy gradient algorithm based on the Actor-Critic architecture to solve the optimization problem, the Actor network is represented as:

[0014] ,

[0015] in The parameters are Actor network with input as state , the output is a deterministic action;

[0016] The Critic network is represented as:

[0017] ,

[0018] in The parameters are Critic network, the input is state and actions , the output is the state-action value function value, represents the real number space;

[0019] The strategy update process is expressed as:

[0020] ;

[0021] in It is a behavioral strategy The resulting state distribution is is the policy gradient, Behavioral Strategy The resulting state distribution The expectations below, is the gradient of the Critic network to the action; is the gradient of the Actor network with respect to its parameters;

[0022] The value function update process is expressed as:

[0023] ,

[0024] in, is the loss function of the Critic network, For the experience replay buffer, is the target Q value, For instant rewards, is the discount factor, and are the target network parameters, From the buffer Randomly draw multiple samples from Calculate the loss value of all sampled samples Take expectations;

[0025] To handle continuous space-time domains, use function approximation:

[0026] ;

[0027] in is the selected radial basis function, For the Precast beam weights, For the Radial basis functions; is the number of radial basis functions.

[0028] Furthermore, the layout optimization unit uses a deep deterministic policy gradient algorithm based on the Actor-Critic architecture to solve the optimization problem. The process of obtaining the final layout solution specifically includes:

[0029] Step 1: Initialize the Actor Network and Critic Network ;

[0030] Step 2: Initialize target network parameters , ;

[0031] Step 3: Initialize the experience replay buffer ;

[0032] Step 4: Learn the optimal precast beam layout strategy by repeatedly interacting with the simulation environment. In this process, select actions based on the current state and the learned strategy, execute these actions and observe the results, and then store the experience in the experience playback buffer; update the Critic network corresponding to its value estimation and the Actor network corresponding to the decision strategy by randomly sampling from these experiences; optimize the layout strategy and maximize the objective function;

[0033] Step 5: Calculate the objective function value of the current layout solution ; Calculate rewards ; Update status Includes: Update Use Actor network output; recalculate ;and ;

[0034] Step 6: Repeat steps 4 to 5 until the set number of times is reached to obtain the prefabricated beam. The optimal decision function is:

[0035] ;

[0036] After determining the decision function, the corresponding crane instruction matrix As the final implementation plan.

[0037] Furthermore, step 4 specifically includes:

[0038] Step 4.1: Initialize the spatiotemporal noise process For motion exploration;

[0039] Step 4.2: Receive initial observation state ;

[0040] Step 4.3: For To the maximum time , perform the following process:

[0041] Select Action Execute actions based on current policy and exploration noise , observe the reward and the new state ; will convert Save to ;from Randomly sample a mini-batch of transformations ;set up ; is the subscript index; the parameters of the Critic network are updated by the following formula:

[0042] ;

[0043] Update the parameters of the Actor network using the following formula:

[0044] ;

[0045] The parameters of the target network are soft-updated using the following formula:

[0046] , ;

[0047] It is a soft update parameter.

[0048] Furthermore, the prefabricated beams constructed by the BIM data acquisition and processing unit The multi-scale feature tensor of is:

[0049] ;

[0050] in, is the geometric feature tensor, where For long, For width, for high; is the material characteristic tensor, where , and The location Elastic modulus, Poisson's ratio and density at ; is the cross-section characteristic tensor, where , and are the moment of inertia, torsion constant and section modulus distributions, respectively.

[0051] Furthermore, the model building unit, based on the multi-scale tensor and site parameters, establishes a site multi-physics field coupling model based on geomechanics, hydrology and heat conduction through the following formula:

[0052] ;

[0053] ;

[0054] ;

[0055] in, is the gradient operator; is the total stress tensor; is the displacement vector; is the water head; is temperature; is the permeability coefficient; is the specific water storage rate; is the Biot coefficient; is the thermal conductivity coefficient; is the specific heat capacity; is the heat source term; for The second derivative of for The first derivative of For the second order norm; is the foundation density; For tensors rank, is the norm operation.

[0056] Furthermore, the model building unit calculates the first Dynamic bearing capacity distribution of precast beams The formula is as follows:

[0057] ;

[0058] in, is the ultimate bearing capacity distribution; is the water head distribution; is the temperature distribution; is the terrain elevation function; , , and are critical stress, hydraulic head, temperature and slope respectively; is the absolute value operator; is the initial temperature.

[0059] Furthermore, the model building unit, based on the multi-scale tensor and site parameters, calculates the first Dynamic feasibility index of precast beams :

[0060] ;

[0061] in, is the Heaviside step function; is the available area; For precast beams Stress distribution of is the first-order natural frequency of the coupled system; is the critical frequency difference; For the Precast beams in position Yield stress; For the The quality of the precast beams; For the The first order natural frequency of each precast beam; For the The spatial domain occupied by a precast beam.

[0062] Furthermore, the objective function of the optimization problem defined by the layout optimization unit is expressed using the following formula:

[0063] ;

[0064] in, is the total number of precast beams.

[0065] The precast beam yard layout optimization system based on BIM and parametric database integration of the present invention has the following beneficial effects: the traditional precast beam yard layout scheme is usually based on static planning, while the present invention realizes real-time optimization in the layout process by acquiring data from BIM and parametric database in real time and combining multi-physical field coupling model. Through this system, the geometric characteristics, material properties and cross-sectional characteristics of the precast beam can be dynamically coupled and analyzed with the physical parameters of the site (such as bearing capacity, topography, groundwater level, temperature, etc.). This dynamic coupling enables the system to cope with changes in foundation bearing capacity, groundwater level fluctuations and other environmental conditions during construction. Whenever the physical conditions of the site change, the system can recalculate the layout plan according to real-time data to ensure that the placement position of the precast beam is always in a safe and stable state. Such dynamic response capability not only improves construction efficiency, but also significantly reduces the potential risks caused by unpredictable environmental changes. The present invention realizes accurate evaluation and optimization of precast beams in site layout by calculating the dynamic feasibility index and bearing capacity distribution of precast beams. The dynamic feasibility index combines bearing capacity, site geometry, stress distribution and structural vibration response to comprehensively evaluate the feasibility of precast beams at specific locations and times. Through this dynamic feasibility assessment, the system can avoid placing precast beams in low bearing capacity areas or high stress areas, ensuring that the layout plan always meets safety and stability requirements. In addition, the introduction of dynamic bearing capacity distribution allows the system to continuously adjust the layout plan according to changes in the site's bearing capacity, avoiding potential dangers caused by insufficient bearing capacity in traditional layout plans. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0067] Figure 1 A schematic diagram of the system structure of a precast beam yard layout optimization system based on BIM and parametric database integration provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0068] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0069] Example 1: Reference Figure 1 , a precast beam yard layout optimization system based on the integration of BIM and parametric database, the system comprising: a BIM data acquisition and processing unit, used to extract the geometric feature tensor, material feature tensor and cross-sectional feature tensor of the precast beam from the BIM parametric database, and use the geometric feature tensor, material feature tensor and cross-sectional feature tensor as an element in a vector respectively to construct a multi-scale feature tensor of the precast beam; a model construction unit, used to establish a site multi-physics field coupling model based on geomechanics, hydrology and heat conduction based on the multi-scale tensor and site parameters; based on the multi-physics field coupling model, calculate the dynamic bearing capacity distribution; based on the multi-scale tensor and site parameters, calculate the dynamic feasibility index for each precast beam and position combination; a layout optimization unit, used to transform the optimization problem into a continuous spatiotemporal Markov decision process, define the objective function of the optimization problem based on the dynamic bearing capacity distribution and dynamic feasibility index obtained by the model construction unit, solve the optimization problem using a deep deterministic policy gradient algorithm based on the Actor-Critic architecture, find the instruction matrix of the crane corresponding to the optimal policy function, as the final execution plan, to complete the optimization layout of the precast beam yard.

[0070] Specifically, the BIM data acquisition and processing unit is responsible for extracting the geometric feature tensor, material feature tensor and cross-sectional feature tensor of the precast beam from the BIM parametric database. These feature tensors represent the different properties of the precast beam in terms of geometric structure, material properties and cross-sectional shape, respectively, and constitute the multi-scale feature information of the precast beam. The introduction of BIM technology provides accurate and comprehensive basic data support for the layout optimization of precast beams. In the existing layout optimization technology, data acquisition usually relies on manual measurement or static architectural design drawings, which is time-consuming and cannot adapt to dynamic changes in actual construction. BIM technology integrates the geometric, spatial and material information of buildings and their components in the form of three-dimensional digital models, greatly improving the efficiency of automatic data acquisition and ensuring the accuracy of data. By extracting the multi-scale feature tensor of the precast beam, the system of the present invention can accurately capture the detailed information of the geometric form, material properties and cross section of each precast beam. The geometric feature tensor includes geometric information such as the length, width and height of the precast beam, which determines the occupied space and installation conditions of the precast beam in the site. The material characteristic tensor describes the physical and mechanical properties of the precast beam, such as the elastic modulus, density, compressive strength, etc. of the material, ensuring the stability and bearing capacity of the precast beam under different stress conditions. The cross-sectional characteristic tensor further refines the local structure of the precast beam, especially playing a vital role in the distribution of bearing capacity and structural strength assessment. These tensor data are integrated together in the form of multidimensional vectors to form a multi-scale feature description of the precast beam. Compared with the existing technology, this tensor-based multi-scale feature representation method can better handle precast beams with complex shapes and properties, and improves the flexibility and adaptability of the system in dealing with different sizes, materials and cross-section combinations.

[0071] After obtaining the multi-scale characteristic tensor of the precast beam, the model building unit of the present invention establishes a multi-physical field coupling model of the site based on geomechanics, hydrology and heat conduction by combining the multi-scale characteristic tensor and site parameters. The traditional precast beam site layout optimization method usually only considers a single physical field, such as geological conditions or static bearing capacity. Such a model is difficult to fully reflect the real situation when dealing with complex sites. Especially when the construction site has unstable geological conditions or extreme climatic conditions, the single physical field model is prone to deviation, which leads to unreasonable layout plans and may even bring safety hazards. The multi-physical field coupling model of the present invention fully considers the interaction between multiple physical fields such as geomechanics, hydrological conditions and heat conduction, and can dynamically simulate the changes of the site under different times and conditions. The geomechanical model is mainly used to evaluate the bearing capacity and stability of the foundation of the site, combined with the weight and structural characteristics of the precast beam, to ensure the stability of each precast beam at a specific position. The hydrological model is used to analyze the impact of factors such as groundwater level, permeability and humidity on the site, especially in the rainy season or in areas with high groundwater levels, changes in hydrological conditions may greatly affect the bearing capacity and structural safety of the site. The heat conduction model is used to evaluate the thermal stress of the site materials and the thermal expansion effect of the precast beams under different temperature conditions. Especially in high or low temperature environments, the heat exchange between the precast beams and the site may cause additional stress or deformation, which is often overlooked in traditional layout optimization methods. The present invention can comprehensively evaluate the dynamic changes of the site by coupling these physical fields to ensure the long-term stability and safety of the layout scheme. After calculating the dynamic bearing capacity distribution of the site based on the multi-physical field coupling model, the system further combines the multi-scale characteristic tensor of the precast beams to calculate the dynamic feasibility index of each precast beam and position combination. The dynamic feasibility index is a new evaluation standard that comprehensively considers factors such as the size, weight, material properties of the precast beams, the bearing capacity, geological stability and operating space of the site, and aims to evaluate the installability and operability of each precast beam at a specific location. In the prior art, the feasibility evaluation of the layout usually adopts static geometric calculations and local bearing capacity analysis, which is difficult to adapt to changes in site conditions. The present invention, through the introduction of the dynamic feasibility index, can not only reflect the changes of the site in different time periods in real time, but also predict possible risks in the future construction process. For example, the feasibility index of a precast beam is high at the current moment, but as the groundwater level rises or the temperature changes, the bearing capacity of the location may decrease, resulting in a lower feasibility index. By calculating the dynamic feasibility index, the system can detect these potential problems in advance, avoid unreasonable layout choices, and improve the safety and sustainability of the layout plan.

[0072] In the layout optimization unit, the present invention adopts a continuous spatiotemporal Markov decision process to solve the layout optimization problem. Traditional layout optimization algorithms usually rely on linear programming or heuristic algorithms, which have certain effects when dealing with static problems, but often seem to be powerless when facing dynamic, multi-variable layout optimization problems. The continuous spatiotemporal Markov decision process is an optimization method based on probabilistic state transfer and reward mechanism, which can effectively deal with uncertainty and dynamic changes in the layout process. In the system of the present invention, the dynamic bearing capacity distribution and dynamic feasibility index of the site are regarded as state variables in the decision-making process, and the layout operation of each prefabricated beam is regarded as an action. Through the Markov decision process, the system can select the optimal action at each time step and predict future state changes. This method can not only handle complex dynamic environments, but also continuously optimize the layout plan over time. After defining the objective function of layout optimization, the system uses a deep deterministic policy gradient algorithm based on the Actor-Critic architecture for solution. Although traditional optimization algorithms such as genetic algorithms or particle swarm algorithms can solve layout optimization problems to a certain extent, since these algorithms rely on random search, the solution efficiency is low and it is difficult to handle dynamic changes in high-dimensional space. The deep deterministic policy gradient algorithm can efficiently solve the optimal strategy in continuous space by combining the advantages of deep learning and reinforcement learning. The Actor network in the Actor-Critic architecture is responsible for generating layout instructions, while the Critic network evaluates these instructions to ensure the optimality and feasibility of the layout plan. In the present invention, the system finds the optimal policy function through the deep deterministic policy gradient algorithm, generates the instruction matrix of the crane, and executes the layout plan accordingly. Compared with traditional heuristic algorithms, the deep deterministic policy gradient algorithm not only has higher solution efficiency, but also can maintain high robustness and adaptability in the face of dynamic changes.

[0073] Embodiment 2: When the layout optimization unit transforms the optimization problem into a continuous spatiotemporal Markov decision process, the state space is defined as:

[0074] ,

[0075] The action space is:

[0076] ;

[0077] The reward function is:

[0078] ;

[0079] in, express System status at the moment; is the decision function, is the gradient of the decision function, precast beam The decision function is , indicating precast beam In time Place in position probability; for Time location Dynamic bearing capacity distribution at precast beams exist Time location The dynamic bearing capacity distribution at ; for Time location Dynamic feasibility index of precast beams exist Time location The dynamic feasibility index at ; for The command matrix of the crane at time , each element in the command matrix represents the An execution action parameter of the crane at the moment; for System actions at all times; To perform actions Back from state Rewards received; is the gradient of the objective function of the optimization problem.

[0080] Specifically, the definition of the state space combines the probability of a precast beam at a specific location, the dynamic bearing capacity distribution, and the dynamic feasibility index to comprehensively describe the real-time status of the precast beam yard. The decision function of the precast beam reflects the probability of placing a specific precast beam at a certain location at a certain point in time, which is evaluated based on multi-dimensional input data such as geometric characteristics, material properties, and physical constraints of the construction site. In this dynamic layout system, the site bearing capacity and feasibility index are key input parameters. The dynamic bearing capacity distribution considers the impact of geological conditions, hydrological conditions, and other environmental factors at different locations and times on the site bearing capacity, ensuring that each precast beam can obtain sufficient foundation support when placed. At the same time, the dynamic feasibility index is used to evaluate the installation feasibility of the precast beam at a specific location, combining information such as the operating space, surrounding environmental constraints, and the operational feasibility of the crane. By integrating these dynamic information, the system can accurately evaluate the layout feasibility of the precast beam. The design of the action space reflects the specific operational decisions of the system at each time point, which mainly consists of the crane's operating instructions. The crane's instruction matrix determines how to adjust the placement of the precast beam within different time steps. In order to achieve the optimal layout, the system calculates the specific operations to be performed in real time according to the feedback of the state space in each time step. Traditional layout methods usually rely on static rules or single algorithms, which are difficult to cope with complex working conditions or changes in site conditions. The present invention converts the layout problem into a Markov decision process, allowing the system to make adaptive adjustments in each time step, ensuring that it can respond to environmental changes in real time during the construction progress and maintain the rationality and safety of the layout. The design of the reward function is directly related to the optimization goal of the system and is mainly used to measure the effectiveness of a certain layout operation. In each time step, the system evaluates the current state according to the operations performed and calculates the corresponding reward value. The core of the reward function is to determine whether a certain layout decision can improve the efficiency and safety of the overall layout through the gradient change of the objective function. By dynamically calculating the reward, the system can gradually optimize the layout decision and find the optimal solution for the precast beam yard layout. Based on this reward mechanism, the system can self-learn and adjust to continuously improve the overall quality of the layout.

[0081] Example 3: When the layout optimization unit uses a deep deterministic policy gradient algorithm based on the Actor-Critic architecture to solve the optimization problem, the Actor network is represented as:

[0082] ,

[0083] in The parameters are Actor network with input as state , the output is a deterministic action;

[0084] The Critic network is represented as:

[0085] ,

[0086] in The parameters are Critic network, the input is state and actions , the output is the state-action value function value;

[0087] The strategy update process is expressed as:

[0088] ;

[0089] in It is a behavioral strategy The resulting state distribution is is the policy gradient, Behavioral Strategy The resulting state distribution The expectations below, is the gradient of the Critic network to the action; is the gradient of the Actor network with respect to its parameters;

[0090] The value function update process is expressed as:

[0091] ,

[0092] in, is the loss function of the Critic network, For the experience replay buffer, is the target Q value, For instant rewards, is the discount factor, and These are target network parameters;

[0093] To handle continuous space-time domains, use function approximation:

[0094] ;

[0095] in is the selected radial basis function, For the Precast beam weights, For the Radial basis functions; is the number of radial basis functions.

[0096] Specifically, the Actor network plays a vital role in the entire optimization process. The task of the Actor network is to generate a deterministic action that represents the decision made by the system on the layout in the current state. Specifically, the system receives the current state information of the venue , which includes the geometric characteristics, material properties, cross-sectional properties of the precast beams, as well as the dynamic bearing capacity distribution and feasibility index of the site. This information reflects the overall situation of the precast beams and the site at the current time step and is the basis for the system to generate layout decisions. Based on these inputs, the Actor network outputs a deterministic action , this action indicates how the system decides to adjust the layout in the current state. For example, the action It can indicate how a crane should move a precast beam to a specific location, or how to adjust the placement of a precast beam according to the load-bearing capacity distribution of the site. Traditional layout optimization methods often rely on static planning algorithms that cannot fully cope with the dynamic changes in the site environment. Through deep reinforcement learning, the Actor network can adjust the layout of the site according to the current state at each time step. Dynamically generate layout decisions. This feature makes the system more flexible and adaptable, and can quickly adjust when the site conditions change. In addition, the actions generated by the Actor network are not random, but deterministic, which means that the system is based on the same state each time. The same actions are generated every time, ensuring repeatable and reliable layout decisions.

[0097] During the layout optimization process, the Critic network is responsible for evaluating the layout decisions generated by the Actor network. The input of the Critic network includes the current state and actions , output a Value, this The value is a state-action value function, which represents the long-term reward that the system can obtain after executing a certain layout action in the current state. The goal of the Critic network is to learn and estimate the effect of the current layout decision. For example, if a prefabricated beam is placed on a site with low bearing capacity, which may cause instability or safety hazards in the future, then the Critic network The value of will be relatively low, reflecting the negative impact of this decision. On the contrary, if the precast beam is reasonably placed in a location with high bearing capacity and high feasibility, the critic network The higher the value, the higher the system expects the layout to bring higher long-term rewards. The Critic network evaluates the actions of the Actor network, and the feedback information generated is in turn used to update the parameters of the Actor network, thereby gradually optimizing the layout strategy. Specifically, the policy gradient of the Actor network is calculated by To update, the feedback of the Critic network Help the Actor network learn how to generate actions better. After each layout decision, the system will adjust the Actor network based on the feedback of the Critic network to ensure that the layout decisions generated in the future can continue to approach the global optimal solution. This strategy update process allows the system to adaptively adjust the layout during the construction process and gradually improve the layout strategy through reinforcement learning. In this architecture, the loss function of the Critic network is It is one of the core of system learning. The Critic network calculates the current Values ​​and Goals The difference between the values ​​is used to update its own parameters. The value not only takes into account the immediate return of the current layout decision , including the discounted value of future returns ,in is a discount factor used to balance the weights of current rewards and future rewards. This means that when the system optimizes the layout, it not only pays attention to the effect of the current layout, but also predicts the long-term impact of the layout in the future construction process. Through this mechanism, the system can gradually improve the global optimality of the layout strategy. It is worth noting that the learning process of the Critic network is carried out through the experience replay mechanism. Experience replay buffer The system stores the state, action, and feedback information of the system over a period of time. The system is trained by randomly sampling data from the buffer. The purpose of this is to break the time correlation of the data and improve the generalization ability of the system. In the layout optimization scenario, site conditions, construction requirements, and characteristics of precast beams may change over time. Through the experience playback mechanism, the system can learn from past decisions and make more optimized decisions under new conditions. This mechanism greatly improves the learning efficiency and stability of the system, ensuring that the system can achieve the global optimal layout in a complex construction environment. In addition, in order to solve the continuous space-time problem in layout optimization, the system introduces radial basis functions as an approximation tool. Layout state of precast beams Through radial basis function The system generates layout decisions based on the linear combination of these basis functions. The introduction of radial basis functions can effectively capture the characteristics of spatial changes, especially when dealing with complex continuous space layout problems. The radial basis function provides a more flexible function approximation capability. This approximation technology can improve the layout accuracy of the system under complex site conditions and ensure that precast beams can be reasonably placed at different times and spatial positions.

[0098] Embodiment 4: The layout optimization unit uses a deep deterministic policy gradient algorithm based on the Actor-Critic architecture to solve the optimization problem. The process of obtaining the final layout solution specifically includes:

[0099] Step 1: Initialize the Actor Network and Critic Network ;

[0100] Step 2: Initialize target network parameters , ;

[0101] Step 3: Initialize the experience replay buffer ;

[0102] Step 4: Learn the optimal precast beam layout strategy by repeatedly interacting with the simulation environment. In this process, select actions based on the current state and the learned strategy, execute these actions and observe the results, and then store the experience in the experience playback buffer; update the Critic network corresponding to its value estimation and the Actor network corresponding to the decision strategy by randomly sampling from these experiences; optimize the layout strategy and maximize the objective function;

[0103] Step 5: Calculate the objective function value of the current layout solution ; Calculate rewards ; Update status Includes: Update Use Actor network output; recalculate ;and ;

[0104] Step 6: Repeat steps 4 to 5 until the set number of times is reached to obtain the prefabricated beam. The optimal decision function is:

[0105] ;

[0106] After determining the decision function, the corresponding crane instruction matrix As the final implementation plan.

[0107] Specifically, this method relies on the collaborative work of Actor and Critic networks to make layout decisions. The main task of the Actor network is to generate decision actions, that is, how to arrange precast beams in the current state. Including the geometric characteristics of the site, bearing capacity distribution , and the site feasibility index , all this information is obtained in real time through BIM and parametric database. , the system is based on the current state Generate Action , that is, the placement of the prefabricated beams and the instructions for the crane. This process enables the system to generate the optimal layout action based on the current environmental conditions through the training and optimization of the Actor network. At the same time, the Critic network Used to evaluate the effectiveness of the current layout decision. The Critic network evaluates the state and actions The combination of The value is the state-action value function. This value represents the long-term reward that the system may obtain after executing a certain layout decision in the current state. For example, when the precast beam is placed in a location with higher bearing capacity and better feasibility, the system will evaluate the long-term benefits of this layout decision, such as site stability, safety, and construction efficiency improvement, and thus assign a higher value to this decision. The task of the Critic network is to help the system learn how to evaluate different layout decisions and compare different value to optimize subsequent layout selection.

[0108] During the optimization process, the system learns the optimal layout strategy by continuously interacting with the simulation environment. In each time step, the system selects a layout action based on the current state, observes the results after executing the action, and stores these experiences in the experience replay buffer. By randomly sampling from the buffer, the system can break the problem of time correlation and improve the stability and generalization ability of training. This mechanism enables the system to learn from past layout experience and generate better layout decisions under new site conditions. When the system performs an action, it calculates the objective function value of the current layout plan based on the results. , this value indicates the overall optimization degree of the current layout plan, such as the utilization efficiency of the site, the reduction of construction time, and the rationality of the placement of precast beams. The change of the objective function value determines the system's reward for the current layout plan. , the reward is calculated by the current target value and the target value at the previous moment Through this dynamic feedback, the system can gradually optimize the layout strategy, so that each layout decision can improve the effectiveness of the entire solution. As the layout strategy is optimized, the system status will be continuously updated. First, the decision function of the precast beam It will be adjusted according to the output of the Actor network. Secondly, the carrying capacity distribution of the site and feasibility index It will also be recalculated with new layout decisions. This dynamic adjustment ensures that the system can respond to changes in site conditions in real time and update the layout strategy with new status information. Through this continuous status update and feedback, the system can ensure that the layout of precast beams is always in the optimal state.

[0109] The value evaluation of the Critic network is also one of the keys to the whole process. The Critic network helps the system identify which layout decisions are effective and which decisions may bring risks in the future. The update of the Critic network minimizes the current Values ​​and Goals The difference between the values ​​is used to achieve the goal The value not only considers the immediate return of the current layout, but also the long-term effect of the future layout. Through this mechanism, the system can balance the immediate benefits of the current layout with possible future returns, thereby ensuring the global optimality of the overall layout strategy. In order to deal with problems in the continuous space-time domain, the system uses radial basis functions to approximate the decision function of precast beams. The radial basis function is a tool that is very suitable for continuous space approximation, which can effectively capture the dynamic characteristics of precast beams in complex sites. Through function approximation, the system can make a more detailed description of complex site layout problems to ensure that precast beams can be placed in the optimal position. At the same time, the system also uses the target network technology in deep reinforcement learning to ensure stability during the learning process and prevent problems such as gradient explosion or disappearance. Finally, after multiple optimization iterations, the system will converge to an optimal layout plan. Under this plan, the optimal decision function for precast beams The optimal command matrix generated by the system through the Actor network is determined to represent the optimal placement position of the precast beam in each time step. , and pass it to the crane as the final execution plan. This process ensures that the layout plan of the precast beam yard is not only optimal under current conditions, but also can cope with changes in site conditions and construction needs in the future.

[0110] Embodiment 5: Step 4 specifically includes:

[0111] Step 4.1: Initialize the spatiotemporal noise process For motion exploration;

[0112] Step 4.2: Receive initial observation state ;

[0113] Step 4.3: For To the maximum time , perform the following process:

[0114] Select Action Execute actions based on current policy and exploration noise , observe the reward and the new state ; will convert Save to ;from Randomly sample a mini-batch of transformations ;set up ; is the subscript index; the parameters of the Critic network are updated by the following formula:

[0115] ;

[0116] Update the parameters of the Actor network using the following formula:

[0117] ;

[0118] The parameters of the target network are soft-updated using the following formula:

[0119] , ;

[0120] It is a soft update parameter.

[0121] Specifically, in step 4.1, the system initializes the spatiotemporal noise process , which is used to enhance the action exploration of the Actor network. The introduction of noise plays a vital role in reinforcement learning because it can help the system conduct extensive exploration in the early stages and avoid falling into local optimality. For the complex precast beam site layout problem, the dynamics and uncertainty of the site require the system to continuously explore different layout solutions and find the optimal strategy. By adding noise, the system's actions in each time step are No longer deterministic, but includes strategy and spatiotemporal noise This combination of actions can effectively expand the exploration space and ensure that the system can maintain sufficient flexibility when facing different site conditions and layout constraints. In step 4.2, the system receives the initial state This state usually includes the geometry, material, and cross-sectional characteristics of the precast beams, as well as the bearing capacity distribution and feasibility index of the current site. Through the integration of BIM and parametric databases, the system can obtain this information in real time, thus forming the basic state for layout optimization. After receiving the initial state, the system begins to enter a loop process, continuously executing actions, observing state changes, and updating strategies until the maximum time step is reached. The core of step 4.3 is that at each time step The layout decision process. In each time step, the system Select Action , the action is determined by the policy network Generate and add spatiotemporal noise , to increase the diversity of exploration. The goal of this step is to try different layout strategies, observe their effects under specific site conditions, and continuously improve the strategy based on the results. After that, the system observes the reward And update the status . Reward function It is calculated based on the results of the current layout decision, and usually reflects factors such as the rationality of the layout, site utilization efficiency, and safety. During the layout process of the precast beam yard, the reward can be measured based on the change in the objective function, such as whether the location of the precast beam meets the bearing capacity requirements and whether it improves the construction efficiency. This observation is used to evaluate the quality of the current layout decision and plays an important role in the subsequent strategy update process. In order to improve learning efficiency, the system uses the experience replay mechanism to record each state, action, reward, and next state. Store to experience replay buffer Experience replay is a commonly used reinforcement learning technique that can break the problem of data time correlation and improve the system's generalization ability by learning from past experience. Randomly sample a small batch of transformation pairs in , used to update the Critic network and the Actor network.

[0122] The update of the critic network is done by minimizing the predicted Values ​​and Goals The error between the values ​​is achieved. value Based on instant rewards The Critic network is calculated by minimizing the loss function , adjust its parameters , ensuring that the long-term effects of different layout strategies can be accurately evaluated. This update process ensures that the system can identify the most effective layout decisions and use them in subsequent policy adjustments. The update of the Actor network is achieved through policy gradients. The goal of the Actor network is to generate layout decisions that can maximize long-term returns. Using the action gradient information provided by the Critic network, the system calculates the gradient of the Actor network , and then adjust the parameters of the Actor network based on these gradients , so that the system can generate better layout decisions in the future state. The soft update mechanism is another key part of this embodiment. It ensures the stability of the learning process by gradually updating the parameters of the target network. In deep reinforcement learning, directly updating the target network may lead to instability in the learning process, and even gradient explosion or disappearance. Soft update introduces a small update factor , gradually bring the parameters of the Actor and Critic networks closer to the target network. Formula and This process is reflected to ensure that the target network parameters change smoothly after each update without causing large fluctuations. This mechanism can effectively improve the stability of learning and ensure that the system can converge to the optimal layout strategy after multiple iterations. Through these steps, the system continuously explores new layout strategies, evaluates their effects, and gradually adjusts and optimizes the parameters of the Actor and Critic networks. Finally, after multiple iterations, the system can converge to an optimal layout strategy. At this time, the optimal decision function of the precast beam is determined, indicating how the precast beams should be arranged to maximize the value of the objective function at a specific time step and position. This optimal strategy can not only improve the construction efficiency of the precast beam yard, but also effectively avoid problems such as insufficient site bearing capacity and safety hazards. The resulting crane command matrix As the final implementation plan, it guides the specific operations during the construction process.

[0123] Example 6: Prefabricated beams constructed by BIM data acquisition and processing unit The multi-scale feature tensor of is:

[0124] ;

[0125] in, is the geometric feature tensor, where For long, For width, for high; is the material characteristic tensor, where , and The positions are Elastic modulus, Poisson's ratio and density at ; is the cross-section characteristic tensor, where , and are the moment of inertia, torsion constant and section modulus distributions, respectively.

[0126] Specifically, the multi-scale feature tensor of the precast beam It consists of three parts: geometric feature tensor , material characteristic tensor and the cross-sectional characteristic tensor These tensors not only contain the basic geometric dimensions of the precast beam, but also the distribution of mechanical properties of its material at different locations and the mechanical properties of the cross section. First, the geometric characteristic tensor describes the geometric dimensions of the precast beam, where , and Represent the length, width and height of the precast beam respectively. These geometric parameters are crucial for the layout optimization of the precast beam yard, because they determine the space occupied by the precast beams in the yard and how to effectively arrange and place different precast beams. Through the accurate description of the geometric characteristics of the precast beams, the system can better evaluate the space occupied by each precast beam in the yard during the layout optimization process, as well as how to avoid interference between beams during construction. At the same time, geometric characteristics also affect the transportation, stacking and installation methods of precast beams, so accurate geometric descriptions provide important support for the overall planning and management of the precast beam yard. Secondly, the material characteristic tensor Describes the material properties of the precast beams, including the elastic modulus , Poisson's ratio and density , these parameters are the precast beams at different positions The material characteristic tensor enables the system to carefully characterize the mechanical properties of precast beams at different positions, especially the mechanical response under complex working conditions. Reflects the stiffness characteristics of precast beams at different positions, Poisson's ratio Describes the lateral deformation of the material under stress, while density This affects the weight and mass distribution of the precast beam. Since the material properties of the precast beam may vary at different locations, these material characteristics are crucial for accurately predicting the stress of the precast beam in the site and evaluating its impact on the site's bearing capacity. By modeling the material characteristics, the system can plan the layout of the precast beam more scientifically based on its mechanical properties to ensure that safety and stability requirements are met during construction. Finally, the section feature tensor It is used to describe the cross-sectional mechanical properties of precast beams. These cross-sectional parameters include the moment of inertia , torsion constant and section modulus , which affect the mechanical response of precast beams under bending, torsion and shear conditions respectively. Determines the bending resistance of the precast beam when subjected to bending, the torsion constant reflects the ability of the precast beam to resist torsion, while the section modulus It describes the resistance of the section under shear force. These cross-sectional features are particularly important for evaluating the mechanical performance of precast beams under different load conditions, especially in complex construction environments. The system needs to use these cross-sectional features to evaluate the stress safety and deformation of precast beams. By integrating the geometric feature tensor, material feature tensor, and cross-sectional feature tensor into a multi-scale feature tensor In the system, the precast beams can be fully described from multiple dimensions. This multi-scale feature representation allows the system to take into account the geometric dimensions, material properties, and cross-sectional mechanical properties of the precast beams during the layout optimization process, ensuring that the placement of each precast beam can take into account site conditions and construction requirements. When dealing with the problem of multi-physics field coupling on the site, this multi-scale feature tensor also provides important data support for the system, allowing the system to more accurately evaluate the interaction between the precast beams and the site.

[0127] Embodiment 7: The model building unit, based on the multi-scale tensor and site parameters, establishes a site multi-physics field coupling model based on geomechanics, hydrology and heat conduction through the following formula:

[0128] ;

[0129] ;

[0130] ;

[0131] in, is the gradient operator; is the total stress tensor; is the displacement vector; is the water head; is temperature; is the permeability coefficient; is the specific water storage rate; is the Biot coefficient; is the thermal conductivity coefficient; is the specific heat capacity; is the heat source term; for The second derivative of for The first derivative of For the second order Norm.

[0132] Specifically, the first equation of the model describes the stress state of the site based on geomechanics. The geometric characteristic tensor contained in the equation , material characteristic tensor and the cross-sectional characteristic tensor The model comprehensively describes the size, mechanical properties and cross-sectional characteristics of the precast beam. By combining these tensors into the site stress equation, the model can calculate the stress conditions of the precast beam at different locations. represents the divergence of stress, is the gravity term, which describes the stress distribution of the foundation under the action of gravity. This equation combines the physical characteristics of the precast beam and the site through the tensor norm , and The combination of describes the interaction between the precast beam and the site. The second derivative of displacement, i.e. the acceleration term, reflects the dynamic response of the precast beam under the bearing capacity of the site. Through this mechanical model, the system can accurately analyze the stress conditions that the precast beam may encounter during the construction process, and then adjust the layout to avoid problems such as foundation instability or precast beam deformation due to excessive stress concentration. The second equation describes the groundwater flow and infiltration phenomenon under hydrological conditions. This equation is based on Darcy's law and uses the permeability coefficient , specific water storage rate And Biot coefficient Describes the flow of groundwater in the site. represents the divergence of the seepage, and The hydrological condition is crucial in the layout of precast beams, because the infiltration of groundwater can significantly affect the bearing capacity and stability of the foundation, especially in the case of rainfall or groundwater level changes, the mechanical response of the site may change significantly. Introduced into this equation, the system can analyze the impact of different precast beams on groundwater infiltration, and then evaluate whether the layout plan is reasonable. In actual construction, this hydrological model helps to predict and prevent foundation settlement or instability caused by groundwater infiltration, thereby ensuring the safe layout of precast beams. The third equation describes the heat conduction phenomenon in the site, especially in high or low temperature environments, the impact of heat exchange between precast beams and the site on the construction process. Indicates the change of site material temperature over time, represents the divergence term of heat conduction, is the thermal conductivity coefficient. By introducing the multi-scale characteristic tensor of the precast beam , the system can evaluate the thermal response of precast beams under different temperature conditions to ensure that the layout can maintain stability under extreme climate conditions. This model is particularly suitable for scenarios where the impact of temperature changes on construction materials and sites needs to be considered, such as the expansion of materials at high temperatures or the contraction at low temperatures. Through the heat conduction model, the system can predict the deformation of precast beams under temperature changes to ensure that they will not crack or deform due to thermal stress during construction. In addition, the heat source term in the equation It reflects the impact of heat generated during construction on the site, ensuring that the system can evaluate the thermal impact under different construction schemes. The coupling of these three physical fields enables the system to accurately optimize the layout of precast beams under complex construction environments. Multi-scale characteristic tensor of precast beams The system provides a comprehensive description of the geometry, material and cross-sectional characteristics of each precast beam, while the geomechanical, hydrological and thermal conduction models evaluate the adaptability of the site from three aspects: mechanical response, groundwater infiltration and thermal conduction. This coupled analysis not only considers the independent response of the site under each physical field, but also analyzes the interaction between them through coupling terms. For example, groundwater infiltration may lead to a reduction in the bearing capacity of the foundation, while thermal conduction under high temperature conditions may further aggravate the settlement problem of the foundation. By comprehensively considering these factors, the system is able to find the optimal precast beam layout solution under complex multi-physical field conditions to ensure the safety and efficiency of the construction process.

[0133] Embodiment 8: A model building unit, based on a multi-physics field coupling model, calculates the Dynamic bearing capacity distribution of precast beams The formula is as follows:

[0134] ;

[0135] in, is the ultimate bearing capacity distribution; is the water head distribution; is the temperature distribution; is the terrain elevation function; , , and are critical stress, hydraulic head, temperature and slope respectively; is the absolute value operator; is the initial temperature.

[0136] Specifically, the dynamic bearing capacity distribution formula Indicates the ultimate bearing capacity distribution, which is the distribution of foundation materials at a specific location. The maximum load that a structure can withstand. This ultimate bearing capacity depends on the geological conditions of the site, such as soil properties, compaction level, etc. Ultimate bearing capacity It is the basis for the entire load-bearing capacity calculation and provides the initial value for the subsequent correction of each physical field. Next, the first term in the formula The ultimate bearing capacity and the section moment of inertia of the precast beam Perform ratio operations. Moment of inertia It reflects the ability of the precast beam section to resist bending and torsion. It is a key parameter in the mechanical properties of precast beams. By combining the ultimate bearing capacity with the moment of inertia, the system can adjust the bearing capacity calculation to ensure that the bending and torsion resistance of the precast beam structure itself is taken into account. This term can be understood as adjusting the strength of the precast beam under local force. Next, the term in the formula Indicates stress Stress is the core parameter in site mechanics. is the critical stress, that is, the maximum stress that the material can withstand. By introducing this exponential function, the system can dynamically evaluate the actual stress state of the prefabricated beam during the construction process. Especially when the stress is close to the critical value, the exponential function will greatly reduce the bearing capacity to prevent overload damage. Torsion constant is related to the stress correction term, which reflects the stiffness of the precast beam under torsion. By combining the stress and torsion constants, the system can effectively evaluate the bearing capacity when the torsion effect is significant. Next, The term is used to describe the impact of hydrological conditions on carrying capacity. is the head distribution, which indicates the change of groundwater level or seepage pressure, and is the critical water head value, which indicates the maximum water pressure that the foundation material can withstand. This term corrects the change in bearing capacity under high water head conditions in the form of an exponential function, ensuring that when groundwater penetration is strong, its weakening effect on the foundation bearing capacity can be accurately evaluated. Section modulus The shear capacity of the precast beam section is described, so the combination of water head and section modulus can effectively describe the influence of groundwater on the stability of the precast beam layout. The temperature effect is also described by the term Make a correction. Here, is the temperature distribution of the site, is the initial temperature, is the critical temperature difference. Temperature changes have a significant impact on the bearing capacity of precast beams, especially in high temperature or extremely cold environments, where the expansion or contraction of materials may lead to changes in mechanical properties. This term can quantify the weakening effect of temperature on bearing capacity when it deviates from the initial state through an exponential decay function, ensuring that the impact of ambient temperature can be considered during layout. Finally, The term introduces the effect of terrain slope. Terrain elevation function Gradient represents the surface slope, is the critical slope value. The slope affects the stability of the precast beam, especially in the case of steep slopes, the stability of the foundation will be greatly affected. Through this correction, the system can weaken the bearing capacity in areas with large slopes to avoid the risk of slippage or overturning due to unstable terrain. Through the correction of the above physical fields, the system can dynamically calculate the bearing capacity distribution of each precast beam at a specific time and location under complex site conditions. This calculation not only takes into account the influence of geomechanical stress, but also combines the influence of hydrological conditions, temperature changes and terrain slope to ensure the safety and stability of each precast beam layout. In the actual construction process, this bearing capacity distribution can provide direct guidance for layout optimization, so that the system can always maintain real-time monitoring and adjustment of the status of precast beams during construction, ensuring that the layout plan can cope with complex site changes and maximize construction efficiency and safety.

[0137] Embodiment 9: The model building unit, based on the multi-scale tensor and site parameters, calculates the first Dynamic feasibility index of precast beams :

[0138] ;

[0139] in, is the Heaviside step function; is the available area; For precast beams Stress distribution of is the first-order natural frequency of the coupled system; is the critical frequency difference; For the Precast beams in position Yield stress; For the The quality of the precast beams; For the The first order natural frequency of each precast beam; For the The spatial domain occupied by a precast beam.

[0140] Specifically, the first term of the dynamic feasibility index formula is The calculation is based on the relationship between the bearing capacity of the precast beam and its own gravity. Indicates at location and time Dynamic load distribution on It is a precast beam The weight of the Heaviside step function Used to determine whether the load-bearing capacity is sufficient to support the weight of the precast beam. is greater than the weight of the precast beam, then Returns 1, indicating that the bearing capacity at this location is sufficient and the precast beam can be safely placed; otherwise, returns 0, indicating that the bearing capacity is insufficient and the precast beam is not suitable for placement at this location. In this way, the system can fully evaluate the bearing capacity of the site before placing the precast beam to ensure that the layout plan will not cause the foundation to become unstable or the precast beam to overturn due to insufficient bearing capacity. Based on the available area of ​​the site Footprint with precast beams The geometry of the precast beam determines its footprint, while the available site area is determined by the actual conditions on site, such as the arrangement of other precast beams, the geometry of the site, and obstacles. Through this calculation, the system can ensure that the precast beams will not be restricted by the site when they are placed. If the available area of ​​the site is large enough to accommodate the precast beams, this item returns 1, indicating that the layout is feasible; otherwise, it returns 0, indicating that the layout is not feasible in the current location. This item is crucial to the overall layout planning of precast beams, especially when the site space is limited. The system can dynamically adjust the layout plan based on the actual available area. The third item The feasibility is evaluated based on the stress distribution. Represents precast beam The stress distribution of represents its yield stress. This term is calculated by comparing the precast beam with the space it occupies. Maximum stress in and minimum yield stress , to determine whether the stress of the precast beam exceeds the yield strength of the material. If the stress of the precast beam exceeds the yield strength, the material may be permanently deformed or damaged, so this term will return 0, indicating that the layout is not feasible; otherwise, it returns 1, indicating that the layout is feasible under stress conditions. This term is crucial to the safety assessment of precast beams to ensure that they can maintain structural integrity under complex loads. Finally, the exponential term in the formula The dynamic response of precast beams is introduced. is the first order natural frequency of the coupled system, and is the first-order natural frequency of a single precast beam. Natural frequency is an important parameter in structural dynamics, which reflects the vibration characteristics of the structure under external excitation. When the natural frequency of the coupled system is close to the natural frequency of the precast beam, resonance may occur, causing the vibration amplitude of the structure to increase, thereby affecting the stability of the precast beam. Therefore, through this exponential correction, the system can weaken the feasibility of the layout and avoid possible dynamic instability problems when the natural frequency of the precast beam is close to the natural frequency of the coupled system.

[0141] Embodiment 10: The objective function of the optimization problem defined by the layout optimization unit is expressed using the following formula:

[0142] ;

[0143] in, is the total number of precast beams.

[0144] Specifically, the system considers the layout of precast beams on site, not only considering the geometric characteristics, material characteristics and cross-sectional characteristics of each precast beam, but also combining specific site conditions such as bearing capacity, terrain changes, groundwater infiltration, temperature effects, etc. Through BIM technology, the system can extract these complex multi-scale features from the parametric database and optimize the calculation according to the actual construction environment. Dynamic Feasibility Index Used to evaluate the precast beams at specific locations and time The index combines the influence of multiple physical fields, such as stress distribution, groundwater level changes, temperature distribution, etc., to ensure that the layout of precast beams can meet the requirements of safety and stability under the condition of multi-physical field coupling. Secondly, the system introduces the bearing capacity distribution To evaluate the bearing capacity of each precast beam at different times and locations. This distribution reflects the maximum load that the site can withstand at the current moment. Since the bearing capacity of the foundation will change dynamically during construction due to changes in soil properties, construction progress, and hydrological conditions, the system must calculate the bearing capacity in real time and adjust the layout plan according to these changes. Insufficient bearing capacity may cause foundation instability or sinking of precast beams, thus affecting the overall construction progress and safety. Therefore, by dynamically evaluating the bearing capacity distribution, the system can effectively avoid possible risks and ensure that each precast beam is placed in a sufficiently safe position. The goal of the optimization problem is to select the optimal layout plan by comparing the dynamic feasibility index and bearing capacity distribution of each precast beam across the entire site and time. Specifically, the system solves the optimization problem so that at each moment and location , feasibility index Distribution of load capacity The difference between the two is maximized. When the feasibility index is much larger than the bearing capacity distribution, it indicates that the location is not only suitable for the placement of precast beams, but also can withstand the load required by the current construction. This optimization strategy ensures that the layout of precast beams in the site not only meets the physical constraints but also improves the construction efficiency. In order to achieve the global optimum, the system integrates the optimization problem in time and space dimensions. This means that the system not only considers the layout at the current moment, but also predicts future changes in site conditions and adjusts the current layout decisions based on these predictions. For example, if the groundwater level rises or the temperature changes in a certain period of time in the future may cause the bearing capacity to decrease, the system will make a response strategy in advance to avoid affecting the stability of the precast beams due to the deterioration of site conditions. Through the dual optimization of time and space, the system ensures that every decision in the layout process is fully considered, minimizing the risks and uncertainties in construction. In addition, the optimization process has a high degree of dynamic adjustment capabilities. During the construction process, the site conditions and the state of the physical field are constantly changing, and the system can respond quickly to these changes through real-time monitoring and data updates. Whether it is the fluctuation of site bearing capacity, or the mechanical response caused by groundwater infiltration or temperature change, the system can re-evaluate the layout plan based on the latest data to ensure that the construction plan can always remain in the optimal state as time goes by. This adaptive capability greatly improves the flexibility and reliability of the precast beam yard layout.

[0145] The present invention is described in detail above. Specific examples are used herein to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea. It should be pointed out that for ordinary technicians in this technical field, without departing from the principle of the present invention, several improvements and modifications can be made to the present invention, and these improvements and modifications also fall within the scope of protection of the claims of the present invention.

Claims

1. The precast beam yard layout optimization system based on BIM and parametric database integration is characterized by: The system includes: a BIM data acquisition and processing unit, which is used to extract the geometric feature tensor, material feature tensor and cross-sectional feature tensor of the precast beam from the BIM parameterized database, and respectively use the geometric feature tensor, material feature tensor and cross-sectional feature tensor as an element in a vector to construct a multi-scale feature tensor of the precast beam; a model construction unit, which is used to establish a site multi-physics field coupling model based on geomechanics, hydrology and heat conduction based on the multi-scale tensor and site parameters; calculate the dynamic bearing capacity distribution based on the multi-physics field coupling model; calculate the dynamic feasibility index for each precast beam and position combination based on the multi-scale tensor and site parameters; a layout optimization unit, which is used to convert the optimization problem into a continuous spatiotemporal Markov decision process, define the objective function of the optimization problem based on the dynamic bearing capacity distribution and dynamic feasibility index obtained by the model construction unit, solve the optimization problem using a deep deterministic policy gradient algorithm based on the Actor-Critic architecture, and find the instruction matrix of the crane corresponding to the optimal policy function as the final execution plan to complete the optimized layout of the precast beam yard.

2. The precast beam yard layout optimization system based on BIM and parametric database integration according to claim 1 is characterized in that: When the layout optimization unit transforms the optimization problem into a continuous spatiotemporal Markov decision process, the state space is defined as: ， The action space is: ； The reward function is: ； in, express System status at the moment; is the decision function, is the gradient of the decision function, precast beam The decision function is , indicating precast beam In time Place in position probability; for Time location Dynamic bearing capacity distribution at precast beams exist Time location The dynamic bearing capacity distribution at ; for Time location Dynamic feasibility index of precast beams exist Time location The dynamic feasibility index at ; for The command matrix of the crane at time , each element in the command matrix represents the An execution action parameter of the crane at the moment; for System actions at all times; To perform actions Back from state Rewards received; is the gradient of the objective function of the optimization problem.

3. The precast beam yard layout optimization system based on BIM and parametric database integration as claimed in claim 2 is characterized in that: When the layout optimization unit uses a deep deterministic policy gradient algorithm based on the Actor-Critic architecture to solve the optimization problem, the Actor network is represented as: ， in The parameters are Actor network with input as state , the output is a deterministic action; The Critic network is represented as: ， in The parameters are Critic network, the input is state and actions , the output is the state-action value function value, represents the real number space; The strategy update process is expressed as: ； in It is a behavioral strategy The resulting state distribution is is the policy gradient, Behavioral Strategy The resulting state distribution The expectations below, is the gradient of the Critic network to the action; is the gradient of the Actor network with respect to its parameters; The value function update process is expressed as: ， in, is the loss function of the Critic network, From the buffer Randomly draw multiple samples from Calculate the loss value of all sampled samples Take expectations, For the experience replay buffer, is the target Q value, For instant rewards, is the discount factor, and These are target network parameters; To handle continuous space-time domains, use function approximation: ； in is the selected radial basis function, For the Precast beam weights, For the Radial basis functions; is the number of radial basis functions.

4. The precast beam yard layout optimization system based on BIM and parametric database integration as claimed in claim 3 is characterized in that: The layout optimization unit uses a deep deterministic policy gradient algorithm based on the Actor-Critic architecture to solve the optimization problem. The process of obtaining the final layout solution includes: Step 1: Initialize the Actor Network and Critic Network ; Step 2: Initialize target network parameters , ; Step 3: Initialize the experience replay buffer ; Step 4: Learn the optimal precast beam layout strategy by repeatedly interacting with the simulation environment. In this process, select actions based on the current state and the learned strategy, execute these actions and observe the results, and then store the experience in the experience playback buffer; update the Critic network corresponding to its value estimation and the Actor network corresponding to the decision strategy by randomly sampling from these experiences; optimize the layout strategy and maximize the objective function; Step 5: Calculate the objective function value of the current layout solution ; Calculate rewards ; Update status Includes: Update Use Actor network output; recalculate ;and ; Step 6: Repeat steps 4 to 5 until the set number of times is reached to obtain the prefabricated beam. The optimal decision function is: ； After determining the decision function, the corresponding crane instruction matrix As the final implementation plan.

5. The precast beam yard layout optimization system based on BIM and parametric database integration as claimed in claim 4 is characterized in that: Step 4 specifically includes: Step 4.1: Initialize the spatiotemporal noise process For motion exploration; Step 4.2: Receive initial observation state ; Step 4.3: For To the maximum time , perform the following process: Select Action Execute actions based on current policy and exploration noise , observe the reward and the new state ; will convert Save to ;from Randomly sample a mini-batch of transformations ;set up ; is the subscript index; the parameters of the Critic network are updated by the following formula: ； Update the parameters of the Actor network using the following formula: ； The parameters of the target network are soft-updated using the following formula: ， ； It is a soft update parameter.

6. The precast beam yard layout optimization system based on BIM and parametric database integration as claimed in claim 5 is characterized in that: Prefabricated beams constructed by BIM data acquisition and processing unit The multi-scale feature tensor of is: ； in, is the geometric feature tensor, where For long, For width, for high; is the material characteristic tensor, where , and The location Elastic modulus, Poisson's ratio and density at ; is the cross-section characteristic tensor, where , and are the moment of inertia, torsion constant and section modulus distributions, respectively.

7. The precast beam yard layout optimization system based on BIM and parametric database integration according to claim 6 is characterized in that: The model building unit, based on multi-scale tensors and site parameters, establishes a site multi-physics field coupling model based on geomechanics, hydrology and heat conduction through the following formula: ； ； ； in, is the gradient operator; is the total stress tensor; is the displacement vector; is the water head; is temperature; is the permeability coefficient; is the specific water storage rate; is the Biot coefficient; is the thermal conductivity coefficient; is the specific heat capacity; is the heat source term; for The second derivative of for The first derivative of For the second order norm; is the foundation density; For tensors rank, is the norm operation.

8. The precast beam yard layout optimization system based on BIM and parametric database integration according to claim 7 is characterized in that: The model building unit calculates the Dynamic bearing capacity distribution of precast beams The formula is as follows: ； in, is the ultimate bearing capacity distribution; is the water head distribution; is the temperature distribution; is the terrain elevation function; , , and are critical stress, hydraulic head, temperature and slope respectively; is the absolute value operator; is the initial temperature.

9. The precast beam yard layout optimization system based on BIM and parametric database integration as claimed in claim 8, characterized in that: The model building unit, based on the multi-scale tensor and site parameters, calculates the first Dynamic feasibility index of precast beams : ； in, is the Heaviside step function; is the available area; For precast beams Stress distribution of is the first-order natural frequency of the coupled system; is the critical frequency difference; For the Precast beams in position Yield stress; For the The quality of the precast beams; For the The first order natural frequency of each precast beam; For the The spatial domain occupied by a precast beam.

10. The precast beam yard layout optimization system based on BIM and parametric database integration according to claim 9, characterized in that: The objective function of the optimization problem defined by the layout optimization unit is expressed using the following formula: ； in, is the total number of precast beams.

Images