A BIM-based intelligent optimization method and system for building construction pipelines

Abstract

This invention discloses a BIM-based intelligent optimization method and system for building construction pipelines, relating to the field of computer platform load balancing technology. The method includes collecting building data, pipeline layout information, and construction constraints from a BIM model; setting optimization objectives by combining building information with construction constraints; constructing a reinforcement learning environment based on the pipeline layout information in the BIM model; training an intelligent optimization agent using a deep reinforcement learning algorithm based on the constructed reinforcement learning environment; and adaptively adjusting the optimization objectives according to different needs at different construction stages. The pipeline layout scheme in the BIM model is updated based on the reinforcement learning optimization results, and the rationality of the scheme is evaluated. The method of this invention analyzes building data, pipeline layout information, and construction constraints in a BIM model.

Description

Technical Field

[0001] This invention relates to the fields of Building Information Modeling (BIM) and Artificial Intelligence Optimization Technology, specifically to a BIM-based intelligent optimization method and system for building construction pipelines. Background Technology

[0002] In the field of building construction, pipeline layout is a crucial factor affecting project quality and construction efficiency. Traditional building construction pipeline layout mainly relies on manual design and rule optimization. Construction workers typically adjust pipeline routes manually according to design drawings to meet construction specifications and functional requirements. However, this method suffers from several drawbacks: high difficulty in manual adjustments, low optimization efficiency, difficulty in automatically resolving spatial conflicts, difficulty in globally optimizing the layout scheme, complex construction constraints, difficulty in dynamically adapting to the needs of different construction stages, lack of intelligent optimization tools, and difficulty in automatically updating optimization results to the BIM model.

[0003] In recent years, Building Information Modeling (BIM) technology has been increasingly applied in construction engineering, with its 3D visualization, parametric modeling, and collision detection capabilities significantly improving the accuracy of pipeline layout. However, current BIM primarily relies on static modeling and cannot be dynamically optimized using artificial intelligence, resulting in a significant need for manual intervention during construction.

[0004] Meanwhile, reinforcement learning and deep reinforcement learning, as important branches of artificial intelligence, have made significant progress in areas such as path optimization, automatic scheduling, and intelligent decision-making. Combining deep reinforcement learning with BIM can enable pipeline optimization to have self-learning and adaptive adjustment capabilities, thereby greatly improving optimization efficiency and construction feasibility. Summary of the Invention

[0005] In view of the above-mentioned problems, the present invention is proposed.

[0006] Therefore, the technical problem solved by this invention is that existing methods for layout and optimization of building construction pipelines rely on manual adjustments, have low optimization efficiency, are difficult to adapt to the needs of different construction stages, and have difficulty automatically writing the optimization results back to the BIM model.

[0007] To address the aforementioned technical problems, this invention provides the following technical solution: a BIM-based intelligent optimization method for building construction pipelines, comprising collecting building data, pipeline layout information, and construction constraints from a BIM model, and setting optimization targets by combining building information with construction constraints.

[0008] A reinforcement learning environment is constructed based on the pipeline layout information in the BIM model. Based on the constructed reinforcement learning environment, a deep reinforcement learning algorithm is used to train an intelligent optimization agent, which adaptively adjusts the optimization target according to the different needs of the construction stage.

[0009] The pipeline layout scheme in the BIM model is updated based on the results of reinforcement learning optimization, and the rationality of the scheme is evaluated.

[0010] The training of intelligent optimization agents using deep reinforcement learning algorithms includes training the intelligent agent based on a deep Q-network and using a state-action-reward-policy optimization framework to dynamically optimize pipeline layout; the agent is optimized through path adjustment, rotation angle optimization, layer height adjustment, and replanning operations.

[0011] The adaptive adjustment optimization objectives include designing an adaptive weight adjustment strategy, dynamically adjusting the weights for reducing conflicts, optimizing paths, and lowering construction costs, and improving the learning ability of the optimization model through experience playback and strategy gradient updates.

[0012] As a preferred embodiment of the BIM-based intelligent optimization method for building construction pipelines described in this invention, the method for collecting building data, pipeline layout information, and construction constraints from the BIM model includes:

[0013] The IFC format is used to parse and obtain information on floor plans, walls, beams and columns, room distribution, equipment layout, and different types of pipeline information such as power, water supply and drainage, and HVAC from the BIM model. The building data and pipeline information are then used to construct a dataset.

[0014] Spatial constraints, pipeline crossing rules, and safety regulations are extracted as construction constraints, and the extracted construction constraints are used to construct a set of construction constraints.

[0015] The process involves parsing the IFC file using the IFC format and loading IfcProject as the overall data entry point for the building.

[0016] Extract the floor number, height, and floor area from IfcBuildingStorey and store them in the floor dataset.

[0017] Iterate through IfcWallStandardCase, IfcColumn, and IfcSpace to parse the location, thickness, and material of the walls, the size, height, and location of the columns, and the room name, location, and area data, and store them in the building dataset.

[0018] Parse IfcPipeSegment, IfcCableSegment, and IfcDuctSegment to identify the start, end, and path of the power pipeline, extract pipeline attributes, and store the duct path information into the pipeline dataset.

[0019] The set of construction constraints includes: setting minimum spacing constraints and construction slope constraints based on spatial limitations; setting intersection conflict constraints based on pipeline intersection rules; and setting pipeline support structure constraints based on safety regulations.

[0020] As a preferred embodiment of the BIM-based intelligent optimization method for building construction pipelines described in this invention, the step of setting optimization objectives by combining building information and construction requirements includes: constructing a multi-objective optimization function to formulate optimization objectives based on the set of construction constraints and the datasets of building data and pipeline information through weighted summation; and adjusting the weights adaptively based on the different priorities of the optimization objectives at different construction stages through an adaptive weight adjustment strategy.

[0021] As a preferred embodiment of the intelligent optimization method for building construction pipelines based on BIM described in this invention, the step of constructing a reinforcement learning environment based on pipeline layout information in the BIM model includes: constructing a reinforcement learning state space using pipeline layout information, building data, and construction constraints in the BIM model; representing the optimization state of the construction pipelines using the reinforcement learning state space; defining a state transition function; and performing optimization operations based on the current state using the state transition function to enter a new optimization state and find the optimal pipeline layout.

[0022] A preferred approach to constructing the state space for reinforcement learning is:

[0023] S t ={P t ,B,C,F,R}

[0024] Among them, S t P represents the state space for reinforcement learning. t A represents pipeline layout information, B represents building structure information, C represents construction constraints, F represents floor information, and R represents room information.

[0025] The state transition function is defined as follows: the optimized state of the construction pipeline is represented by a reinforcement learning state space, and the intelligent optimization agent takes optimization actions based on the state in the reinforcement space to enter a new optimized state.

[0026] A preferred approach to defining the state transition function is as follows:

[0027] S t+1 =f(S) t A t )

[0028] Where f represents the state transition function, A t Indicates based on S t The optimization actions taken, S t+1 It indicates the state at the next moment.

[0029] As a preferred embodiment of the BIM-based intelligent optimization method for building construction pipelines described in this invention, the step of training the intelligent optimization agent using a deep reinforcement learning algorithm includes initializing a Q-network based on the state space S. t Multiple pipeline layout schemes are sampled from BIM analysis data as training samples. One preferred scheme for the training samples is:

[0030] D={(S t A t ,R t ,S t+1 )}

[0031] Among them, S t Indicates the current pipeline layout status, A t R represents the action to optimize the agent's execution at the current moment. t R represents the reward value calculated based on the pipeline optimization objective. t Let represent the reward value at time step t, and D represent the training sample.

[0032] Using the ε-greedy strategy to select A t Using the ε-greedy strategy to select A t One preferred solution is:

[0033]

[0034] in, Denotes the optimal decision, Q(S) t A) represents the Q-value function, and A represents all possible actions.

[0035] Execute A t Enter S t+1 Update the Q-value, adjust the optimization strategy to maximize the cumulative reward, store the experience samples and repeat the steps until the Q-value converges, obtain the optimal pipeline optimization strategy, and terminate the training.

[0036] A preferred approach to updating the Q value is:

[0037]

[0038] Among them, Q(S) t A t ) represents the current state S t Select action A t The Q value, η represents the learning rate, and R t Indicates immediate reward, γ represents the discount factor. This represents the optimal reward in the future.

[0039] Termination of training includes the termination conditions for training the optimized agent, the cumulative reward convergence reaching the maximum number of training steps, and the optimized agent successfully finding the optimal pipeline layout.

[0040] One optimal solution for cumulative reward convergence is:

[0041]

[0042] in, Q represents the cumulative calculation from the first training round to the Tth training round. prev (S t A t Q(S) represents the Q-value from the previous training round. t A t ) represents the Q-value of the current training, and ∈ represents the convergence threshold.

[0043] Finding the optimal pipeline layout involves maximizing the cumulative reward obtained throughout the reinforcement learning process. One preferred solution for finding the optimal pipeline layout is:

[0044]

[0045] Among them, P * This represents the optimal pipeline layout scheme. This indicates that the objective function is to be solved. R represents the cumulative reward from t=1 to T. t This represents the reward value at time step t.

[0046] As a preferred embodiment of the BIM-based intelligent optimization method for building construction pipelines described in this invention, the adaptive adjustment of the optimization objective includes adopting a dynamic weighted strategy based on the construction stage, and adjusting the optimization objective of reinforcement learning through BIM construction data analysis, pipeline optimization constraint changes, and real-time feedback mechanisms.

[0047] The dynamic weighting strategy based on construction phases includes adopting an adaptive weight adjustment strategy, defining dynamic weights for the optimization objective, and a preferred scheme for defining the dynamic weights for the optimization objective is as follows:

[0048]

[0049] Among them, w k The adaptive weights, α, represent the optimization objective. k The rate at which the control weights are adjusted is represented, T0 represents the construction phase transition time, and t represents the construction time.

[0050] Based on the obtained adaptive weights, an adjustment function for the optimization objective is constructed. A preferred scheme for constructing the adjustment function is as follows:

[0051] Ftotal =w1F conflict +w2F length +w3F cost

[0052] Among them, F total Let F represent the overall objective function, w1, w2, and w3 represent the weight coefficients of the objective function, and F represent the weight coefficients of the objective function. conflict F represents the pipeline conflict optimization objective. length F represents the pipeline length optimization objective. cost This indicates the goal of optimizing construction costs.

[0053] As a preferred embodiment of the BIM-based intelligent optimization method for building construction pipelines described in this invention, the adaptive adjustment of the optimization objective further includes: evaluating the effectiveness of the optimization strategy in real time during the reinforcement learning process, calculating the adaptability score of the optimization scheme and feeding back the optimization objective adjustment, dynamically adjusting the optimization objective according to different construction stages, and optimizing the training process of the agent.

[0054] One preferred method for calculating the fitness score of the optimization scheme is:

[0055] S opt =w1S conflict +w2S length +w3S cost

[0056] Among them, S opt S represents the optimized score. conflict S represents the pipeline conflict resolution rate. length S represents the percentage reduction in pipeline path length after optimization. cost This represents the percentage reduction in construction costs after optimization. opt If the value is below the set threshold, the weight w is adjusted. k , and re-optimize the target allocation.

[0057] The optimization objectives are dynamically adjusted according to different construction stages. When it is detected that the optimization scheme does not meet the current construction needs, adjustments are made. If there are many pipeline conflicts, the weight of pipeline conflicts is increased. If there is a lot of waste of construction materials, the weight of construction costs is increased. If the optimization process does not meet the shortest path principle, the weight of pipeline path is increased.

[0058] Optimizing the agent training process includes, through S t Calculate S opt The target adjustment factor is calculated. If the target adjustment factor is greater than ∈, it means that the optimization target has not been achieved, and the optimization weights are adjusted.

[0059] A preferred method for calculating the target adjustment factor is as follows:

[0060] δ k =Sopt-prev -S opt

[0061] Where, δ k S represents the target adjustment factor. opt-prev This represents the score of the previous optimization objective.

[0062] One preferred approach to adjusting and optimizing weights is:

[0063] w k ←w k +η·δ k

[0064] Among them, w k The weights represent the current optimization objective, η represents the adjustment step size, and δ represents the weights. k This represents the adjustment factor for the optimization objective.

[0065] As a preferred embodiment of the intelligent optimization method for building construction pipelines based on BIM described in this invention, the method of updating the pipeline layout scheme in the BIM model includes: parsing the optimized pipeline layout data, updating the BIM data based on the IFC structure and performing conflict detection and verification, storing the updated BIM pipeline data and visualizing the optimization scheme, and exporting integrated construction data.

[0066] As a preferred embodiment of the intelligent optimization method for building construction pipelines based on BIM described in this invention, the rationality evaluation of the scheme includes: parsing the optimized BIM model, extracting the spatial location data of all pipelines, performing collision analysis using a BIM collision detection tool, checking pipelines against building structures, equipment, and other pipelines, and automatically adjusting pipeline paths when conflicts are found.

[0067] Another objective of this invention is to provide a BIM-based intelligent optimization system for building construction pipelines, which can dynamically optimize the layout of construction pipelines through an intelligent optimization agent based on deep reinforcement learning. This solves the problems of current BIM-based construction pipeline optimization technologies, such as reliance on manual adjustments, low optimization efficiency, and difficulty in adapting to the needs of different construction stages.

[0068] As a preferred embodiment of the BIM-based intelligent optimization system for building construction pipelines described in this invention, it includes a data acquisition and optimization target setting module, a reinforcement learning environment construction and training module, and an optimization result application and BIM model update module.

[0069] The data acquisition and optimization target setting module is used to collect building data, pipeline layout information and construction constraints in the BIM model, and set optimization targets by combining building information with construction constraints.

[0070] The reinforcement learning environment construction and training module is used to construct a reinforcement learning environment based on the pipeline layout information in the BIM model. Based on the constructed reinforcement learning environment, a deep reinforcement learning algorithm is used to train an intelligent optimization agent, which adaptively adjusts the optimization target according to the different needs of the construction stage.

[0071] The optimization results are applied to the BIM model update module to update the pipeline layout scheme in the BIM model based on the results of reinforcement learning optimization, and to evaluate the rationality of the scheme.

[0072] The beneficial effects of this invention are as follows: The intelligent optimization method for building construction pipelines based on BIM provided by this invention analyzes building data, pipeline layout information, and construction constraints in the BIM model to construct a reinforcement learning environment. This allows the intelligent optimization agent to be trained based on real construction data, avoiding the problems of traditional methods relying on fixed rules or manual adjustments, and improving the level of optimization intelligence. Using a deep Q-network for training enables the optimization agent to dynamically adjust its optimization strategy based on different needs at different construction stages, exhibiting higher adaptability and optimization capabilities compared to traditional heuristic algorithm-based optimization methods. By constructing a dynamic optimization weight adjustment strategy, the optimization agent can prioritize reducing conflicts in the early stages of construction, optimize pipeline paths in the middle stages, and reduce construction costs in the later stages, ensuring that the optimization scheme remains optimal at different construction stages. This invention achieves better results in terms of optimization intelligence, improved optimization efficiency, and enhanced construction adaptability. Attached Figure Description

[0073] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0074] Figure 1 The first embodiment of the present invention provides an overall flowchart of a BIM-based intelligent optimization method for building construction pipelines. Detailed Implementation

[0075] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0076] Example 1, referring to Figure 1As an embodiment of the present invention, a BIM-based intelligent optimization method for building construction pipelines is provided, comprising:

[0077] S1: Collect building data, pipeline layout information and construction constraints from the BIM model, and set optimization goals by combining building information with construction constraints.

[0078] In the process of optimizing building construction pipelines, the effectiveness of the optimization strategy depends on accurate building data, pipeline layout information, and construction constraints. To ensure the integrity and accuracy of the data, the Industry FoundationClasses format is used for parsing to extract key information such as floors, walls, pipelines, and equipment contained in the BIM model, construct the dataset required for optimization, and formulate optimization objectives that meet the project requirements in conjunction with construction constraints.

[0079] Furthermore, by parsing the IFC format, information on floor plans, walls, beams and columns, room distribution, equipment layout, and different types of pipeline information such as electricity, water supply and drainage, and HVAC is obtained from the BIM model. The building data and pipeline information are then used to construct a dataset.

[0080] Spatial constraints, pipeline crossing rules, and safety regulations are extracted as construction constraints, and the extracted construction constraints are used to construct a set of construction constraints.

[0081] The process involves parsing the IFC file using the IFC format and loading IfcProject as the overall data entry point for the building.

[0082] Extract the floor number, height, and floor area from IfcBuildingStorey and store them in the floor dataset.

[0083] Iterate through IfcWallStandardCase, IfcColumn, and IfcSpace to parse the location, thickness, and material of the walls, the size, height, and location of the columns, and the room name, location, and area data, and store them in the building dataset.

[0084] Parse IfcPipeSegment, IfcCableSegment, and IfcDuctSegment to identify the start, end, and path of the power pipeline, extract pipeline attributes, and store the duct path information into the pipeline dataset.

[0085] It should be noted that by collecting building data, pipeline layout information, and construction constraints from the BIM model, and by setting optimization objectives based on building information and construction constraints, the optimization algorithm is ensured to be based on real construction data, thereby improving the practical feasibility of the optimization results.

[0086] The optimization objectives can be dynamically adjusted, improving the adaptability of the optimization strategy. This reduces manual intervention, increases optimization efficiency, and deeply integrates BIM technology with artificial intelligence optimization technology, thereby enhancing the level of intelligent construction.

[0087] S2: Construct a reinforcement learning environment based on the pipeline layout information in the BIM model. Based on the constructed reinforcement learning environment, use deep reinforcement learning algorithms to train an intelligent optimization agent and adaptively adjust the optimization target according to different needs in the construction stage.

[0088] In the process of optimizing pipeline construction, the effectiveness of the optimization strategy depends not only on accurate data input, but also on the learning ability of the intelligent optimization agent, enabling it to dynamically adjust the optimization objectives at different construction stages. By constructing a reinforcement learning environment, training the optimization agent, and adjusting the optimization strategy based on construction needs, the optimization results are ensured to possess adaptability, construction feasibility, and global optimality.

[0089] Furthermore, a reinforcement learning state space is constructed using pipeline layout information, building data, and construction constraints in the BIM model. The state space of reinforcement learning represents the optimized state of the construction pipeline. A state transition function is defined, and optimization operations are performed based on the current state to enter a new optimized state, thereby finding the optimal pipeline layout.

[0090] A preferred approach to constructing the state space for reinforcement learning is:

[0091] S t ={P t ,B,C,F,R}

[0092] Among them, S t P represents the state space for reinforcement learning. t A represents pipeline layout information, B represents building structure information, C represents construction constraints, F represents floor information, and R represents room information.

[0093] The state transition function is defined as follows: the optimized state of the construction pipeline is represented by a reinforcement learning state space, and the intelligent optimization agent takes optimization actions based on the state in the reinforcement space to enter a new optimized state.

[0094] A preferred approach to defining the state transition function is as follows:

[0095] S t+1 =f(S) t A t )

[0096] Where f represents the state transition function, A t Indicates based on S t The optimization actions taken, S t+1 It indicates the state at the next moment.

[0097] Furthermore, the Q-network is initialized based on the state space S. t Multiple pipeline layout schemes are sampled from BIM analysis data as training samples. One preferred scheme for the training samples is:

[0098] D={(S t A t ,R t ,S t+1 )}

[0099] Among them, S t Indicates the current pipeline layout status, A t R represents the action to optimize the agent's execution at the current moment. t R represents the reward value calculated based on the pipeline optimization objective. t Let represent the reward value at time step t, and D represent the training sample.

[0100] Using the ε-greedy strategy to select A t Using the ε-greedy strategy to select A t One preferred solution is:

[0101]

[0102] in, Denotes the optimal decision, Q(S) t A) represents the Q-value function, and A represents all possible actions.

[0103] Execute A t Enter S t+1 Update the Q-value, adjust the optimization strategy to maximize the cumulative reward, store the experience samples and repeat the steps until the Q-value converges, obtain the optimal pipeline optimization strategy, and terminate the training.

[0104] A preferred approach to updating the Q value is:

[0105]

[0106] Among them, Q(S) t A t ) represents the current state S t Select action A t The Q value, η represents the learning rate, and R t Indicates immediate reward, γ represents the discount factor. This represents the optimal reward in the future.

[0107] Termination of training includes the termination conditions for training the optimized agent, the cumulative reward convergence reaching the maximum number of training steps, and the optimized agent successfully finding the optimal pipeline layout.

[0108] One optimal solution for cumulative reward convergence is:

[0109]

[0110] in, Q represents the cumulative calculation from the first training round to the Tth training round. prev (S t A t Q(S) represents the Q-value from the previous training round. t A t ) represents the Q-value of the current training, and ∈ represents the convergence threshold.

[0111] Finding the optimal pipeline layout involves maximizing the cumulative reward obtained throughout the reinforcement learning process. One preferred solution for finding the optimal pipeline layout is:

[0112]

[0113] Among them, P * This represents the optimal pipeline layout scheme. This indicates that the objective function is to be solved. R represents the cumulative reward from t=1 to T. t This represents the reward value at time step t.

[0114] Furthermore, a dynamic weighted strategy based on the construction phase is adopted, which adjusts the optimization objectives of reinforcement learning through BIM construction data analysis, pipeline optimization constraint changes, and real-time feedback mechanisms.

[0115] The dynamic weighting strategy based on construction phases includes adopting an adaptive weight adjustment strategy, defining dynamic weights for the optimization objective, and a preferred scheme for defining the dynamic weights for the optimization objective is as follows:

[0116]

[0117] Among them, w k The adaptive weights, α, represent the optimization objective. k The rate at which the control weights are adjusted is represented, T0 represents the construction phase transition time, and t represents the construction time.

[0118] In the early stages of construction, priority should be given to optimizing pipeline conflicts to ensure a reasonable pipeline layout. Conflict optimization has a greater weight than path optimization and cost optimization.

[0119] During the mid-construction phase, pipeline routes are optimized to reduce the total pipeline length and lower construction difficulty. The weight of route optimization is greater than that of conflict optimization and cost optimization.

[0120] In the later stages of construction, optimize construction costs to ensure optimal material and labor expenses. Cost optimization has a greater weight than conflict optimization and path optimization.

[0121] Based on the obtained adaptive weights, an adjustment function for the optimization objective is constructed. A preferred scheme for constructing the adjustment function is as follows:

[0122] F total =w1F conflict +w2F length +w3F cost

[0123] Among them, F total Let F represent the overall optimization objective function, w1, w2, and w3 represent the weight coefficients of the optimization objective, and F... conflict F represents the pipeline conflict optimization objective. length F represents the pipeline length optimization objective. cost This indicates the goal of optimizing construction costs.

[0124] It should be noted that the effectiveness of the optimization strategy is evaluated in real time during the reinforcement learning process, the adaptability score of the optimization scheme is calculated and the optimization target is adjusted accordingly, and the optimization target is dynamically adjusted according to different construction stages to optimize the training process of the agent.

[0125] One preferred method for calculating the fitness score of the optimization scheme is:

[0126] S opt =w1S conflict +w2S length +w3S cost

[0127] Among them, S opt S represents the optimized score. conflict S represents the pipeline conflict resolution rate. length S represents the percentage reduction in pipeline path length after optimization. cost This represents the percentage reduction in construction costs after optimization. opt If the value is below the set threshold, the weight w is adjusted. k , and re-optimize the target allocation.

[0128] The optimization objectives are dynamically adjusted according to different construction stages. When it is detected that the optimization scheme does not meet the current construction needs, adjustments are made. If there are many pipeline conflicts, the weight of pipeline conflicts is increased. If there is a lot of waste of construction materials, the weight of construction costs is increased. If the optimization process does not meet the shortest path principle, the weight of pipeline path is increased.

[0129] Optimizing the agent training process includes, through S t Calculate S opt The target adjustment factor is calculated. If the target adjustment factor is greater than ∈, it means that the optimization target has not been achieved, and the optimization weights are adjusted.

[0130] A preferred method for calculating the target adjustment factor is as follows:

[0131] δ k =S opt-prev -S opt

[0132] Where, δ k S represents the target adjustment factor. opt-prev This represents the score of the previous optimization objective.

[0133] One preferred approach to adjusting and optimizing weights is:

[0134] w k ←w k +η·δ k

[0135] Among them, w k The weights represent the current optimization objective, η represents the adjustment step size, and δ represents the weights. k This represents the adjustment factor for the optimization objective.

[0136] It should also be noted that by extracting data on walls, columns, and spatial distribution, the optimization agent can comprehensively consider spatial constraints to ensure that pipeline routes do not affect the building structure or functional areas. This reduces manual intervention, increases the degree of automation in optimization, and allows the optimization agent to autonomously learn optimization strategies, thereby improving construction efficiency.

[0137] Reinforcement learning models can iteratively optimize strategies under different construction environments, finding the optimal pipeline path through trial and error. The ε-greedy strategy is used for action selection, enabling the optimization agent to explore extensively in the initial stages and gradually converge to the global optimum, avoiding suboptimal results caused by early incorrect path selection. The training mechanism of reinforcement learning ensures that the optimization strategy can adapt to complex building environments, improving the generalization ability of the optimization scheme and making it applicable to various construction scenarios.

[0138] The optimization objectives can be dynamically adjusted to adapt to the needs of different construction stages. In the early stages of construction, priority is given to minimizing pipeline conflicts, ensuring reasonable wiring, and avoiding intersections that could affect construction progress. In the middle stages, pipeline routes are optimized to reduce total pipeline length and improve construction efficiency. In the later stages, the focus is on optimizing construction costs, reducing material and labor expenses, and improving construction economics. The dynamic weighting strategy automatically adjusts the weights of the optimization objectives based on the construction stage, ensuring that the optimization strategy always meets construction needs, reducing adjustment costs during construction, and improving construction efficiency.

[0139] The optimized agent can gradually find the optimal strategy, ensuring the stability and feasibility of the final solution. By maximizing cumulative rewards, the optimized agent can continuously adjust the optimization plan to achieve the optimal result. The application of the experience replay mechanism reduces the volatility during the training process, making the optimization results stable.

[0140] S3: Update the pipeline layout scheme in the BIM model based on the results of reinforcement learning optimization, and evaluate the rationality of the scheme.

[0141] The optimized pipeline layout data is analyzed, BIM data is updated based on the IFC structure, and conflict detection and verification are performed. The updated BIM pipeline data is stored, the optimized scheme is visualized, and integrated construction data is exported.

[0142] After analyzing and optimizing the BIM model, the spatial location data of all pipelines is extracted. A BIM collision detection tool is used to perform collision analysis to check the pipelines against the building structure, equipment, and other pipelines. When a conflict is found, the pipeline path is automatically adjusted.

[0143] Furthermore, the optimized pipeline layout data is parsed and matched with IFC semantic information. Based on the IFC structure, the BIM data is updated, modifying attributes such as pipeline location, direction, and diameter. Conflict detection and verification are performed to ensure the new layout complies with construction specifications. The updated BIM data is stored and version managed, the optimization scheme is visualized, the 3D view is updated, and construction simulation is supported.

[0144] Export construction data, generate CAD construction drawings and interface data with the management system, and ultimately realize intelligent optimization, real-time updates and automatic construction guidance of construction pipelines, thereby improving the level of intelligent application of BIM in construction management.

[0145] Example 2 is an embodiment of the present invention, which provides a BIM-based intelligent optimization method for building construction pipelines. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0146] First, in this embodiment, to verify the effectiveness of the intelligent optimization method for building construction pipelines based on BIM and reinforcement learning, construction pipeline data of a high-rise building was selected for testing. This building contains multiple floors and involves the layout of three types of pipelines: electricity, water supply and drainage, and HVAC. The test objective is to reduce pipeline conflicts, optimize pipeline paths, and lower construction costs through intelligent optimization agents.

[0147] First, based on BIM model data analysis, architectural information for each floor was extracted, including floor number, area, structural walls, beams and columns, and room distribution. Simultaneously, construction pipeline layout information was extracted, including parameters such as pipeline type, start point, end point, pipeline diameter, and path length. Combined with construction safety regulations, a set of construction constraints was established. This data was used to construct the state space for reinforcement learning, providing the basic input for the optimization process.

[0148] Subsequently, based on the construction of a reinforcement learning environment, a DQN model was used to train an intelligent optimization agent, enabling it to dynamically adjust optimization objectives at different construction stages. The optimization agent was trained using reinforcement learning based on BIM data, selecting the optimal pipeline path through an ε-greedy strategy and continuously adjusting the optimization strategy to reduce pipeline conflicts and optimize path layout. During training, the state of the pipeline path was continuously updated, and the optimization agent underwent trial and error in the construction environment, ultimately converging to the optimal pipeline layout scheme under the condition of maximizing cumulative reward.

[0149] After optimization, the BIM model was updated, including modifying pipeline layout information, adjusting pipeline routes, and optimizing pipeline spacing. The optimized BIM model was stored in the IFC structure and exported to CAD construction drawings. Simultaneously, BIM clash detection tools were used to perform conflict detection to ensure that the optimized pipeline scheme met construction specifications.

[0150] Table 1: Experimental Data

[0151]

[0152] As can be seen from the experimental data, the method of this invention has a significant optimization effect compared with traditional pipeline optimization methods. Before optimization, the pipeline layout generally had many conflicts, with an average of 18.5 conflicts per test case. After optimization, the number of pipeline conflicts was significantly reduced to 2.5, a reduction of approximately 86.5%. This indicates that the reinforcement learning optimization agent performs excellently in reducing pipeline conflicts and effectively improves the rationality of pipeline layout.

[0153] Furthermore, the average total length of the pipeline was 520 meters, which was reduced to 460 meters after optimization, a decrease of approximately 11.5%. This demonstrates that the optimization algorithm can effectively reduce redundant pipeline layout, optimize pipeline routes, and improve construction efficiency. The reduction in the total pipeline length after optimization also indirectly reduces construction costs, making the construction budget more controllable.

[0154] Regarding construction costs, the average construction cost before optimization was 861,700 yuan, while after optimization it decreased to 746,700 yuan, a reduction of approximately 115,000 yuan on average, saving 13.3% of construction costs. The reduction in construction costs mainly comes from the optimization of pipeline routes, which reduces the required pipeline materials and shortens the construction time, thereby reducing labor and material costs.

[0155] This experiment verifies that the intelligent optimization method for building construction pipelines, which combines BIM data analysis with reinforcement learning, can effectively reduce pipeline conflicts, optimize pipeline routes, lower construction costs, and improve construction efficiency.

[0156] This optimization method combines the visualization capabilities of BIM with the adaptive optimization capabilities of reinforcement learning, making construction pipeline optimization more intelligent and precise, while reducing manual intervention, improving construction quality, and providing construction teams with better construction management solutions.

[0157] Example 3, an embodiment of the present invention, provides a BIM-based intelligent optimization system for building construction pipelines, including a data acquisition and optimization target setting module 100, a reinforcement learning environment construction and training module 200, and an optimization result application and BIM model update module 300.

[0158] The data acquisition and optimization target setting module 100 is used to collect building data, pipeline layout information and construction constraints in the BIM model, and set optimization targets by combining building information with construction constraints.

[0159] The reinforcement learning environment construction and training module 200 is used to construct a reinforcement learning environment based on the pipeline layout information in the BIM model. Based on the constructed reinforcement learning environment, a deep reinforcement learning algorithm is used to train an intelligent optimization agent, which adaptively adjusts the optimization target according to the different needs of the construction stage.

[0160] The optimization results application and BIM model update module 300 is used to update the pipeline layout scheme in the BIM model based on the optimization results of reinforcement learning, and to evaluate the rationality of the scheme.

[0161] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0162] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0163] More specific examples (a non-exhaustive list) of computer-readable media include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which programs can be printed, because programs can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0164] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc. It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A BIM-based intelligent optimization method for building construction pipelines, characterized in that, include: Collect building data, pipeline layout information and construction constraints from the BIM model, and set optimization objectives by combining building information with construction constraints. A reinforcement learning environment is constructed based on the pipeline layout information in the BIM model. Based on the constructed reinforcement learning environment, a deep reinforcement learning algorithm is used to train an intelligent optimization agent. The optimization target is adaptively adjusted according to the different needs of the construction stage. The pipeline layout scheme in the BIM model is updated based on the results of reinforcement learning optimization, and the rationality of the scheme is evaluated. The training of intelligent optimization agents using deep reinforcement learning algorithms includes training the intelligent agent based on a deep Q-network and using a state-action-reward-policy optimization framework to dynamically optimize pipeline layout. The agent is optimized through path adjustment, rotation angle optimization, floor height adjustment, and replanning operations; The adaptive adjustment optimization objectives include designing an adaptive weight adjustment strategy, dynamically adjusting the weights for reducing conflicts, optimizing paths, and reducing construction costs, and improving the learning ability of the optimization model through experience playback and strategy gradient updates. The method of training an intelligent optimization agent using deep reinforcement learning algorithms includes... Initialize the Q network based on state space Multiple pipeline layout schemes are sampled from BIM analysis data as training samples. One preferred scheme for the training samples is: in, Indicates the current pipeline layout status. This indicates the action to be optimized by the agent at the current moment. This represents the reward value calculated based on the pipeline optimization objectives. Indicates the first The reward value at each time step. Indicates training samples; Using the ε-greedy strategy for selection Using an ε-greedy strategy for selection One preferred solution is: in, This represents the optimal decision. Represents the Q-value function, Indicates all possible actions; implement Enter Update the Q value, adjust the optimization strategy to maximize the cumulative reward, store the experience samples and repeat the steps until the Q value converges, obtain the optimal pipeline optimization strategy, and terminate the training. A preferred approach to updating the Q value is: in, Indicates the current state Select Action Q value, Indicates the learning rate. Indicates an immediate reward. Indicates the discount factor. Indicates the optimal reward in the future; Termination of training includes the termination conditions for training the optimized agent, the cumulative reward convergence reaching the maximum number of training steps, and the optimized agent successfully finding the optimal pipeline layout. One optimal solution for cumulative reward convergence is: in, This indicates the period from the first round of training to the second round. Cumulative calculation of training rounds, This represents the Q-value from the previous training round. This represents the Q-value during the current training. Indicates the convergence threshold; Finding the optimal pipeline layout involves maximizing the cumulative reward obtained throughout the reinforcement learning process. One preferred solution for finding the optimal pipeline layout is: in, This represents the optimal pipeline layout scheme. This indicates that the objective function is to be solved. Indicates from arrive The sum of accumulated rewards, Indicates the first The reward value for each time step.

2. The intelligent optimization method for building construction pipelines based on BIM as described in claim 1, characterized in that: The collection of building data, pipeline layout information, and construction constraints from the BIM model includes... The BIM model stores floor plans, walls, beams and columns, room distribution, equipment layout information, and different types of pipeline information such as power, water supply and drainage, and HVAC by parsing the IFC format, and then constructs a dataset by combining the building data and pipeline information. Spatial constraints, pipeline crossing rules, and safety regulations are extracted as construction constraints, and the extracted construction constraints are used to construct a set of construction constraints. The process involves parsing the IFC file using the IFC format and loading IfcProject as the overall data entry point for the building. Extract the floor number, height, and floor area from IfcBuildingStorey and store them in the floor dataset. Iterate through IfcWallStandardCase, IfcColumn, and IfcSpace to parse the location, thickness, and material of the walls, the size, height, and location of the columns, and the room name, location, and area data, and store them in the building dataset; Parse IfcPipeSegment, IfcCableSegment, and IfcDuctSegment to identify the start, end, and path of power lines, extract pipeline attributes, and store duct path information into the pipeline dataset; The set of construction constraints includes: setting minimum spacing constraints and construction slope constraints based on spatial limitations; setting intersection conflict constraints based on pipeline intersection rules; and setting pipeline support structure constraints based on safety regulations.

3. The intelligent optimization method for building construction pipelines based on BIM as described in claim 1 or 2, characterized in that: The optimization objectives set by combining building information with construction constraints include... Based on the set of construction constraints and the datasets of building data and pipeline information, a multi-objective optimization function is constructed by weighted summation to formulate optimization objectives. According to the different priorities of optimization objectives at different construction stages, an adaptive weight adjustment strategy is used to adjust the weights.

4. The intelligent optimization method for building construction pipelines based on BIM as described in claim 3, characterized in that: The construction of the reinforcement learning environment based on the pipeline layout information in the BIM model includes... A reinforcement learning state space is constructed using pipeline layout information, building data, and construction constraints in the BIM model. The optimized state of the construction pipeline is represented by the reinforcement learning state space. A state transition function is defined, and optimization operations are performed based on the current state to enter a new optimized state, thereby finding the optimal pipeline layout. A preferred approach to constructing the state space for reinforcement learning is: in, This represents the state space for reinforcement learning. This indicates pipeline layout information. Indicates building structure information, Indicates construction constraints. Indicates floor information. Indicates room information; The defined state transition function includes: the optimized state of the construction pipeline is represented by a reinforcement learning state space; and the intelligent optimization agent takes optimization actions based on the reinforcement space state to enter a new optimized state. A preferred approach to defining the state transition function is as follows: in, Represents the state transition function. Indicates based on Optimization actions taken, It indicates the state at the next moment.

5. The intelligent optimization method for building construction pipelines based on BIM as described in claim 1, 2, or 4, characterized in that: The adaptive adjustment and optimization objectives include: A dynamic weighted strategy based on the construction phase is adopted, and the optimization target of reinforcement learning is adjusted through BIM construction data analysis, pipeline optimization constraint changes, and real-time feedback mechanism. The dynamic weighting strategy based on construction phases includes adopting an adaptive weight adjustment strategy, defining dynamic weights for the optimization objective, and a preferred scheme for defining the dynamic weights for the optimization objective is as follows: in, The adaptive weights represent the optimization objective. This indicates the rate at which the control weights are adjusted. Indicates the time point of transition between construction phases. Indicates the construction period; Based on the obtained adaptive weights, an adjustment function for the optimization objective is constructed. A preferred scheme for constructing the adjustment function is as follows: in, Denotes the overall optimization objective function. , , The weight coefficients represent the optimization objective. This indicates the pipeline conflict optimization objective. This indicates the pipeline length optimization objective. This indicates the goal of optimizing construction costs.

6. The intelligent optimization method for building construction pipelines based on BIM as described in claim 5, characterized in that: The adaptive adjustment optimization objective also includes... The effectiveness of the optimization strategy is evaluated in real time during the reinforcement learning process. The adaptability score of the optimization scheme is calculated and the optimization target is adjusted accordingly. The optimization target is dynamically adjusted according to different construction stages to optimize the training process of the agent. One preferred method for calculating the fitness score of the optimization scheme is: in, Indicates the optimized score. This indicates the pipeline conflict resolution rate. This indicates the percentage reduction in pipeline path length after optimization. This represents the percentage reduction in construction costs after optimization. If the value falls below a set threshold, the weight will be adjusted. Re-optimize target allocation; The optimization objectives are dynamically adjusted according to different construction stages. When the optimization scheme is detected to not meet the current construction needs, adjustments are made. If there are many pipeline conflicts, the weight of pipeline conflicts is increased. If there is a lot of waste of construction materials, the weight of construction cost is increased. If the optimization process does not meet the shortest path principle, the weight of pipeline path is increased. Optimizing the agent training process includes, through calculate And calculate the target adjustment factor. When the target adjustment factor is greater than This indicates that the optimization objective has not been achieved, and the optimization weights should be adjusted. A preferred method for calculating the target adjustment factor is as follows: in, Indicates the target adjustment factor. This represents the score of the previous optimization objective; One preferred approach to adjusting and optimizing weights is: in, This indicates the weight of the current optimization objective. Indicates adjusting the step size. This represents the adjustment factor for the optimization objective.

7. The intelligent optimization method for building construction pipelines based on BIM as described in claim 6, characterized in that: The updated pipeline layout scheme in the BIM model includes... The optimized pipeline layout data is analyzed, BIM data is updated based on the IFC structure, and conflict detection and verification are performed. The updated BIM pipeline data is stored, the optimized scheme is visualized, and integrated construction data is exported.

8. The intelligent optimization method for building construction pipelines based on BIM as described in claims 1, 2, 4 or 7, characterized in that: The assessment of the rationality of the proposed solution includes, After analyzing and optimizing the BIM model, the spatial location data of all pipelines is extracted. A BIM collision detection tool is used to perform collision analysis to check the pipelines against the building structure, equipment, and other pipelines. When a conflict is found, the pipeline path is automatically adjusted.

9. A system employing the BIM-based intelligent optimization method for building construction pipelines as described in any one of claims 1 to 8, characterized in that: It includes a data acquisition and optimization target setting module (100), a reinforcement learning environment construction and training module (200), and an optimization result application and BIM model update module (300). The data acquisition and optimization target setting module (100) is used to collect building data, pipeline layout information and construction constraints in the BIM model, and set optimization targets by combining building information with construction constraints; The reinforcement learning environment construction and training module (200) is used to construct a reinforcement learning environment based on the pipeline layout information in the BIM model. Based on the constructed reinforcement learning environment, a deep reinforcement learning algorithm is used to train an intelligent optimization agent, and the optimization target is adaptively adjusted according to the different needs of the construction stage. The optimization results are applied to the BIM model update module (300) to update the pipeline layout scheme in the BIM model based on the results of reinforcement learning optimization, and to evaluate the rationality of the scheme.

Images