A safety control method for building component dismantling sequence integrating genetic algorithm, Bayesian reasoning and machine learning

Abstract

This invention provides a method for safely controlling the demolition sequence of building components that integrates genetic algorithms, Bayesian reasoning, and machine learning. This method uses topological sorting to establish a directed acyclic graph of component dependencies. Mechanical simulation and genetic algorithm optimization are combined to generate feasible, low-risk dismantling sequence paths. Bayesian reasoning is used to dynamically perceive structural risks, supplemented by machine learning models to enhance the stability of risk identification, achieving a coordinated linkage between dismantling sequence path generation and risk control. The invention's data interaction and decision-making linkage form a closed-loop control chain driven by risk assessment to regulate the dismantling sequence path. By integrating genetic algorithms, Bayesian reasoning, and machine learning, this method achieves dynamic optimization and full-process safety assurance for building structure demolition.

Description

Technical Field

[0001] The present invention relates to the field of building technology, and in particular to a building component dismantling sequence safety control method integrating genetic algorithm, Bayesian reasoning and machine learning. Background Art

[0002] With the increasing demand for urban construction and renewal, building demolition has become an essential and unavoidable part of modern engineering. Traditional building dismantling methods rely heavily on experience and manual judgment, making safety issues particularly prominent when dealing with complex high-rise structures. During the dismantling process, due to the complex dependencies between structural components, incorrect dismantling sequences can lead to structural instability or even collapse, posing a significant safety hazard. Therefore, effectively controlling risks and optimizing the dismantling sequence during the demolition process to ensure construction safety are pressing challenges in current building dismantling technology.

[0003] However, existing building disassembly methods typically prescribe a fixed disassembly sequence and lack real-time monitoring and adjustment of the building structure's status, making it difficult to cope with stress changes that occur during the disassembly process. Furthermore, although real-time monitoring technology is relatively mature, effectively utilizing real-time data (such as stress, displacement, and vibration) to assess the risk of structural instability remains a blind spot in existing technologies. Furthermore, the aging of building structures, changes in material properties, and noise in real-time monitoring data introduce substantial uncertainties that are difficult for existing methods to address. Summary of the Invention

[0004] (1) Technical issues to be resolved

[0005] Based on this, the present invention provides a method for safely controlling the disassembly sequence of building components by integrating genetic algorithms, Bayesian reasoning, and machine learning to solve the problems mentioned in the background technology, such as the lack of monitoring and adjustment of the real-time status of the structure due to the fixed disassembly sequence, insufficient real-time data analysis capabilities, and inability to handle uncertainty factors.

[0006] (2) Technical solution

[0007] To achieve the above objectives, the present invention provides a method for safely controlling the demolition sequence of building components by integrating genetic algorithms, Bayesian reasoning, and machine learning, comprising:

[0008] S1: Use topological sorting to analyze the dependencies between components in a building structure and construct a directed acyclic graph of the disassembly order;

[0009] S2: In the directed acyclic graph, randomly select The disassembly sequence schemes that meet the topological constraints are used as the initial population of the genetic algorithm, where ; Based on mechanical simulation, the safety of the state of the disassembly sequence scheme is evaluated, and combined with the optimization operation of the genetic algorithm, the actual disassembly scheme is obtained;

[0010] S3: During the dismantling process actually adopted, the monitoring data of the building structure is combined with Bayesian reasoning and machine learning to conduct dynamic risk prediction and continuous correction of the building structure status.

[0011] Furthermore, the S2 includes:

[0012] S201: In a directed acyclic graph, randomly select The disassembly sequence schemes that meet the topological constraints are taken as candidate schemes, and the candidate schemes are taken as the initial population of the genetic algorithm;

[0013] Disassembly sequence scheme using component sequence vector in directed acyclic graph express:

[0014]

[0015] in, Indicates the disassembly sequence plan; Indicates the Step 1: The number of the component to be removed; represents the number of disassembly steps in the disassembly sequence scheme, where ;

[0016] S202: Performing a safety assessment on the status of the candidate solution based on mechanical simulation and calculating the fitness of the candidate solution;

[0017] In the disassembly sequence scheme, for each disassembly state , use the finite element analysis method to simulate the stress distribution of the building structure and output the components to be disassembled in the next step The actual stress value , and calculate Safety factor :

[0018]

[0019] in, Indicates that the next step is to disassemble the component The yield strength of the material; Indicates that the next step is to disassemble the component The actual stress of

[0020] Then the fitness function of the entire disassembly sequence scheme is The definition is as follows:

[0021]

[0022] in, Indicates the disassembly sequence plan; Indicates the number of steps in the disassembly sequence plan; Indicates the current status The corresponding safety factor;

[0023] S203: Determine whether the fitness is within a safe range. If so, use the candidate solution with the highest fitness as the actual disassembly sequence solution. If not, proceed to the next step.

[0024] S204: Screening for the highest fitness The candidate solutions are optimized by genetic algorithm, where and ;

[0025] S205: Using the result of the optimization operation as a new candidate solution, verifying whether the new candidate solution satisfies the topology constraint. If so, proceed to S202; if not, proceed to S204.

[0026] Furthermore, the S3 includes:

[0027] S301: During the disassembly process actually adopted, obtain the current status Monitoring data of building structures ;

[0028] S302: Based on the monitoring data ; Calculate the corresponding Bayesian posterior probability As a Bayesian risk score ,in Indicates that a structural instability event occurs;

[0029] S303: The monitoring data Input the trained machine learning model and obtain the risk score value output by the machine learning model ;

[0030] S304: Comprehensive Bayesian risk score and the risk score value output by the machine learning model , for the current state Conduct dynamic risk prediction and continuous correction.

[0031] Furthermore, the state in S301 Monitoring data The definition of is as follows:

[0032]

[0033] in, It represents the maximum displacement of all remaining components in the demolition sequence, reflecting the most serious deformation of the building structure in the current state; It represents the maximum value of the ratio of the actual strain of all remaining components in the demolition sequence to the yield strain of their materials, and is used to reflect the potential damage risk of the most dangerous component; Indicates the peak acceleration at the key parts of the building structure, reflecting the degree of inertial impact on the building structure. The key parts are selected by experts through analysis and judgment; It is the maximum horizontal inclination angle of the current building structure as a whole or a typical floor, used to reflect the overturning tendency of the structure; It indicates the maximum ratio of residual stress to material ultimate strength at key parts of the current building structure, and is used to measure the residual bearing capacity of the building structure.

[0034] Furthermore, in S302, based on the monitoring data , use the Bayesian inference formula to evaluate the current state The posterior probability , the formula is as follows

[0035]

[0036] in, Indicates that a structural instability event occurs; Indicates status monitoring data; Represents the static prior probability, which is determined by the hierarchical weight and safety factor of the component to be disassembled in the next step in the directed acyclic graph The smaller the topological depth of the component to be disassembled next, The smaller the value, the higher the prior probability is assigned; Represents the likelihood probability, which is obtained by the statistics of a large number of simulation samples accumulated during the genetic algorithm optimization process in step S2. In order to avoid the monitoring value in the multidimensional feature space It is difficult to completely overlap with the simulation sample, so it can be expanded appropriately The tolerance range, that is, will be in the multidimensional feature space with The Euclidean distance is less than The simulation sample is used as ,in is a preset positive number; represents the probability of monitoring, which can be generated through statistical analysis; Represents the posterior probability, which serves as the basis for risk assessment of the current state;

[0037] Receive new monitoring data After that, the static prior probability is no longer used to calculate the posterior probability of the new state. Instead, the posterior probability of the previous state is used as the prior probability of the new state, as shown below:

[0038] .

[0039] Furthermore, the training of the machine learning model in S303 includes:

[0040] In each state , collect monitoring data characteristics of building structures , and record the posterior probability obtained by Bayesian inference , build a sample The training set composed of

[0041] First, for each sample Calculate its risk score The scoring function comprehensively considers the topological depth and Bayesian posterior probability of the component to be disassembled in the next step. The formula is as follows:

[0042]

[0043] in, Indicates the topological depth of the component to be disassembled in the next step; is a minimum non-zero constant, with a value of 0.001; Indicates recent The sliding average of the posterior probability is used to reflect the difference between the current state and the previous state. The average risk of the state after the disassembly step; if the current number of disassembly steps is insufficient ,but The value is the actual number of steps. is the weighting coefficient used to express the "structural dependency level" and “Instability Probability Assessment” Assign weights between the two types of information and satisfy , Set by empirical data or training verification results;

[0044] The historical monitoring data is used as input and clustered in high-dimensional state space using K-means clustering method to obtain Clusters , and the center of each cluster; for each sample ,based on Similarity and risk score with the corresponding cluster center , calculate the sample Representative scores , the representativeness score is used to determine the priority of samples in the training set; the formula for the representativeness score is as follows:

[0045]

[0046] in, Represents a sample Monitoring data characteristics in and the cluster center of the cluster to which it belongs The Euclidean distance of is the weight coefficient and ;

[0047] In each cluster, keep the score The highest number of samples are used as representative samples to form a training subset; the training subset is input into the machine learning model for training.

[0048] Furthermore, in S304, during the dismantling process, when the machine learning model has been fully trained, the "joint discrimination mechanism" is activated, and the rules are as follows:

[0049] like , then the current risk judgment is recognized, that is, the posterior probability obtained by Bayesian reasoning is considered correct;

[0050] like , then the current risk judgment is not recognized, that is, it is believed that there is a problem with the posterior probability obtained by Bayesian reasoning, and the demolition operation is suspended. After manual review, a second judgment is made;

[0051] in, is the preset threshold.

[0052] Furthermore, the S1 includes:

[0053] S101: Acquire and analyze relevant data of the building structure, determine the interdependence of each component during the demolition process and establish a dependency matrix;

[0054] S102: Based on the dependency matrix, topologically sort the components and draw a directed acyclic graph.

[0055] Furthermore, the optimization operation of the genetic algorithm in S204 includes crossover and mutation.

[0056] (3) Beneficial effects

[0057] From the above technical solution, it can be seen that the method for safely controlling the demolition sequence of building components proposed by the present invention, which integrates genetic algorithm, Bayesian reasoning and machine learning, has the following beneficial effects:

[0058] 1. Topological sorting is used to screen the disassembly sequence path, and a genetic algorithm is used to optimize the short-step disassembly plan. Combined with mechanical simulation, the stability of each disassembly state is evaluated to ensure that each disassembly operation is within a controllable range, significantly reducing the risk of structural instability caused by improper disassembly sequence and improving the structural safety of the disassembly process.

[0059] 2. Real-time monitoring status data is collected and a Bayesian inference model is used to recursively update the instability probability of the current state. Simultaneously, a machine learning model quickly provides auxiliary judgments, achieving dual verification. Once the risk exceeds the threshold, operations are automatically suspended and an optimization plan is regenerated, ensuring closed-loop risk control and enabling dynamic risk identification and immediate response.

[0060] 3. The present invention can continuously receive field data and dynamically learn historical judgment results, so that the system has the ability of self-learning and model evolution. It can effectively deal with structural response changes caused by unpredictable factors such as material aging, environmental disturbances, and construction errors, improve the robustness and reliability of the overall system, and enhance the adaptability of the disassembly process to uncertain factors. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] The features and advantages of the present invention will be more clearly understood by referring to the accompanying drawings, which are schematic and should not be construed as limiting the present invention in any way. In the accompanying drawings:

[0062] Figure 1 This is a flow chart of the building disassembly sequence safety control method that integrates genetic algorithm, Bayesian reasoning and machine learning;

[0063] Figure 2 A flow chart of the present invention combining mechanical simulation and genetic algorithm to form a disassembly plan with minimal risk;

[0064] Figure 3 Schematic diagram of the joint risk identification using Bayesian reasoning and machine learning in the present invention. DETAILED DESCRIPTION

[0065] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0066] like Figure 1 As shown, the present invention provides a building disassembly sequence safety control method that integrates topological sorting, optimization algorithm, Bayesian reasoning and machine learning, including:

[0067] S1: Use topological sorting to analyze the dependencies between components in the building structure and construct a directed acyclic graph of the disassembly order. The specific steps include:

[0068] S101: Acquire and analyze relevant data of the building structure, determine the interdependence of each component during the demolition process and establish a dependency matrix;

[0069] To ensure the safety of the demolition process, relevant building data must first be obtained. This data includes, but is not limited to, design drawings (i.e., the building's design drawings, which include the layout of each floor's load-bearing structure and non-load-bearing walls) and detailed structural information (including material types, component connections, component size, and strength). This data allows for a comprehensive analysis of the building's structural characteristics and identifies the interdependencies between components during demolition.

[0070] The dependencies between components are determined based on the following principles:

[0071] 1. Principle of dependence of load-bearing structure: There is a dependence relationship between the load-bearing structural components of a building. Beams depend on columns, floor slabs depend on beams, secondary beams depend on main beams, shear walls depend on peripheral frames, and structural connecting beams depend on walls. The dependent components must be demolished later.

[0072] 2. The principle of prioritizing demolition of the superstructure; in multi-story buildings, the load-bearing components of the upper floors must be demolished first to avoid affecting the stability of the lower floors.

[0073] 3. Principle of last demolition of lateral force resisting components; lateral force resisting components play a key role in providing the overall lateral stability of the building. When demolishing, lateral force resisting components should be demolished as late as possible after other components in the same situation, otherwise it may cause local instability or even overall collapse.

[0074] 4. Principle of prioritizing demolition of fragile and high-risk areas: When multiple components are closely connected, demolition should start from the most fragile or dangerous areas to avoid accidental collapse or instability during the demolition process.

[0075] Combine the interdependencies of each component to define a dependency matrix .in, =1 indicates component After dismantling, the components Traverse all building components and build a dependency matrix.

[0076] S102: Based on the dependency matrix, topologically sort the components and draw a directed acyclic graph.

[0077] Through the dependency matrix Components are topologically sorted to generate a removal order that complies with dependencies. The resulting directed acyclic graph (DAG) shows the priorities and dependencies between components, providing a clear reference path for subsequent removal operations.

[0078] After completing the construction of the directed acyclic graph, the concept of topological depth is further introduced to measure the hierarchical influence of each component in the overall structural dominance relationship. Topological depth reflects how many upstream components a component is controlled by and the degree to which it is relatively "backward" or "dominated" in the entire disassembly path. For each component node in the directed acyclic graph, , defining its topological depth From any source node to The longest path length, that is, starting from any source node in the directed acyclic graph to The number of edges contained in the longest path. The source node indicates that there is no predecessor node and can be directly removed without considering the removal order of other parts.

[0079] The present invention utilizes the topological depth of a component as one of the scores for measuring its importance in a directed acyclic graph, thereby participating in the subsequent estimation modeling of a priori instability probability.

[0080] S2: In the directed acyclic graph, randomly select A disassembly sequence scheme that satisfies the topological constraints is used as the initial population of the genetic algorithm; the safety of the state of the disassembly sequence scheme is evaluated based on mechanical simulation, and combined with the optimization operation of the genetic algorithm, the actual disassembly scheme is obtained;

[0081] In a directed acyclic graph, one or more disassembly sequence plans (i.e., candidate plans) that satisfy topological constraints (component priorities and dependencies) are randomly selected as the initial population for the genetic algorithm. Each candidate plan consists of several disassembly steps. Mechanical simulation methods are then used to evaluate the structural state and instability risk of each candidate plan. Ultimately, the optimal disassembly sequence is selected using a fitness function. Since the goal of the building disassembly sequence problem is to find a set of component disassembly sequences that satisfies topological dependencies and optimizes overall safety, this problem is a constrained permutation optimization problem.

[0082] like Figure 2 As shown, S2 specifically includes:

[0083] S201: In a directed acyclic graph, randomly select The disassembly sequence schemes that meet the topological constraints (comply with the priority and dependency relationship between components) are selected as candidate schemes, and the candidate schemes are used as the initial population of the genetic algorithm;

[0084] The disassembly order of building components is modeled as a permutation optimization problem, and the genes of the genetic algorithm are the component sequence vectors in the directed acyclic graph. :

[0085]

[0086] in, Indicates the The number of the component to be dismantled. The length is , represents a sequence of dismantling components (i.e., dismantling sequence scheme). Different sequences will affect the stress distribution and instability risk of the building structure. Indicates the number of disassembly steps in the disassembly sequence plan.

[0087] The encoding method of the genetic algorithm uses the permutation vector of component numbers to represent the dismantling order, ensuring that each dismantling order scheme corresponds to a complete and non-repetitive dismantling sequence.

[0088] S202: Performing a safety assessment on the status of the candidate solution based on mechanical simulation and calculating the fitness of the candidate solution;

[0089] In each disassembly scheme, for each disassembly state , using the finite element analysis (FEM) method to simulate the stress distribution of the building structure. The mechanical simulation takes into account the influence of external loads and boundary conditions on the structural stress and outputs the components to be disassembled in the next step. The actual stress value , and calculate its safety factor. Safety factor The formula is as follows:

[0090] (1)

[0091] in, Indicates that the next step is to disassemble the component The yield strength of the material; Indicates that the next step is to disassemble the component actual stress.

[0092] During the assessment of each disassembly step, the safety factor is used to determine the instability risk of the next component to be disassembled. If the safety factor of a state is close to 1, the building structure is close to instability; if the safety factor of a state is much greater than 1, the building structure is stable.

[0093] Fitness function Used to evaluate the overall security of each disassembly sequence scheme. In each scheme, the current state (That is, the components to be disassembled in the next step are ) corresponds to a safety factor of , and the security of each state is interrelated. Therefore, the fitness function should reflect the coherence and cumulative risk of multiple states. The fitness function is defined as follows:

[0094] (2)

[0095] in, Indicates the disassembly sequence plan; Indicates the number of steps in the disassembly sequence plan; Indicates the current status The corresponding safety factor; The solution is shown in formula (1).

[0096] Through this step-by-step multiplication, the fitness function can reflect the coherent risk of each state in the disassembly plan. If the safety factor of a certain step is low, the fitness value will drop significantly, thus guiding people to avoid high-risk plans.

[0097] S203: Determine whether the fitness is within a safe range. If so, use the candidate solution with the highest fitness as the actual disassembly sequence solution (i.e., the disassembly solution with the lowest risk). If not, proceed to the next step.

[0098] The safety range is set artificially based on experience.

[0099] S204: Screening for the highest fitness The candidate solutions are optimized by genetic algorithm;

[0100] The optimization operations include crossover and mutation operations. The crossover adopts the PMX (partial matching crossover) algorithm, and the mutation is to randomly exchange, insert or reverse the elements in the demolition sequence (i.e., the construction number) in a certain proportion to generate a new demolition sequence.

[0101] S205: Using the result of the optimization operation as a new candidate solution, verifying whether the new candidate solution satisfies the topology constraint. If so, go to S202; if not, go to S204.

[0102] To ensure that all components conform to the dominant dependency relationship between building components, after each generation of genetic optimization operation generates a new disassembly sequence scheme, a topological constraint check is performed on the new disassembly sequence scheme.

[0103] After multiple rounds of iterative optimization using a genetic algorithm, the disassembly sequence with the highest fitness value was selected as the actual disassembly sequence. This scheme maximizes the safety of each state during the disassembly process and minimizes the risk of instability.

[0104] S3: During the dismantling process actually adopted, the monitoring data of the building structure is combined with Bayesian reasoning and machine learning to conduct dynamic risk prediction and continuous correction of the building structure status.

[0105] During the demolition process, in order to accurately identify and dynamically control the risk of structural instability during building demolition, a joint enhanced risk identification mechanism combining Bayesian reasoning and machine learning was introduced based on "real-time monitoring". S3 specifically includes:

[0106] S301: During the process of the least risk disassembly solution, obtain the current status Monitoring data of building structures ;

[0107] Get the current status by deploying multiple types of sensors Monitoring data of the building structure (including all remaining components in the demolition sequence plan) , is a multidimensional vector, which is defined as follows:

[0108] (3)

[0109] in, It represents the maximum displacement of all remaining components in the demolition sequence (i.e., demolition sequence plan), reflecting the most serious deformation of the building structure in the current state; It represents the maximum value of the ratio of the actual strain of all remaining components in the demolition sequence to the yield strain of their materials, and is used to reflect the potential damage risk of the most dangerous component; Indicates the peak acceleration at the key parts of the building structure, reflecting the degree of inertial impact on the building structure. The key parts are selected by experts through analysis and judgment; It is the maximum horizontal inclination angle of the current building structure as a whole or a typical floor, used to reflect the overturning tendency of the structure; It indicates the maximum ratio of residual stress to material ultimate strength at key parts of the current building structure, and is used to measure the residual bearing capacity of the building structure.

[0110] S302: Based on the monitoring data ; Calculate the corresponding Bayesian posterior probability As a Bayesian risk score ;

[0111] Based on multidimensional vector , use the Bayesian inference formula to evaluate the current state The posterior probability , the formula is as follows:

[0112] (4)

[0113] in, Indicates that a structural instability event occurs; Indicates status monitoring data; Represents the static probability of instability (i.e., prior probability) that is independent of monitoring data, which is determined by the hierarchical weight and safety factor of the component to be disassembled in the directed acyclic graph in the next step The smaller the topological depth of the component to be disassembled next, The smaller the value, the higher the prior probability of instability is assigned; Indicates that monitoring data is obtained during instability The probability (i.e. likelihood probability) is obtained by the statistics of a large number of simulation samples accumulated during the genetic algorithm optimization process in step S2. In order to avoid the monitoring value in the multidimensional feature space It is difficult to completely overlap with the simulation sample, so it can be expanded appropriately The tolerance range, that is, will be in the multidimensional feature space with The Euclidean distance is less than The simulation sample is used as ,in is a preset positive number; Indicates current monitoring data The total probability of occurrence in all states (i.e., the monitoring probability) can be generated through statistical analysis; It represents the posterior probability after combining prior knowledge and state monitoring, which serves as the basis for risk assessment of the current state.

[0114] In addition, set the instability risk threshold ,when When the risk response mechanism is triggered, the current demolition operation is suspended, prompting temporary reinforcement and data review. The system then returns to step S2 to regenerate a new demolition sequence plan. This process is repeated to ensure that each demolition state remains within the risk control range.

[0115] when In order to improve the temporal continuity and prediction robustness of state risk assessment, the present invention introduces a recursive update mechanism in Bayesian reasoning. After that, the static prior instability probability is no longer used to calculate the posterior probability of the new state. Instead, the posterior probability of the previous state is used as the prior probability of the new state, as shown below:

[0116] (5)

[0117] By dynamically adjusting the prior probability, the continuity and traceability of the risk assessment results in the time dimension are guaranteed, and emergencies that arise during the dismantling process can be responded to more flexibly.

[0118] S303: The monitoring data Input the trained machine learning model to obtain the machine learning risk score value ;

[0119] In order to improve the efficiency of risk identification in long-term operation, the present invention uses machine learning methods to continuously learn the Bayesian reasoning process. ), collect monitoring data characteristics of building structures , and record the risk score value (i.e., posterior probability) obtained by Bayesian inference ;Build sample The training set composed of .

[0120] In order to construct a high-quality training set that is representative and discriminative, the present invention proposes a risk-guided cluster sampling mechanism to screen out representative samples with high risk focus and state coverage from historical samples accumulated during the disassembly process for model training of machine learning.

[0121] First, for each sample Calculate its risk score The scoring function comprehensively considers the topological depth and Bayesian posterior probability of the component to be disassembled in the next step. The formula is as follows:

[0122] (6)

[0123] in, Indicates the topological depth of the component to be disassembled in the next step; is a minimum non-zero constant, with a value of 0.001; Indicates recent The sliding average of the posterior probability is used to reflect the difference between the current state and the previous state. The average risk of the state after the first step of disassembly is calculated, thereby improving the sensitivity of the training sample sampling strategy to the risk evolution trend in subsequent sampling. The default value is 4; if the number of current disassembly steps is insufficient ,but The value is the actual number of steps. is the weighting coefficient used to express the "structural dependency level" and “Instability Probability Assessment” Assign weights between the two types of information and satisfy , Set by empirical data or training verification results, by default it can be set to 、 , this default setting means that the posterior risk is the main reference factor, but the topological position still has influence.

[0124] The historically accumulated monitoring data feature vector As input, K-means clustering method is used to cluster in high-dimensional state space to obtain Clusters , and the center of each cluster. For each sample , based on its similarity to the corresponding cluster center and risk score , calculate the representativeness score of the sample , the representativeness score is used to determine the priority (i.e., weight) of the sample in the training set. The formula for the representativeness score is as follows:

[0125] (7)

[0126] in, Represents a sample The monitoring data feature vector in and the cluster center of the cluster to which it belongs The Euclidean distance of is the weight coefficient and , the default value is set is 0.3, is 0.7.

[0127] In each cluster, keep the score The top several samples (in this example, 1000) are used as representative samples to form a training subset. This training subset ensures both broad sample coverage and focused representation of high-risk areas, providing an efficient training data foundation for the machine learning model (in this example, the GBDT model). This training subset is used to train a gradient boosted decision tree (GBDT) model, gradually learning the risk discrimination model used in Bayesian reasoning.

[0128] The GBDT model approximates the Bayesian output results through the integration and iteration of multiple rounds of weak classifiers. It has good nonlinear modeling capabilities and small sample learning effects. It can effectively simulate the complex mapping relationships implicit in the Bayesian reasoning process. After accumulating enough samples, it forms a fast and efficient risk auxiliary identifier. After training is completed, the GBDT model can independently output the machine learning risk score value. .

[0129] S304: Comprehensive Bayesian risk score and the risk score value output by the machine learning model , for the current state Conduct dynamic risk prediction and continuous correction.

[0130] like Figure 3 As shown in the figure, during the demolition process, when the machine learning model has been fully trained, the "joint discrimination mechanism" is activated, and the rules are as follows:

[0131] like , then the current risk judgment is recognized (that is, the posterior probability obtained by Bayesian reasoning is considered correct);

[0132] like , the current risk judgment is not recognized (that is, it is believed that there is a problem with the posterior probability obtained by Bayesian reasoning), the demolition operation is suspended, and a second judgment is made after manual review;

[0133] in, is the preset threshold.

[0134] Considering that the initial stage of the building component demolition sequence safety control method that integrates genetic algorithm, Bayesian reasoning and machine learning has not yet accumulated enough historical samples for training the machine learning model, a staged discrimination mechanism is set, and its rules are as follows:

[0135] 1. When the number of historical samples used to train the machine learning model does not reach the preset threshold When , only the posterior probability of Bayesian inference is used to judge the risk;

[0136] 2. When the number of historical samples used to train the machine learning model reaches a preset threshold , and the mean absolute error MAE is less than 0.05, the above-mentioned joint discrimination mechanism of Bayesian reasoning and machine learning is adopted.

[0137] The present invention provides a sequential safety control method throughout the entire building dismantling process. This method establishes a directed acyclic graph of component dependencies through topological sorting, combines mechanical simulation with genetic algorithm optimization operations to generate feasible, low-risk dismantling sequence paths, and dynamically perceives structural risks using a Bayesian inference mechanism. This method, supplemented by a machine learning model to enhance the stability of risk identification, achieves a coordinated linkage between dismantling sequence path generation and risk control. The present invention's data interaction and decision-making linkage form a closed-loop control chain driven by risk assessment to regulate the dismantling sequence path. By integrating genetic algorithms, Bayesian inference, and machine learning, this method achieves dynamic optimization and full-process safety assurance for building structure dismantling.

[0138] The above description is only a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with this technical field can easily think of changes or replacements within the technical scope disclosed by the present invention, which should be covered by the scope of protection of the present invention.

Claims

1. A building component demolition sequence safety control method integrating genetic algorithm, Bayesian reasoning and machine learning, characterized in that: include: S1: Use topological sorting to analyze the dependencies between components in a building structure and construct a directed acyclic graph of the disassembly order; S2: In the directed acyclic graph, randomly select The disassembly sequence schemes that meet the topological constraints are used as the initial population of the genetic algorithm, where ; Based on mechanical simulation, the safety of the state of the disassembly sequence scheme is evaluated, and combined with the optimization operation of the genetic algorithm, the actual disassembly scheme is obtained; The disassembly sequence scheme uses the component sequence vector in the directed acyclic graph express: in, Indicates the disassembly sequence plan; Indicates the Step 1: The number of the component to be removed; Indicates the number of disassembly steps in the disassembly sequence plan; The safety assessment of the status of the disassembly sequence scheme based on mechanical simulation includes: In the disassembly sequence scheme, for each disassembly state , use the finite element analysis method to simulate the stress distribution of the building structure and output the components to be disassembled in the next step The actual stress value , and calculate Safety factor : in, Indicates that the next step is to disassemble the component The yield strength of the material; Indicates that the next step is to disassemble the component The actual stress of S3: During the dismantling process actually adopted, the monitoring data of the building structure is combined with Bayesian reasoning and machine learning to conduct dynamic risk prediction and continuous correction of the building structure status.

2. The method according to claim 1, characterized in that The S2 includes: S201: In a directed acyclic graph, randomly select The disassembly sequence schemes that meet the topological constraints are taken as candidate schemes, and the candidate schemes are taken as the initial population of the genetic algorithm; S202: Calculating the fitness of the candidate solution based on the safety factor; Fitness function of the entire disassembly sequence scheme The definition is as follows: in, Indicates the disassembly sequence plan; Indicates the number of steps in the disassembly sequence plan; Indicates the current status The corresponding safety factor; S203: Determine whether the fitness is within a safe range. If so, use the candidate solution with the highest fitness as the actual disassembly sequence solution. If not, proceed to the next step. S204: Screening for the highest fitness The candidate solutions are optimized by genetic algorithm; S205: Using the result of the optimization operation as a new candidate solution, verifying whether the new candidate solution satisfies the topology constraint. If so, go to S202; if not, go to S204.

3. The method according to claim 1, characterized in that The S3 includes: S301: During the disassembly process actually adopted, obtain the current status Monitoring data of building structures ; S302: Based on the monitoring data ; Calculate the corresponding Bayesian posterior probability As a Bayesian risk score ; S303: The monitoring data Input the trained machine learning model and obtain the risk score value output by the machine learning model ; S304: Comprehensive Bayesian risk score and the risk score value output by the machine learning model , for the current state Conduct dynamic risk prediction and continuous correction.

4. The method according to claim 3, characterized in that The state in S301 Monitoring data The definition of is as follows: in, It represents the maximum displacement of all remaining components in the demolition sequence, reflecting the most serious deformation of the building structure in the current state; It represents the maximum value of the ratio of the actual strain of all remaining components in the demolition sequence to the yield strain of their materials, and is used to reflect the potential damage risk of the most dangerous component; Indicates the peak acceleration at the key parts of the building structure, reflecting the degree of inertial impact on the building structure. The key parts are selected by experts through analysis and judgment; It is the maximum horizontal inclination angle of the current building structure as a whole or a typical floor, used to reflect the overturning tendency of the structure; It indicates the maximum ratio of residual stress to material ultimate strength at key parts of the current building structure, and is used to measure the residual bearing capacity of the building structure.

5. The method according to claim 4, characterized in that In S302, based on the monitoring data , use the Bayesian inference formula to evaluate the current state The posterior probability , the formula is as follows in, Indicates that a structural instability event occurs; Indicates status monitoring data; Represents the static prior probability, which is determined by the hierarchical weight and safety factor of the component to be disassembled in the next step in the directed acyclic graph The smaller the topological depth of the component to be disassembled next, The smaller the value, the higher the prior probability is assigned; Represents the likelihood probability, which is obtained by the statistics of a large number of simulation samples accumulated during the genetic algorithm optimization process in step S2. In order to avoid the monitoring value in the multidimensional feature space It is difficult to completely overlap with the simulation sample, and it will be in the multi-dimensional feature space. The Euclidean distance is less than The simulation sample is used as ,in is a preset positive number; represents the probability of monitoring, which can be generated through statistical analysis; Represents the posterior probability, which serves as the basis for risk assessment of the current state; Receive new monitoring data After that, the static prior probability is no longer used to calculate the posterior probability of the new state. Instead, the posterior probability of the previous state is used as the prior probability of the new state, as shown below: 。 6. The method according to claim 5, characterized in that The training of the machine learning model in S303 includes: In each state , collect monitoring data characteristics of building structures , and record the posterior probability obtained by Bayesian inference , build a sample The training set composed of First, for each sample Calculate its risk score The scoring function comprehensively considers the topological depth and Bayesian posterior probability of the component to be disassembled in the next step. The formula is as follows: in, Indicates the topological depth of the component to be disassembled in the next step; is a minimum non-zero constant, with a value of 0.001; Indicates recent The sliding average of the posterior probability is used to reflect the difference between the current state and the previous state. The average risk of the state after the disassembly step; if the current number of disassembly steps is insufficient ,but The value is the actual number of steps. is the weighting coefficient used to express the "structural dependency level" and "instability probability assessment" Assign weights between the two types of information and satisfy , Set by empirical data or training verification results; The historical monitoring data is used as input and clustered in high-dimensional state space using K-means clustering method to obtain Clusters , and the center of each cluster; for each sample ,based on Similarity and risk score with the corresponding cluster center , calculate the sample Representative scores , the representativeness score is used to determine the priority of samples in the training set; the formula for the representativeness score is as follows: in, Represents a sample Monitoring data characteristics in and the cluster center of the cluster to which it belongs The Euclidean distance of is the weight coefficient and ; In each cluster, keep the score The highest number of samples are used as representative samples to form a training subset; the training subset is input into the machine learning model for training.

7. The method according to claim 6, characterized in that In S304, during the dismantling process, when the machine learning model has been fully trained, the "joint discrimination mechanism" is activated, and the rules are as follows: like , then the current risk judgment is recognized, that is, the posterior probability obtained by Bayesian reasoning is considered correct; like , then the current risk judgment is not recognized, that is, it is believed that there is a problem with the posterior probability obtained by Bayesian reasoning, and the demolition operation is suspended. After manual review, a second judgment is made; in, is the preset threshold.

8. The method according to claim 1, characterized in that Said S1 comprises: S101: Acquire and analyze relevant data of the building structure, determine the interdependence of each component during the demolition process and establish a dependency matrix; S102: Based on the dependency matrix, topologically sort the components and draw a directed acyclic graph.

9. The method according to claim 2, characterized in that The optimization operation of the genetic algorithm in S204 includes crossover and mutation.

Images