Earthquake-damaged building demolition scheme rapid making method based on genetic algorithm and reinforcement learning

Abstract

The invention belongs to the field of earthquake engineering and civil engineering, and particularly relates to an earthquake damage building demolition scheme rapid making method based on a genetic algorithm and reinforcement learning. The method comprises the following steps of S1, establishing a data information base of a target earthquake damage demolition building, wherein the data information base mainly comprises various building structure types, structure plane arrangement, building site types and beam column component sizes; s2, establishing a structure progressive collapse analysis finite element model according to the collected information; s3, according to the collected information, determining the range of components allowed to be dismantled during collapse analysis of the dismantled structure; s4, according to the progressive collapse analysis finite element model, calculating a single dismantling effect; and S5, selecting a genetic algorithm or a Q-learning reinforcement learning algorithm according to whether the actual situation needs to consider the influence of the demolition sequence, and obtaining an optimal demolition scheme. The method can be used for quickly formulating a demolition scheme of a seriously damaged building in restoration and reconstruction after an earthquake disaster.

Description

Technical Field

[0001] The present invention belongs to the fields of earthquake engineering and civil engineering, and particularly relates to a method for quickly formulating a demolition plan for earthquake-damaged buildings based on genetic algorithms and reinforcement learning. Background Art

[0002] After an earthquake, severely damaged building structures can no longer be used and become standing "ruins". In order to quickly repair and reconstruct earthquake-affected areas, it is necessary to quickly formulate an effective demolition plan for earthquake-damaged buildings. The accurate identification of weak points is an essential prerequisite for the demolition of earthquake-damaged buildings. This step requires integrating professional knowledge in the engineering field and deeply analyzing the complexity of the target building structure. Through systematic structural analysis, engineers and demolition experts can determine potential weak points, including multiple factors such as material properties, structural strength, load distribution, and structural connections. Such detailed analysis not only helps to accurately identify possible damage points but also helps to evaluate which areas are most vulnerable to damage, thereby formulating the most appropriate demolition strategy to ensure the success and safety of the demolition project of earthquake-damaged buildings. However, an incorrect building demolition plan may lead to serious property losses and prevent the target structure from collapsing as expected in one go, which means additional engineering and resources are required to clear the debris and delays the progress of post-disaster repair and reconstruction. Therefore, the main breakthrough of the present invention is to quickly formulate a demolition plan for earthquake-damaged buildings through genetic algorithms, improve the accuracy of weak point identification, and the feasibility of the demolition plan.

[0003] The identification of weak points plays a crucial role in the demolition project of earthquake-damaged buildings because it directly affects the safety and efficiency of the demolition process. Traditional methods for identifying weak points usually rely on empirical methods or numerical simulation methods of mechanical analysis. Although these two types of methods are widely used, there are some problems. Traditional methods for identifying weak points usually require a large amount of modeling and analysis work, which can lead to problems of complexity and time consumption. Traditional finite element analysis methods require detailed three-dimensional models and complex calculations in building demolition projects, which may consume a large amount of time and resources. The accuracy of the model is crucial for the effectiveness of traditional methods. Traditional numerical simulation methods require accurate material properties and structural parameters in building demolition projects, otherwise it may lead to inaccurate predictions. Traditional methods are inefficient in dealing with large-scale data. Methods based on numerical simulation may require a large amount of computing time when dealing with large building demolition projects, which may not be applicable to rapid decision-making and planning in actual engineering.

[0004] In recent years, the development of machine learning and data-driven approaches has been rapid. By using machine learning or data-driven means to quickly identify weak points, not only can the efficiency and accuracy of engineering be improved, but also the cumbersome modeling and analysis work can be significantly saved. The advantage of this method lies in its ability to utilize large-scale data and intelligent algorithms to automatically identify potential weak points, thereby greatly shortening the time and resources required for the identification process. Summary of the Invention

[0005] Aiming at the problems of insufficient accuracy and low efficiency in formulating demolition plans using traditional methods based on mechanical analysis or numerical simulation, the present invention aims to provide a method for quickly formulating a demolition plan by identifying weak points during the demolition of earthquake-damaged buildings based on genetic algorithms or the Q-learning algorithm of reinforcement learning, so as to solve the defects and deficiencies of the above traditional calculation methods.

[0006] Therefore, the technical solution invented in this paper is as follows: A method for quickly formulating a demolition plan for earthquake-damaged buildings based on genetic algorithms, and the required process parameters include the following: A method for quickly formulating a demolition plan for earthquake-damaged buildings based on genetic algorithms, including the following steps: S1: Establish a database of data information for the target earthquake-damaged building to be demolished, mainly including various building structure types, structural plane layouts, building site categories, and beam-column member sizes; S2: Establish a finite element model for progressive collapse analysis of the structure according to the collected information; S3: Determine the range of components that are allowed to be demolished during the collapse analysis of the demolished structure according to the collected information; S4: Calculate the single demolition effect according to the finite element model of progressive collapse analysis; S5: Select the corresponding optimization algorithm according to whether the influence of the demolition sequence needs to be considered in actual situations to obtain the optimal demolition plan: If the influence of the demolition sequence is not considered, select the genetic algorithm to optimize the selection of the demolition range of building components; if the influence of the demolition sequence is considered, select the Q-learning reinforcement learning algorithm to optimize the selection of the demolition range and demolition sequence of building components.

[0007] Preferably, the specific method of step S2 is as follows: Use finite element analysis software, combine multi-scale modeling methods and the structural information in step S1 to establish a progressive collapse analysis model of the earthquake-damaged building structure.

[0008] Preferably, the genetic algorithm mainly includes population selection, chromosome crossover, and chromosome mutation; First, encode based on the demolition model, randomly generate an initial population; calculate the individual fitness, perform selection operations and mutation operations, and return the optimization result after reaching the termination condition; an individual represents each building demolition plan, a chromosome represents the encoding of each demolition plan, a gene represents the encoding element, that is, whether each column is to be demolished, and the viability represents the result of each plan in the fitness function.

[0009] Preferably, the ratio of the vertical displacement of the earthquake-damaged building after each demolition to the number of demolished columns is used as the evaluation value of the fitness function, and the set parameters of the columns demolished each time are used as the independent variables of the algorithm. The set parameters of the demolition objects obtained by iterative calculation using the genetic algorithm are further cyclically calculated using the iterated parameter values until the results meet the requirements of the final demolition effect.

[0010] Preferably, the specific method for encoding based on the demolition model is to use binary vectors for encoding: 。

[0011] Preferably, the Q-learning reinforcement learning algorithm includes initializing the relevant parameters of the Q-Learning algorithm, selecting a polynomial decay function to control the change of the learning rate, constructing a Q-table to store state-action values, using the ε-greedy strategy to select an action, and updating the estimated value of the corresponding state-action value in the Q-table according to the current state, action, reward, next state, and end flag. The optimization solution of the objective function is completed by searching in the Q-table to find the action with the highest Q value.

[0012] The beneficial effects of the present invention are as follows: Compared with the traditional empirical method, traditional mechanical analysis, and numerical simulation methods, the present invention proposes an encoding method based on matrix positions, where different positions represent different components to be demolished in the structure, which can more intuitively represent the conversion between the demolished components and the genes of the population individuals. In addition, the present invention proposes to use the genetic algorithm or the Q-learning algorithm to solve the optimal building demolition plan, which has the advantages of global search, high search efficiency, and high accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 is a schematic flow chart of the method for quickly formulating a building demolition plan based on the genetic algorithm and reinforcement learning of the present invention; Figure 2 is a schematic diagram of the prototype structure of the method for quickly formulating a building demolition plan based on the genetic algorithm and reinforcement learning of the present invention; Figure 3 is the coding intention of the method for quickly formulating a building demolition plan based on the genetic algorithm and reinforcement learning of the present invention; Figure 4 is a schematic flow chart of the genetic algorithm; Figure 5 is a schematic flow chart of the Q-learning reinforcement learning algorithm; Figure 6 is a simple example of the genetic algorithm; Figure 7 is a simple example of the Q-learning reinforcement learning algorithm, where (a) matrix R; (b) Q-table. Detailed implementation mode

[0014] To make the purpose, technical solution and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. The specific embodiments described herein are only used to explain the present invention and are not used to limit the scope of the present invention.

[0015] Unless otherwise defined, all technical terms and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field of the present invention. The terms used in the specification of the present invention herein are only for the purpose of describing specific embodiments and are not intended to limit the present invention.

[0016] As Figure 1 shown, a method for quickly formulating a demolition plan for damaged buildings based on a genetic algorithm is characterized by including the following steps: Step1: Establish a database of data information for the target demolished building, mainly including various building structure types, structural plane layouts, building site categories, beam-column member sizes, etc.

[0017] Step2: Establish a finite element model for progressive collapse analysis of the structure according to the collected information. Establish a steel frame finite element model according to the database of data information for the target demolished building in Step 1. Use OpenSees finite element software to establish a numerical model of the damaged building.

[0018] As Figure 2 shown, the prototype of the finite element model in this embodiment is a four-story steel frame structure. The steel beams and steel columns are simulated using Beam elements, the reinforced concrete slabs are simulated using Shell elements, and the joints are simulated using macroscopic elements. In the design of the structural lateral force resistance system, it is designed according to the relevant requirements of the special moment-resisting frame system with reduced flange beam hinges. The moment-resisting frames are arranged in the side spans of the structure. Other structural parts are designed according to the vertical force resistance system. The columns and beam sections of the vertical force resistance systems of the three structures all adopt W14×90 and W24×55, and the connections of the internal vertical force resistance frames adopt conventional single-shear plate beam-column connections. The beam-columns adopt ASTM A992 steel (nominal yield strength of 345 MPa and elastic modulus of 2×10 11 N / m2)

[0019] Step3: Determine the allowable range of components to be demolished during the progressive collapse analysis of the demolished structure according to the collected information.

[0020] As Figure 3 shown, all the load-bearing columns of the structure (a total of 16) are within the allowable range of components to be demolished.

[0021] Step4: Calculate the single demolition effect according to the finite element model for progressive collapse analysis.

[0022] AsFigure 3 As shown in the figure, all load-bearing columns are numbered in the order from the lower floor to the upper floor and from left to right. The specific component numbers to be demolished are determined according to random numbers. The corresponding components are demolished through the "Remove" command in OpenSees, and the demolished objects set in the working condition are removed. Gravity analysis is performed on the remaining structure, and the vertical displacement of the structure after the components are demolished is calculated, which is used as the evaluation index for the demolition effect this time.

[0023] Step5: Determine whether to specifically select the genetic algorithm or the Q-learning reinforcement learning algorithm for optimization and solution according to whether the influence of the demolition sequence needs to be considered in actual situations: (1) Based on the genetic algorithm, optimize the selection of the demolition component plan when demolishing damaged buildings to obtain an optimal demolition plan. Here, the demolition plan specifically refers to the set of demolition points selected each time. First, encode based on the demolition model, then randomly generate an initial population, calculate the individual fitness, and then perform selection and mutation operations. Finally, return the optimization result after reaching the termination condition.

[0024] As Figure 3 shown, the specific method of encoding based on the demolition model is to use binary vectors for encoding: ; Among them, size represents the total number of demolition objects. Taking a four-story steel frame structure as an example, there are a total of 4×4 = 16 load-bearing columns; all demolition objects are sorted according to the floor and position. The 4 columns on the first floor are numbered 1-4 in order from left to right, and the load-bearing columns on the remaining floors are numbered in the same way. represents whether the load-bearing column numbered i in the demolition objects is used as a demolition point in this demolition. If , it means that the i-th column is used as a demolition point in this demolition; if , it means that the i-th column is not used as a demolition point in this demolition; for example, the chromosome of a certain demolition model is , which means that the third column on the first floor and the first column on the second floor are demolished.

[0025] As Figure 4 shown, the specific steps of the genetic algorithm are as follows: (1) Population initialization: According to this encoding method, within the set range, randomly generate the information of the demolition objects to obtain an initial population composed of several chromosomes, and perform appropriate stretching processing on the population to increase individual differences and ensure the convergence accuracy; (2) Define the fitness function: The fitness value is related to the demolition efficiency of a single demolition, and is determined by the vertical displacement of the structure and the number of demolitions. The higher the ratio of the vertical displacement of the structure to the number of demolitions, the higher the fitness; the fitness function is as follows: (1); Among them, in formula (1), represents the demolition object coding information for a single demolition,[ represents the structural vertical displacement distance for a single demolition,[ represents the number of columns demolished in a single demolition; (3) Selection operation: The roulette wheel selection method is adopted for the selection operation. The probability of each chromosome individual being selected is proportional to the value of its fitness function. The optimized individuals are inherited to the next generation or new individuals are generated through pairing and crossover and then inherited to the next generation; (4) Crossover operation: Taking the floor as the unit, the four columns on each floor are regarded as a gene segment. The number of gene segments on each chromosome is equal to the number of floors. The two-point crossover method is selected to act on the population, and the gene segments between any positions of two chromosomes are randomly exchanged, and the search range is determined in the crossover operator; (5) Mutation operation: The mutation operator acts on each gene segment of the chromosome according to a certain mutation probability. By adjusting the mutation probability, observing and calculating the convergence curve, ensuring the convergence speed, and avoiding the problem of overfitting; (6) Evolution operation: The optimal demolition plan is searched by evolution. After the population undergoes selection, crossover, and mutation operations, the next generation population is obtained.

[0026] The genetic algorithm is used to solve the building demolition plan without considering the demolition order. For example, Figure 6 as shown. Assume that the population consists of two individuals. The implementation steps are as follows according to Figure 4 the flowchart shown: Population initialization: Randomly generate two individuals in the population. The encodings of the individuals are respectively, ; ; Crossover operation: Taking the floor as the unit, the four columns on each floor are regarded as a gene segment. The number of gene segments on each chromosome is equal to the number of floors. The two-point crossover method is selected to act on the population, and the gene segments between the corresponding positions of two chromosomes are randomly exchanged. Here, the gene segments corresponding to the third floor are exchanged; Mutation operation: The mutation operator acts on each gene segment of the chromosome according to a certain mutation probability. Here, the mutation operation is tentatively not performed; Calculate the fitness function: Demolish the components at the corresponding positions according to the individual encoding, and calculate the fitness function value through the OpenSees function.

[0027] Selection operation: The roulette wheel selection method is adopted for the selection operation. The probability of each chromosome individual being selected is proportional to the value of its fitness function. The optimized individuals are inherited to the next generation or new individuals are generated through pairing and crossover and then inherited to the next generation. The optimization results are shown in Table 1 below.

[0028] Table 1 shows the optimization results of the genetic algorithm .

[0029] As Figure 5 shown, the specific steps of Q-learning are as follows: (1), Initialize relevant parameters: Given the exploration rate , learning rate and discount factor , and set the number of algorithm iterations N to initialize the state-action value function Q(s,a), and assign all function values in this function to 0; (2), Select an action: Use the -greedy strategy to select an action, and obtain a new state and the corresponding reward value r according to the current state and the selected action; (3), Update the state: Each time the state is updated, update the estimated value of the state-action value corresponding to the current state, action, reward, and next state in the Q table. The specific update method is: Q s , a ← Q s , a + α [ R + γ max Q s ' , a - Q s , a ] (2); Among them, in formula (2), represents the Q value corresponding to the action with the largest Q value among all possible actions taken in state ; (4) Optimization: By searching in the Q-table, find the action with the highest Q-value, which is the optimal action in the current state. This action will be used as the next action of the agent. The state corresponding to this action is the optimal solution state of the problem, completing the optimization and solution of the objective function. The Q-table stores the Q-values of state-action pairs, representing the cumulative reward that can be obtained by taking a certain action from a certain state. S5 needs to search in the Q-table for the action with the highest Q-value in the current state; for the current state s, traverse all possible actions a, and then find the action a* with the highest Q-value; (5) Optimization and solution: After finding the optimal action a*, use it as the next action of the agent, and at the same time use the state corresponding to a* as the optimal solution state of the problem. This state can minimize the objective function. Since the Q-value represents the cumulative reward obtained by taking a certain action, the action corresponding to the highest Q-value will obtain the largest cumulative reward in the long term, thus making the objective function the optimal building demolition plan.

[0030] The Q-learning algorithm is used to solve the building demolition plan considering the demolition sequence. The case is as Figure 7 shown below: State space: The set of environmental states obtained by the agent , using the same coding method as in step S6, and using whether the column at each demolition point is demolished as the basis for state division, representing the i-th result after demolition. The total number of state function is as follows: (3); Among them, in formula (3) represents the total number of demolishable points; Action space: The set of environmental states obtained by the agent , using which column at the demolition point is demolished as the basis for action division. The total number of actions is equal to ; Reward function: Q-learning requires the agent to continuously try and error to complete the learning process. During this period, an immediate reward value is obtained by making a good or bad evaluation of the action selected in the state . Here, the vertical displacement of the structure is used as the reward value; Initialize relevant parameters: Given the exploration rate , learning rate and discount factor , and set the number of algorithm iterations N to initialize the state-action value function Q(s,a), and assign all function values in this function to 0; Select action: Use The -greedy policy selects an action and obtains a new state based on the current state and the selected action. And the corresponding reward value r. Here, state 0 (without demolition treatment) is selected as the initial state. Update the state: Each time the state is updated, the estimated value of the state-action value in the Q-table is updated according to the current state, action, reward, and the next state. The initial state in the previous step corresponds to k possible behaviors (transfer to 16 states with a total of 1 column demolished through k possible actions). Randomly, we select to transfer to state 3 (demolish the third column on the first floor), and then observe the fourth row of matrix R (corresponding to state 3), which corresponds to k - 1 possible behaviors: transfer to 15 states with a total of 2 columns demolished. Here, the learning rate is set to 0.1. Q(0, 3)= R(0, 3)+ 0.1*(-max{|Q(3, 1)|, |Q(3, 2)|, … |Q(3, 15)|}= -0.18 - 0.1×2.97 = -0.477; Optimization: By searching in the Q-table, find the action with the highest Q value, that is, the optimal action in the current state, and use it as the next action of the agent. The state corresponding to this action is the optimal solution state of the problem, completing the optimization solution of the objective function. The Q-table stores the Q values of state-action pairs, indicating the cumulative reward that can be obtained by taking a certain action from a certain state. S5 needs to search in the Q-table for the action with the highest Q value in the current state; for the current state s, traverse all possible actions a, and then find the action a* with the highest Q value. Optimization solution: After finding the optimal action a*, use it as the next action of the agent, and at the same time use the state corresponding to a* as the optimal solution state of the problem. This state can make the objective function reach the minimum value. Since the Q value represents the cumulative reward obtained by taking a certain action, the action corresponding to the highest Q value will obtain the largest cumulative reward in the long term, thus making the objective function the optimal demolition plan for the building.

[0031] The above is only a hypothetical implementation case of the present invention and is not intended to limit the present invention. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for quickly formulating a demolition plan for earthquake-damaged buildings based on a genetic algorithm, characterized in that, It includes the following steps: S1: Establish a database of data information on target earthquake-damaged buildings to be demolished, mainly including various building structure types, structural plane layouts, building site categories, and beam-column member sizes; S2: Establish a finite element model for progressive collapse analysis of the structure according to the collected information; S3: Determine the allowable range of components to be demolished during the collapse analysis of the demolished structure according to the collected information; S4: Calculate the effect of a single demolition according to the finite element model for progressive collapse analysis; S5: Select the corresponding optimization algorithm according to whether the influence of the demolition sequence needs to be considered in the actual situation to obtain the optimal demolition plan: if the influence of the demolition sequence is not considered, select the genetic algorithm to optimize the selection of the demolition range of building components; if the influence of the demolition sequence is considered, select the Q-learning reinforcement learning algorithm to optimize the selection of the demolition range and demolition sequence of building components.

2. The method for rapidly formulating a demolition plan for earthquake-damaged buildings based on genetic algorithm and reinforcement learning according to claim 1, wherein, The specific method of step S2 is as follows: Use finite element analysis software, combine the multi-scale modeling method and the structural information in step S1 to establish a progressive collapse analysis model of the earthquake-damaged building structure.

3. The method for quickly formulating a demolition plan for earthquake-damaged buildings based on genetic algorithm and reinforcement learning according to claim 1, characterized in that, The genetic algorithm mainly includes population selection, chromosome crossover, and chromosome mutation; first, encode based on the demolition model and randomly generate an initial population; calculate the individual fitness, perform selection operations and mutation operations, and return the optimization result after reaching the termination condition; An individual represents each demolition plan for an earthquake-damaged building, a chromosome represents the encoding of each demolition plan, a gene represents the encoding element, that is, whether each column is to be demolished, and the viability represents the result of each plan in the fitness function.

4. The method for rapidly formulating a demolition plan for earthquake-damaged buildings based on genetic algorithms and reinforcement learning according to claim 3, wherein, Take the ratio of the vertical displacement of the earthquake-damaged building after each demolition to the number of demolished columns as the evaluation value of the fitness function, take the set parameters of each demolished column as the independent variable of the algorithm, use the genetic algorithm to perform iterative calculations to obtain the set parameters of the demolition object, and use the iterated parameter values to further perform cyclic calculations until the result meets the requirements of the final demolition effect.

5. The method for quickly formulating a demolition plan for earthquake-damaged buildings based on genetic algorithm and reinforcement learning according to claim 3, wherein, The specific method of encoding based on the demolition model is to use binary vectors for encoding: .

6. The method for quickly formulating a demolition plan for earthquake-damaged buildings based on genetic algorithm and reinforcement learning according to claim 1, wherein The Q-learning reinforcement learning algorithm includes initializing the relevant parameters of the Q-Learning algorithm, selecting a polynomial decay function to control the change of the learning rate, constructing a Q-table to store state-action values, using the ε-greedy strategy to select an action, and updating the estimated value of the corresponding state-action value in the Q-table according to the current state, action, reward, next state, and end flag, and complete the optimization solution of the objective function by searching in the Q-table and finding the action with the highest Q value.

Images