A method and system for optimizing measurement points of venue lattice shell roof

Abstract

The present invention is a method and system for optimizing measuring points of a stadium grid shell roof, which relates to the technical field of engineering structures, including taking all nodes of the stadium grid shell roof as an initialization population, and dividing the initialization population into multiple populations to simultaneously perform genetic iteration operations; constructing a stadium grid shell roof measuring point layout optimization model based on a non-dominated recombination genetic algorithm, and taking the fitness of the population as an optimization target; setting the fitness function based on a MAC matrix, and selecting the non-diagonal elements in the minimized MAC matrix as the criterion of the optimization function; setting the number of iterations, crossover probability, and mutation probability of the non-dominated recombination genetic algorithm according to the optimization function criterion; when the number of iterations meets the set conditions, the optimal solution is obtained, which is the optimal solution for measuring points of the stadium grid shell roof. The effect of sensor layout and the speed of solution are improved, especially for dealing with large-scale candidate measuring point problems of complex grid shell structures.

Description

Technical Field

[0001] The present invention relates to the field of safety engineering technology, and in particular to a method and system for optimizing measuring points of a stadium lattice shell roof. Background Art

[0002] With the development of modern society and the improvement of people's living standards, the construction of various sports stadiums has received more and more attention. These stadiums are not only used as venues for sports events, but also carry a variety of functions such as cultural performances and commercial activities. Therefore, it is particularly important to ensure the safety and operational efficiency of such structures. As a spatial structure with high strength, light weight and beautiful appearance, the lattice shell structure is widely used in the roof design of sports stadiums. However, the complexity of the lattice shell structure and the large number of measurement points to be selected have brought great challenges to the optimal arrangement of sensors.

[0003] When dealing with measurement point solutions for complex lattice shell structures, existing sensor placement methods have major problems such as poor sensor placement optimization, slow solution speed, and lack of research on external excitation identification. In the lattice shell structure of a sports stadium, the scale of candidate measurement points is huge, and usually dozens or even hundreds of sensors need to be arranged. Due to the huge number of degrees of freedom of the actual structural model, traditional optimization algorithms often show problems of low computational efficiency and slow solution speed when dealing with such a large number of candidate measurement points. In order to reduce computational complexity, existing studies often use simplified models to reduce the number of candidate measurement point locations. However, for heterogeneous spatial lattice shell structures, this simplified processing is difficult and may affect the accuracy of the optimization results. Therefore, there is an urgent need for a high-performance optimization algorithm to solve the problem of large-scale candidate measurement points and improve the effect and efficiency of sensor placement. Summary of the invention

[0004] In order to overcome the shortcomings of the above-mentioned prior art, such as weak adaptability and insufficient forecasting accuracy, the main purpose of the present invention is to provide a method and system for optimizing measurement points of a venue grid shell roof.

[0005] To achieve the above object, the present invention adopts the following technical solution, which is implemented by a sensor optimization arrangement method based on a multi-population non-dominated recombination genetic algorithm (MP-RGA), comprising the following steps:

[0006] The population initialization divides the measurement point positions into multiple populations (population 1, population 2, ... population N) during the initialization phase, and performs genetic iteration operations at the same time to improve the algorithm's optimization ability. The number of multiple populations N needs to be set during population initialization. In theory, the more multiple populations the better, but when the total number of individuals in the population remains unchanged, too many population divisions will result in a small number of individuals in a single population, which seriously reduces the algorithm's optimization search ability.

[0007] The calculated fitness is reflected by the fitness value calculated by the fitness function f1. When optimizing the monitoring points of the complex lattice shell structure of the stadium described in the technical solution, a MAC matrix is ​​constructed based on the modal vector estimated at the measuring point position, and the non-diagonal elements in the minimized MAC matrix are selected as the fitness function, that is:

[0008]

[0009] Where, MAC ij represents the element in the i-th row and j-th column of the MAC matrix; represents the maximum value in the non-diagonal unit of the MAC matrix; X represents the measurement point location layout plan.

[0010] The reverse learning generates a reverse population corresponding to the original population, retains individuals with better fitness than the original population, and eliminates poor individuals in the original population, thereby achieving a full search of the solution space.

[0011] The population communication is implemented through the population communication ratio α and the population communication frequency f. The population communication ratio and the population communication frequency affect the evolutionary direction and similarity of individuals in different populations during the multi-population communication process. Increasing the population communication ratio and the communication frequency will increase the number of excellent individuals obtained by the worst population in each generation from the best population, and the update speed of the worst population to the best population will be accelerated.

[0012] The optimal population refers to the population according to the fitness function f fit Evaluate population 1, population 2, … population N to obtain the optimal population.

[0013] The worst population refers to the population according to the fitness function f fit Evaluate population 1, population 2, … population N to obtain the worst population.

[0014] The improved population refers to introducing elite individuals from the best population into the worst population to replace a corresponding number of poor individuals.

[0015] The adaptive population size refers to dynamically adjusting the size of the population according to a scaling ratio of 0.5-1.5. When the average fitness of the parent individuals of the population is better than the average fitness of the parent individuals of the previous generation population, it means that the parent of the previous generation population has achieved a better search effect, and the population size can be reduced. On the contrary, the iteration has achieved a poor search effect, and the population size should be increased to improve the search ability. Specifically, it is determined according to the fitness value and the sigmoid function according to the following steps:

[0016] (a) Calculate the fitness change rate

[0017] (b) Calculate the scaling ratio β = sigmoid(5×α) + 0.5

[0018] (c) Determine the updated population size G N =G N-1 ×β

[0019] Among them, f fit_(N-1) represents the fitness value of the N-1 generation, f fit_(N) represents the fitness value of the Nth generation, α represents the fitness change rate, β represents the scaling ratio, G N-1 represents the population size of the N-1 generation, G N represents the population size of generation N.

[0020] The adaptive crossover refers to ensuring the diversity of individuals through the non-dominated recombination method. In the adaptive non-dominated recombination operation, each iteration performs non-dominated recombination on a part of the individuals in the population to ensure population diversity and avoid the complete destruction of excellent gene fragments in the population. The proportion of non-dominated recombination can be determined by measuring the similarity of the population fitness variance. The greater the similarity, the more individuals in the population undergo non-dominated recombination. The calculation formula is:

[0021] N r =-ln(D(fitness))·T·N parent

[0022] Where N r represents the number of individuals undergoing non-dominated recombination, D(fitness) represents the variance of the fitness of the parental individuals, T represents a constant 1 / 20, and N parent Represents the number of parent individuals in the population.

[0023] The adaptive mutation refers to the use of a smaller mutation probability in the early stage of optimization and a larger mutation probability in the later stage. This mutation strategy can ensure that the algorithm converges quickly to the optimal solution in the early stage and is conducive to jumping out of local convergence in the later stage. Its adaptive conversion function also uses the sigmoid function, and the specific calculation is as follows.

[0024]

[0025] In the formula, M is a constant of 0.1, which means that the probability of mutation generally does not exceed 0.1.

[0026] The optimal arrangement scheme refers to the optimal scheme obtained when a specified number of iterations are completed and the set conditions are met.

[0027] Compared with the prior art, the beneficial effects of the present invention are as follows: by optimizing the measurement point positions, it is ensured that the measurement point arrangement scheme can more effectively reflect the dynamic characteristics of the structure, thereby improving the accuracy and reliability of structural health monitoring. By minimizing the non-diagonal elements in the MAC (Modal Assurance Criterion) matrix, the discrimination between modal vectors is optimized, modal confusion is reduced, and the accuracy of modal parameter identification is improved. The robustness and efficiency of the algorithm are improved, and the search efficiency and global search capability are improved by dividing a large population into multiple populations and performing genetic iteration operations at the same time. Through the exchange of individuals between populations, richer genetic diversity is introduced, which helps to jump out of the local optimal solution and find the global optimal solution. The design of the fitness function takes into account the characteristics of the MAC matrix, so that the algorithm can effectively evaluate and guide specific structural optimization problems. Dynamic adjustment is performed according to the degree of fitness convergence, ensuring that the algorithm finds a satisfactory solution in sufficient time while avoiding unnecessary calculations. Through adaptive adjustment, the diversity of the population is maintained, while ensuring the convergence of the algorithm. The population size is dynamically adjusted according to the change of the fitness value, which helps to maintain an appropriate search range and search accuracy at different stages of the algorithm. By introducing elite individuals from the optimal population into the worst population, the convergence process of the algorithm is accelerated while maintaining the excellent genetic characteristics of the population.

[0028] In summary, the layout of the measuring points of the venue's grid shell roof has been optimized, the performance of the structural health monitoring system has been improved, and the efficiency of the algorithm and the reliability of the results have been ensured. The realization of these technical effects is of great significance for the long-term health monitoring and safety assessment of engineering structures. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] The drawings described herein are used to provide further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute improper limitations on the present application.

[0030] Figure 1 It is a schematic diagram of the process structure of the non-dominated recombination genetic algorithm of the present invention;

[0031] Figure 2 It is a schematic diagram comparing the effective effects of various algorithms in the embodiments of the present invention. DETAILED DESCRIPTION

[0032] The present invention will be further described below in conjunction with the accompanying drawings and implementation modes.

[0033] The optimization of monitoring points for complex lattice shell structures of sports stadiums is achieved by proposing a sensor optimization layout method based on multi-population non-dominated recombination genetic algorithm (MP-RGA). Figure 1 .

[0034] Population initialization divides the measurement point locations into multiple populations (population 1, population 2, ... population N) during the initialization phase, and performs genetic iteration operations at the same time to improve the algorithm's optimization ability. The number of populations N needs to be set during population initialization. In theory, the more populations the better, but when the total number of individuals in the population remains unchanged, too many population divisions will result in a small number of individuals in a single population, which will seriously reduce the algorithm's optimization search ability.

[0035] It is lower than the calculated fitness, which is reflected by the fitness value calculated by the fitness function f1. In this technical solution, when optimizing the monitoring points of the complex grid shell structure of the stadium, the MAC matrix is ​​constructed based on the modal vector estimated at the measuring point position, and the non-diagonal elements in the minimized MAC matrix are selected as the fitness, that is:

[0036]

[0037] In the formula, MAC ij Represents the element in the i-th row and j-th column of the MAC matrix; represents the maximum value in the non-diagonal unit of the MAC matrix; X represents the sensor layout scheme. The measurement point location is the sensor location. The purpose of this fitness function is to minimize the sum of the attribute differences between the measurement points, thereby optimizing the measurement point layout scheme.

[0038] Based on the measurement point locations, the MAC matrix is ​​constructed, including the following steps:

[0039] Obtain structural dynamic response data at the measuring point and estimate modal parameters, including modal frequency, damping ratio and modal vector, through signal processing;

[0040] Construct an M×M MAC matrix, where M is the number of identified modes;

[0041] For each off-diagonal element (i,j) in the MAC matrix, calculate the dot product of the i-th and j-th mode vectors Divide the dot product result by the product of the norms of the i-th and j-th modal vectors to obtain the normalized MAC value;

[0042] The calculated MAC value is filled into the corresponding position (i, j) of the MAC matrix to obtain the MAC matrix.

[0043] Reverse learning generates a reverse population corresponding to the original population, retains the individuals with better fitness than the original population, and eliminates the poorer individuals in the original population, thereby achieving a full search of the solution space.

[0044] Population communication is implemented through the population communication ratio α and the population communication frequency f. The population communication ratio and the population communication frequency affect the evolutionary direction and similarity of individuals in different populations during the multi-population communication process. Increasing the population communication ratio and the communication frequency will increase the number of excellent individuals obtained by the worst population from the best population in each generation, and the update speed of the worst population to the best population will be accelerated.

[0045] The optimal population refers to the population according to the fitness function f fit Evaluate population 1, population 2, … population N to obtain the optimal population.

[0046] The worst population refers to the population according to the fitness function f fit Evaluate population 1, population 2, … population N to obtain the worst population.

[0047] Improving a population means introducing elite individuals from the best population into the worst population, replacing a corresponding number of poor individuals.

[0048] Adaptive population size refers to dynamically adjusting the size of the population at a scaling ratio of 0.5-1.5. When the average fitness of the parent individuals in the population is better than that of the parent individuals in the previous generation, it means that the parent generation in the previous generation has achieved a good search effect, and the population size can be reduced. On the contrary, the iteration has achieved a poor search effect, and the population size should be increased to improve the search ability. Specifically, according to the fitness value and the sigmoid function, follow the following steps to determine:

[0049] (a) Calculate the fitness change rate

[0050] (b) Calculate the scaling ratio β = aigmoid(5×α)+0.5

[0051] (c) Determine the updated population size G N =G N-1 ×β

[0052] Among them, f fit_(N-1) represents the fitness value of the N-1 generation, f fit_(N) represents the fitness value of the Nth generation, α represents the fitness change rate, β represents the scaling ratio, G N-1 represents the population size of the N-1 generation, G N represents the population size of generation N.

[0053] The adaptive crossover refers to ensuring the diversity of individuals through the non-dominated recombination method. In the adaptive non-dominated recombination operation, each iteration performs non-dominated recombination on a part of the individuals in the population to ensure population diversity and avoid the complete destruction of excellent gene fragments in the population. The proportion of non-dominated recombination can be determined by measuring the similarity of the population fitness variance. The greater the similarity, the more individuals in the population undergo non-dominated recombination. The calculation formula is:

[0054] N r =-ln(D(fitness))·T·N parent

[0055] Where N r represents the number of individuals undergoing non-dominated recombination, D(fitness) represents the variance of the fitness of the parental individuals, T represents a constant 1 / 20, and N parent Represents the number of parent individuals in the population.

[0056] The adaptive mutation refers to the use of a smaller mutation probability in the early stage of optimization and a larger mutation probability in the later stage. This mutation strategy can ensure that the algorithm converges quickly to the optimal solution in the early stage and is conducive to jumping out of local convergence in the later stage. Its adaptive conversion function also uses the sigmoid function, and the specific calculation is as follows.

[0057]

[0058] Among them, M is a constant of 0.1, which means that the probability of mutation generally does not exceed 0.1.

[0059] The optimal arrangement plan is to select the individual with the highest fitness as the optimal solution when the specified number of iterations is completed and the set conditional fitness value converges, and then obtain the optimal plan based on the parameters of the optimal solution.

[0060] Example

[0061] The venue grid shell roof measurement point optimization method of this application was applied to the space grid roof structure of the Xi'an International Football Center. It is planned to deploy 30, 50, and 100 sensors on the roof structure, and use the multi-population non-dominated recombination genetic algorithm (MP-RGA) in this application, as well as the quantum genetic algorithm (QGA) and the particle swarm optimization algorithm (PSO) to optimize the measurement points, and obtain the measurement point optimization solutions respectively. During the comparison process, the initial population size of the multi-population non-dominated recombination genetic algorithm (MP-RGA) was taken as 100, with a total of 3 populations, the population exchange ratio was set to 0.2, and the population exchange frequency was set to once every 10 generations; the quantum genetic algorithm (QGA) and the particle swarm optimization algorithm (PSO) are single populations, so the population size is taken as 300. Other parameters are set as follows: The QGA method uses two-weight quantum bit coding and quantum rotating gate function to realize the evolution of the population, and the length of each gene code is 10; the PSO method updates the population by considering the optimal solution of each generation of the population and the historical optimal solution of the individual. The learning factors C1 and C2 in the update are both 2, and the speed limit is 10 to avoid skipping the optimal solution due to excessive movement steps. For individuals that do not meet the constraints, the QGA method and the PSO method use regeneration to replace them. In order to eliminate the errors in the test process, the above three algorithms are run independently 3 times, and the comparison results are shown in Figure 2. The results show that when 30, 50, and 100 sensors are deployed on the roof structure, the MP-RGA converges the fastest and converges after 200 iterations. The QGA converges the second fastest and converges after about 400 iterations. The PSO converges the slowest and converges after 1000 iterations. In terms of fitness value (f1), the fitness value of MP-RGA is between 0.95 and 1.0, QGA converges between 0.9 and 0.96, and PSO converges between 0.7 and 0.85. After the same number of iterations, the proposed multi-population non-dominated recombination genetic algorithm (MP-RGA) obtains a larger fitness value (f1) than the quantum genetic algorithm (QGA) and the particle swarm optimization algorithm (PSO), indicating that the invented algorithm is more suitable for the measurement point arrangement of the stadium roof structure than the previous method, and the best measurement point optimization scheme is obtained.

[0062] It should be noted that, in the present invention, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprises" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or includes elements inherent to such process, method, article or device.

[0063] The above embodiments are merely examples of the present invention and do not limit the protection scope of the present invention. All designs that are the same or similar to the present invention fall within the protection scope of the present invention.

Claims

1. A method for optimizing measurement points of a venue lattice shell roof, characterized in that: The following steps are involved: Obtaining all measurement point positions of the venue lattice shell roof as an initialization large population, dividing the large population into multiple populations and performing genetic iteration operations simultaneously; The MAC matrix is ​​constructed based on the modal vector estimated by the measuring point position, and the non-diagonal elements in the MAC matrix that are minimized are selected as the fitness function, and the fitness function is used as the optimization target; The number of iterations, crossover probability and mutation probability are preset. When the number of iterations meets the set value and the fitness function converges, the individual with the highest fitness is selected as the optimal solution. According to the parameters of the optimal solution, the optimal solution for the measurement points of the venue lattice shell roof is determined. A non-dominated recombination genetic algorithm is used to perform genetic iteration operations on multiple populations, and the non-dominated recombination genetic algorithm includes the following steps: By generating a reverse population corresponding to the original population, the individuals with better fitness than the original population are retained to obtain a new population. Determine the population exchange ratio based on fitness , Population Communication Frequency , exchange individuals between new populations to obtain populations after communication; Identify the best and worst populations in each population after communication, introduce elite individuals from the best population into the worst population, replace the poor individuals therein, and obtain an improved population; The adaptive population update mechanism is used to update the sizes of multiple populations including the best population, the worst population and the improved population. The individuals of the population are crossed according to the similarity measured by the variance, and the updated population is adaptively mutated by the sigmoid function. Repeat the iterative process to determine whether the number of iterations reaches the set condition.

2. The venue lattice shell roof measurement point optimization method according to claim 1, characterized in that: The fitness function is expressed as: ； in, Indicates the first i Line j Elements of a column; represents the maximum value in the non-diagonal cells of the MAC matrix; X Indicates the measurement point location layout plan.

3. The venue lattice shell roof measurement point optimization method according to claim 1, characterized in that: The construction of the MAC matrix based on the measurement point position includes the following steps: Obtain structural dynamic response data at the measuring point and estimate modal parameters, including modal frequency, damping ratio and modal vector, through signal processing; Construct an M×M MAC matrix, where M is the number of identified modes; For each off-diagonal element in the MAC matrix ( i , j ), calculate the i and j The dot product of the modal vectors ; Divide the dot product result by i and j The product of the norms of the modal vectors is used to obtain the normalized MAC value; Fill the calculated MAC value into the corresponding position of the MAC matrix ( i , j ) to fill and obtain the MAC matrix.

4. The venue lattice shell roof measurement point optimization method according to claim 1, characterized in that: The number of iterations of the non-dominated recombination genetic algorithm is initially set to an integer greater than 1000, and is expanded and adjusted according to the degree of fitness convergence until the fitness converges; The crossover probability of the population in the non-dominated recombination genetic algorithm is determined according to the similarity measured by the variance of the fitness of the population; The mutation probability of the population in the non-dominated recombination genetic algorithm is adaptively adjusted through a sigmoid function adaptive conversion function.

5. The venue lattice shell roof measurement point optimization method according to claim 1, characterized in that: The population size in the non-dominated recombination algorithm is determined by the fitness value and the sigmoid function, and includes the following steps: Get the fitness change rate, the calculation formula is as follows: ； Get the zoom ratio, the calculation formula is as follows: ； Determine the updated population size, the calculation formula is as follows: ； in, represents the fitness value of the N-1 generation, represents the fitness value of N generations, represents the fitness change rate, Indicates the zoom ratio, represents the population size of the N-1 generation, represents the population size of generation N.

6. The venue lattice shell roof measurement point optimization method according to claim 1, characterized in that: The crossover probability of the population in the non-dominated recombination genetic algorithm is determined according to the similarity measured by the variance of the population fitness, and the calculation formula is: ； in, Nr represents the number of individuals undergoing non-dominated recombination, D ( fitness ) represents the variance of the parental individual fitness, T represents the constant 1 / 20, N parent Represents the number of parent individuals in the population.

7. The venue lattice shell roof measurement point optimization method according to claim 1, characterized in that: The mutation probability of the population in the non-dominated recombination genetic algorithm is adaptively adjusted by the sigmoid function adaptive conversion function, and the calculation formula is as follows: ； in, M is a constant of 0.1, indicating that the probability of mutation does not exceed 0.

1.

8. The venue lattice shell roof measurement point optimization method according to claim 1, characterized in that: The identification of the best population and the worst population in each population after the communication includes: The population after communication is The optimal and the worst populations are obtained.

9. A venue lattice shell roof measurement point optimization system, characterized in that: include: A measuring point data acquisition module is used to obtain the positions of all measuring points of the venue lattice shell roof as an initialization large population, and divide the large population into multiple populations to perform genetic iteration operations simultaneously; The measurement point optimization analysis module builds a measurement point layout optimization model for the venue grid shell roof based on the non-dominated recombination genetic algorithm, and takes the fitness of the population as the optimization target; Based on the estimated modal vector of the measuring point position, the MAC matrix is ​​constructed, and the non-diagonal elements in the minimized MAC matrix are selected as the fitness function, which is expressed as: ； in, Indicates the first i Line j Elements of a column; represents the maximum value in the non-diagonal cells of the MAC matrix; X Indicates the measurement point location arrangement plan; sets the number of iterations, crossover probability and mutation probability of the non-dominated recombination genetic algorithm; The optimization scheme acquisition module selects the individual with the highest fitness as the optimal solution when the number of iterations meets the set value and the fitness value converges, and further obtains the optimal solution for the measurement points of the venue grid shell roof; A non-dominated recombination genetic algorithm is used to perform genetic iteration operations on multiple populations, and the non-dominated recombination genetic algorithm includes the following steps: By generating a reverse population corresponding to the original population, the individuals with better fitness than the original population are retained to obtain a new population. Determine the population exchange ratio based on fitness , Population Communication Frequency , exchange individuals between new populations to obtain populations after communication; Identify the best and worst populations in each population after communication, introduce elite individuals from the best population into the worst population, replace the poor individuals therein, and obtain an improved population; The adaptive population update mechanism is used to update the sizes of multiple populations including the best population, the worst population and the improved population. The individuals of the population are crossed according to the similarity measured by the variance, and the updated population is adaptively mutated by the sigmoid function. Repeat the iterative process to determine whether the number of iterations reaches the set condition.

Images