An adaptive model updating method for probabilistic analysis of complex engineering structures
By setting the number of iterations and exponential constants in complex engineering structures, combining adaptive likelihood functions, using random sampling and proposed distribution to generate intermediate model parameters, the problems of insufficient adaptability of algorithms and long calculation time in the existing technology are solved, and the problem of improved The efficiency and adaptability of model correction are provided to provide better guidance for health detection of complex engineering structures.
Patent Information
- Application Number
- CN202411216622.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-02
- Publication Date
- 2024-11-19
- Estimated Expiration
- 2044-09-02
AI Technical Summary
The existing technology has problems such as insufficient algorithm adaptability, inappropriate calculation of likelihood function and long calculation time in parameter in the parameter inversion and health detection of complex engineering structures, resulting in low efficiency of model parameter correction and difficulty in adapting to different models and iteration times. Long question.
By setting the number of algorithm iterations, difference constants and exponential constants, combined with the adaptive likelihood function form, random sampling and proposed distribution are used to generate intermediate model parameters, and the algorithm coefficients and likelihood functions are adaptively adjusted to improve the adaptability of model correction. and computational efficiency.
It realizes the rapid search for the most possible distribution of structural model parameters in complex engineering structures, reduces the calculation time, improves the efficiency and adaptability of model correction, and provides better health detection guidance.
Smart Images

Figure CN118734658B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of model structure parameter inversion, and in particular to an adaptive model updating method for probability analysis of complex engineering structures. Background Art
[0002] Parameter inversion based on test data of complex engineering structures to perform health detection is an economical and reliable method, which has great guiding significance for the safety assessment of complex structures. Complex engineering structures refer to large-scale and complex engineering structures, such as high-rise and super-high-rise structures, long-span bridges and nuclear power plants. Since complex engineering structures are large and have many components, even if scaled models are used for testing, it will take a lot of time and money. In addition, when studying the safety of complex engineering structures, it is generally necessary to load the model structure to destruction. The established model can only obtain a set of structural destruction data. If repeated tests are to be carried out, it will require huge manpower, material resources and financial resources. This makes it very difficult to completely infer the parameter values of complex engineering structures through experiments, and it is difficult to promote the research on health detection of complex engineering structures. In addition, there are many parameters that need to be determined in complex engineering structures. Some material parameters can be determined through material properties tests, but more parameters cannot be determined through tests or specifications. Due to the uncertainty caused by the parameters themselves, measurement errors and environmental conditions, it is more difficult to determine the parameters of complex engineering structures, and it is also necessary to consider the uncertainty of parameters when analyzing complex engineering structures. Finally, if the finite element method and parameter inversion analysis method are used to establish a complex engineering structure model for parameter inversion analysis, the calculation time of the model parameter inversion and the adaptability of the method to the complex engineering model are issues that need to be considered. Therefore, how to establish a small amount of data-driven and highly adaptive model update method for complex engineering structures is an important step in solving the parameter inversion and health detection of complex engineering structures.
[0003] The Bayesian model parameter update method is one of the most popular model parameter update methods. It is a model parameter correction method based on Bayesian theory and Markov Monte Carlo (MCMC) sampling algorithm. Since the 21st century, Bayesian model parameter update algorithm and finite element model have been widely used in parameter inversion and health detection of models. The most popular Bayesian model parameter update method is based on Metropolis-Hastings (MH) sampling algorithm. Due to the limitations of the MH sampling algorithm, some new improved algorithms have also been proposed, such as the transition MCMC (TMCMC) algorithm, the delayed rejection-MCMC (DR-MCMC) algorithm, and the adaptive Metropolis-MCMC (AM-MCMC) algorithm. However, these algorithms all have certain limitations. First, they have not proposed a method for calculating the likelihood function suitable for different models, which means that each time the algorithm is applied to a new model, it is necessary to find a suitable form of the likelihood function again; second, they cannot adaptively find reasonable values for some coefficients in the algorithm, which means that it takes a lot of time to debug these coefficients in the algorithm; finally, they do not consider the problem of long finite element calculation time. In the algorithm, the solution of some algorithm parameters requires multiple iterations. These methods may be suitable for models with short iteration times, but they are not applicable if the iteration time is long. Summary of the invention
[0004] Purpose of the invention: The purpose of the present invention is to provide an adaptive model updating method for probabilistic analysis of complex engineering structures.
[0005] Technical solution: The present invention comprises the following steps:
[0006] (1) Assume that the total number of algorithm iterations is , the number of iterations each time , difference constant , exponential constant , likelihood function , the important parameter matrix of the model ,in , Indicates Prior distribution of important parameters and model parameters , the number of measuring points And the test point response matrix ,in , Indicates The response of the measuring point is ;
[0007] (2) Order , from the prior distribution of the model parameters Random sampling Group prior parameter matrix ,in , Indicates A priori parameter matrix is formed and the likelihood of the corresponding model prior parameters is calculated;
[0008] (3) Calculate the variance constant , and the maximum difference , and The likelihood weight coefficient of each group of prior parameter values under the total iterations and the mean of the weight coefficients ;
[0009] (4) Calculate the variance matrix of the proposed distribution ;
[0010] (5) Generate an intermediate model parameter matrix based on the likelihood weight coefficient of the prior parameter value and the prior parameter ,in , Indicates The intermediate model parameter matrix of the group, the variance matrix based on the intermediate parameter matrix and the proposed distribution Random sampling Group Candidate Value , , Indicates Group candidate value matrix, if If it exceeds the range of the prior distribution, Resample by a factor of 2 to get Finally, the likelihood values of the intermediate parameters and candidate values are calculated respectively. and ;
[0011] (6) Generate a random value from [0,1] ,if , then let , otherwise let ;
[0012] (7) Update , , and ;
[0013] (8) Order ,if ,make , repeat steps (4)-(7) to start a new round of iterations and calculate The last iteration ends; if , then return to step (4).
[0014] Furthermore, the number of iterations in each step (1) is Take 100, the difference constant Pick , exponential constant Pick ;Important parameter matrix The prior distribution of model parameters is determined by screening the parameters that have a greater impact on the model through sensitivity analysis; Obtained through industry standards, journal papers and test data; the number of sensors and the response of the measurement points depends on the test data and the number of measurement points that the user is concerned about; the form of the likelihood function is expressed as:
[0015] ,
[0016] in, The measurement point response matrix representing the simulated data; Represents the test point response matrix; represents the difference constant, which is calculated based on the likelihood of prior sampling; Represents the transpose symbol.
[0017] Furthermore, the variance constant in step (3) is expressed as:
[0018] ,
[0019] ,
[0020] ,
[0021] in, represents variance; Represents the value of the likelihood function for a specific condition; is an exponential constant; according to The quality of the group parameter values is used to find the appropriate , if the parameter values of these groups are poor, then will be smaller, Indicates the worst parameter value The value of .
[0022] Furthermore, the maximum difference value in step (3) is expressed as:
[0023] ,
[0024] in, ; Indicates The measured value of each measuring point; Indicates The simulated value of each measuring point; Indicates the last iteration , if it is the first iteration, this item is not considered.
[0025] Furthermore, the likelihood weight coefficient of each set of prior parameter values in step (3) is It is expressed as:
[0026] .
[0027] Furthermore, the mean of the prior parameter likelihood weight coefficients in step (3) is It is expressed as:
[0028] .
[0029] Furthermore, the variance matrix of the proposed distribution in step (4) is It is expressed as:
[0030] ,
[0031] ,
[0032] ,
[0033] ,
[0034] ,
[0035] in, represents the scale parameter; express The mean matrix of group prior parameters; Represents an incremental constant, which is 1 for the first iteration and increases by 1 for each additional iteration; represents the acceptance rate of candidate values; represents the target acceptance rate of candidate values; Indicates the number of parameters involved in model parameter correction.
[0036] Furthermore, in step (7), the maximum difference is updated after each set of iterations is completed. , likelihood function , likelihood weight and the mean of the likelihood weights .
[0037] Furthermore, in step (8), ,make , and then repeat the iteration once to obtain the posterior distribution of the model parameters, and the weight value of the likelihood is calculated by the following formula:
[0038] ,
[0039] Among them, assuming that the final hour , the iteration ends.
[0040] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages: the present invention sets the number of iterations, the difference constant and the exponential constant, and combines the proposed likelihood function to adaptively find the most likely distribution of the structural model parameters, thereby further providing guidance for the health detection of complex engineering structures; it overcomes the limitations of previous studies such as the need to obtain suitable algorithm coefficients through multiple iterations and the lack of a likelihood function form with good adaptability. By comprehensively utilizing the data in the iteration to adaptively find the algorithm coefficients and likelihood function form suitable for different complex engineering models, the model correction has better adaptability, less calculation time, and improved algorithm efficiency. BRIEF DESCRIPTION OF THE DRAWINGS
[0041] Figure 1 is a flow chart of the present invention;
[0042] Figure 2 It is the finite element model of the prestressed concrete containment;
[0043] Figure 3 This is the change diagram of the minimum value of the overall error of 30 measuring points during the iteration process;
[0044] Figure 4 It is the frequency histogram and fitting result diagram of each parameter after model correction;
[0045] Figure 5 This is the error distribution diagram of 30 measurement points during the iteration process. DETAILED DESCRIPTION
[0046] The technical solution of the present invention is further described below in conjunction with the accompanying drawings.
[0047] like Figure 1 As shown, the present invention comprises the following steps:
[0048] (1) Assume that the total number of algorithm iterations is , the number of iterations each time , difference constant , exponential constant , likelihood function , the important parameter matrix of the model ,in , Indicates Prior distribution of important parameters and model parameters , the number of measuring points And the test point response matrix ,in , Indicates The response of the measuring point is ;
[0049] The user does not need to determine the number of iterations in step (1). Take 100, the difference constant Pick , exponential constant Pick ;Important parameter matrix The prior distribution of model parameters is determined by screening the parameters that have a greater impact on the model through sensitivity analysis; Obtained through industry standards, journal papers and test data; the number of sensors and the response of the measurement points depends on the test data and the number of measurement points that the user is concerned about; the form of the likelihood function is expressed as:
[0050]
[0051] in, The measurement point response matrix representing the simulated data; Represents the test point response matrix; represents the difference constant, which is calculated based on the likelihood of prior sampling; Represents the transpose symbol.
[0052] (2) Order , from the prior distribution of the model parameters Random sampling Group prior parameter matrix ,in , Indicates A priori parameter matrix is formed and the likelihood of the corresponding model prior parameters is calculated;
[0053] (3) Calculate the variance constant , and the maximum difference , and The likelihood weight coefficient of each group of prior parameter values under the total iterations and the mean of the weight coefficients ;
[0054] The variance constant is expressed as:
[0055]
[0056]
[0057]
[0058] in, represents variance; Represents the value of the likelihood function for a specific condition; is an exponential constant; according to The quality of the group parameter values is used to find the appropriate , if the parameter values of these groups are poor, then will be smaller, Indicates the worst parameter value The value of .
[0059] The maximum difference is expressed as:
[0060]
[0061] in, ; Indicates The measured value of each measuring point; Indicates The simulated value of each measuring point; Indicates the last iteration , if it is the first iteration, this item is not considered.
[0062] The likelihood weight coefficient of each group of prior parameter values It is expressed as:
[0063]
[0064] The mean of the prior parameter likelihood weight coefficients It is expressed as:
[0065]
[0066] (4) Calculate the variance matrix of the proposed distribution ;
[0067] Variance matrix of the proposed distribution It is expressed as:
[0068]
[0069]
[0070]
[0071]
[0072]
[0073] in, represents the scale parameter; express The mean matrix of group prior parameters; Represents an incremental constant, which is 1 for the first iteration and increases by 1 for each additional iteration; represents the acceptance rate of candidate values; represents the target acceptance rate of candidate values; Indicates the number of parameters involved in model parameter correction.
[0074] (5) Generate an intermediate model parameter matrix based on the likelihood weight coefficient of the prior parameter value and the prior parameter ,in , Indicates The intermediate model parameter matrix of the group, the variance matrix based on the intermediate parameter matrix and the proposed distribution Random sampling Group Candidate Value , , Indicates Group candidate value matrix, if If it exceeds the range of the prior distribution, Resample by a factor of 2 to get Finally, the likelihood values of the intermediate parameters and candidate values are calculated respectively. and ; When generating intermediate model parameters, when the likelihood weight of the prior model parameters accounts for a larger proportion, the generated intermediate model parameter value will be larger. The number of intermediate model parameter values that are the same as the prior model parameter values is Candidate values are generated from a normal distribution with the intermediate model parameters as mean and the variance matrix of the proposed distribution as variance.
[0075] (6) Generate a random value from [0,1] ,if , then let , otherwise let ;
[0076] (7) Update , , and ; Update the maximum difference at the end of each set of iterations , likelihood function , likelihood weight and the mean of the likelihood weights .
[0077] (8) Order ,if ,make , repeat steps (4)-(7) to start a new round of iterations and calculate The last iteration ends; if , then return to step (4);
[0078] when ,make , and then repeat the iteration once to obtain the posterior distribution of the model parameters, and the weight value of the likelihood is calculated by the following formula:
[0079]
[0080] Among them, assuming that the final hour , the iteration ends.
[0081] Take a prestressed concrete containment shell as an example. Figure 2 The finite element model of the prestressed concrete containment is shown in Table 1. In this example analysis, there are 5 important parameters of the prestressed concrete containment model, and their distribution means and variances are shown in Table 1. Parameter distribution of prestressed concrete containment.
[0082]
[0083] In addition, let the number of iterations be , the difference constant , exponential constant The internal pressure of the containment shell increases from 0 MPa to 1.42 MPa, and 15 points where the containment shell reaches functional damage (1.29 MPa) and structural damage (1.42 MPa) are selected to form the measurement point response. , the location distribution of 15 measuring points and the numbers of 30 measuring points are shown in Table 2 The location distribution of 15 measuring points and the numbers of 30 measuring points are shown in Table 2 .
[0084]
[0085] After defining the above values, the algorithm starts to run automatically. The algorithm ends after 7 iterations, and the total error of the updated optimal parameters is 3.858 mm, which is 40.26% lower than the total error of 6.458 mm before iteration. Figure 3 The figure shows the change of the minimum value of the overall error of 30 measurement points found in each iteration during the parameter iteration process. Figure 4 The figure shows the frequency histogram and fitting results of each parameter after model correction. Figure 4 The 95% confidence interval of each parameter distribution obtained by fitting in the above equation is obtained. 100 sets of parameter values are randomly selected to obtain the error distribution diagram of 30 measuring points. Figure 5 shown.
[0086] In summary, the present invention is based on the prior distribution of parameters. After the algorithm is iterated, the error of the optimal parameters found is reduced by about 40% compared with before the update. The mean of the error of the randomly generated measurement points of the posterior distribution of the parameters obtained after the update is also close to 0, which proves the efficiency of the algorithm in updating the parameters of complex structural models and can provide guidance for the subsequent health detection of complex engineering structures.
Claims
1. An adaptive model updating method for probabilistic analysis of complex engineering structures, characterized in that: The steps include: (1) Assume that the total number of algorithm iterations is , the number of iterations each time , difference constant , exponential constant , likelihood function , the important parameter matrix of the model ,in , Indicates Prior distribution of important parameters and model parameters , the number of measuring points And the test point response matrix ,in , Indicates The response of the measuring point is ; (2) Order , from the prior distribution of the model parameters Random sampling Group prior parameter matrix ,in , Indicates A priori parameter matrix is formed and the likelihood of the corresponding model prior parameters is calculated; (3) Calculate the variance constant , and the maximum difference , and The likelihood weight coefficient of each group of prior parameter values under the total iterations and the mean of the weight coefficients ; (4) Calculate the variance matrix of the proposed distribution ; (5) Generate an intermediate model parameter matrix based on the likelihood weight coefficient of the prior parameter value and the prior parameter ,in , Indicates The intermediate model parameter matrix of the group, the variance matrix based on the intermediate parameter matrix and the proposed distribution Random sampling Group Candidate Value , , Indicates Group candidate value matrix, if If it exceeds the range of the prior distribution, Resample by a factor of 2 to get Finally, the likelihood values of the intermediate parameters and candidate values are calculated respectively. and ; (6) Generate a random value from [0,1] ,if , then let , otherwise let ; (7) Update , , and ; (8) Order ,if ,make , repeat steps (4)-(7) to start a new round of iterations and calculate The last iteration ends; if , then return to step (4), The number of iterations in each step (1) Take 100, the difference constant Pick , exponential constant Pick ;Important parameter matrix The parameters with the greatest impact on the model are screened through sensitivity analysis to determine the prior distribution of model parameters. Obtained through industry standards, journal papers and test data; the number of sensors and the response of the measurement points depends on the test data and the number of measurement points that the user is concerned about; the form of the likelihood function is expressed as: , in, The measurement point response matrix representing the simulated data; Represents the test point response matrix; represents the difference constant, which is calculated based on the likelihood of prior sampling; represents the transpose symbol, The variance constant in step (3) is expressed as: , , , in, represents variance; represents the value of the likelihood function for a specific condition; is an exponential constant; according to The good or bad group parameter values are used to find , Indicates the worst parameter value The value of .
2. The adaptive model updating method for probabilistic analysis of complex engineering structures according to claim 1, characterized in that: The maximum value of the difference in step (3) is expressed as: , in, ; Indicates The measured value of each measuring point; Indicates The simulated value of each measuring point; Indicates the last iteration , if it is the first iteration, this item is not considered.
3. The adaptive model updating method for probabilistic analysis of complex engineering structures according to claim 1, characterized in that: The likelihood weight coefficient of each set of prior parameter values in step (3) is It is expressed as: 。 4. The adaptive model updating method for probabilistic analysis of complex engineering structures according to claim 1, characterized in that: The mean of the prior parameter likelihood weight coefficients in step (3) It is expressed as: 。 5. The adaptive model updating method for probabilistic analysis of complex engineering structures according to claim 1, characterized in that: The variance matrix of the proposed distribution in step (4) is It is expressed as: , , , , , in, represents the scale parameter; express The mean matrix of group prior parameters; Represents an incremental constant, which is 1 for the first iteration and increases by 1 for each additional iteration; represents the acceptance rate of candidate values; represents the target acceptance rate of candidate values; Indicates the number of parameters involved in model parameter correction.
6. The adaptive model updating method for probabilistic analysis of complex engineering structures according to claim 1, characterized in that: In step (7), the maximum difference is updated after each set of iterations is completed. , likelihood function , likelihood weight and the mean of the likelihood weights .
7. The adaptive model updating method for probabilistic analysis of complex engineering structures according to claim 1, characterized in that: In step (8), ,make , and then repeat the iteration once to obtain the posterior distribution of the model parameters, and the weight value of the likelihood is calculated by the following formula: , Among them, assuming that the final hour , the iteration ends.
Citation Information
Patent Citations
Transient electromagnetic two-dimensional Bayesian inversion method adopting adaptive Thiessen polygon parameterization
CN114896564A
Markov chain Monte Carlo parameter estimation method based on Kalman proposal distribution
CN115081326A
Random structure nonlinear model correction method based on TMCMC algorithm
CN116050208A
Super high-rise building construction simulation method and system combined with Bayesian correction
CN117521202A
