Rapid random model correction method for step-by-step ocean platform structure

Abstract

The invention relates to the technical field of ocean engineering, and provides a rapid random model correction method for a step-by-step ocean platform structure. According to the design parameters of the ocean platform, establishing an ocean platform initial finite element model; selecting to-be-corrected parameters from the design parameters, and determining the maximum value and the minimum value of each to-be-corrected parameter; performing # imgabs0 # sampling on the maximum value and the minimum value of each to-be-corrected parameter to obtain a corrected parameter matrix; substituting the correction parameter matrix into the initial finite element model to obtain an updated finite element model, obtaining a structure frequency vector corresponding to a correction parameter vector, and establishing a Kriging model; calculating an initial finite element model frequency based on the corrected parameter vector and a Kriging model; constructing a distance objective function and a distribution objective function, and correcting the mean value and the standard deviation of the correction parameters twice; and obtaining a corrected ocean platform finite element model based on the finite element model frequency corrected for the second time.

Description

Technical Field

[0001] The invention relates to the technical field of marine engineering, and in particular to a step-by-step marine platform structure rapid random model correction method. Background Art

[0002] Offshore platforms are key infrastructure for the development of marine resources (such as oil and natural gas extraction). They are exposed to complex and harsh marine environments for a long time. They are mainly composed of multiple components such as pile legs, decks and supporting structures. Dynamic response assessment and health monitoring technology of offshore platform structures under complex marine environmental loads are current research hotspots and difficulties. Obtaining a finite element model that can replace the actual offshore platform structure for monitoring and evaluation is the key core.

[0003] Theoretical models such as the initially established offshore platform finite element model often deviate from the actual structure. The factors affecting the deviation mainly come from modeling assumptions (such as material property assumptions, boundary condition assumptions, etc.), manufacturing errors, installation errors, and other aspects. For example, when modeling, it is assumed that the elastic modulus of the material is a fixed value, but the elastic modulus of the actual material may vary due to factors such as production batches and environmental corrosion. Model correction can improve the accuracy of the model, thereby more accurately predicting the response of the platform under various working conditions.

[0004] The traditional model correction method considering the uncertainty of marine structure parameters requires a large number of iterations and multiple calls to the finite element model for optimization and solution. Due to the complex structure of marine structural units and the large number of degrees of freedom, the model correction process often leads to problems such as large amount of calculation and low efficiency. Simplifying the finite element model or finding a proxy model has become an important measure to improve the efficiency of model correction operations. Random model correction based on proxy models was proposed and widely used based on this idea. Its main idea is to use an explicit proxy model to fit the complex implicit relationship between the structural frequency response and the selected parameters to be corrected, so as to achieve the purpose of replacing the finite element model for numerical calculation and simplifying the iterative process. Compared with other proxy models such as polynomial regression and neural networks, the Kriging model has excellent results in processing nonlinear data. For the optimization solution link in the random model correction, the selection of the objective function and correction strategy is crucial to the correction accuracy. The random model correction method based on the traditional or active learning Kriging model uses the Euclidean norm and KL divergence between the measured frequency and the finite element model frequency to construct a multi-objective function. The correction results show that there is a large difference in the variance between the corrected finite element model frequency and the measured frequency. The interference of different objective functions in selecting the optimal solution of the distribution index on the Pareto front is the main reason for this problem.

[0005] Before the random model correction of the structure, the mean and variance of its correction parameters are unknown. The direct structural random model correction method inputs the mean and variance of the correction parameters into the optimization algorithm as unknowns at the same time, and directly performs iterative solutions. Studies have shown that for indicators that measure the difference between the two probability distributions of the measured frequency and the finite element model frequency, such as the Bhattacharyya distance and the KL divergence, when the measured frequency and the finite element model frequency do not overlap, their indicators are meaningless. Therefore, the direct structural random model correction method combines the distribution difference index with the objective function such as the Euclidean norm between the two frequencies, constructs a multi-objective function for optimization and solution, and obtains the optimal solution of the mean and standard deviation of the structural correction parameters at one time. However, for multi-objective optimization, the selection of the optimal solution is constrained by all objective functions, thus forming a Pareto frontier. The Pareto frontier refers to the state where it is impossible to improve one objective without compromising another objective under given constraints. In multi-objective optimization problems, the solution of the Pareto frontier is very important. It represents a trade-off in which no objective can be improved without affecting other objectives. Therefore, when the KL divergence is used to construct the objective function, the selection of the optimal solution is affected by the Euclidean norm, and the optimization effect of the KL divergence cannot be fully exerted, which will inevitably interfere with the correction of the variance between the frequency of the finite element model and the measured frequency.

[0006] Based on the above reasons, it is necessary to study a random correction method for the offshore platform structure model that can reduce the variance difference between the actual frequency and the finite element frequency after correction. Summary of the invention

[0007] The purpose of the present invention is to solve the above technical problems and provide a step-by-step fast random model correction method for an offshore platform structure.

[0008] In order to achieve the above object, in some embodiments of the present invention, the following technical solutions are provided: A step-by-step fast random model correction method for an offshore platform structure, characterized by comprising the following steps: According to the design parameters of the offshore platform, the initial finite element model of the offshore platform is established; Among the design parameters of the offshore platform, parameters to be corrected, and determining the maximum and minimum values ​​of each parameter to be corrected; The maximum and minimum values ​​of each parameter to be corrected are calculated. Sampling times, get the modified parameter matrix ,in For the The correction parameter vectors obtained by sampling times, each correction parameter vector contains correction parameters; among them, is an integer greater than 1; Substituting the correction parameter matrix into the initial finite element model, updating the structural parameters or dimensional parameters corresponding to the correction parameters in the correction parameter matrix to obtain an updated finite element model, performing modal analysis or finite element analysis on the updated finite element model to obtain a structural frequency vector corresponding to the correction parameter vector; establishing a Kriging model based on the correction parameter vector and its corresponding structural frequency vector; Based on the modified parameters and Kriging model, the initial finite element model frequency is calculated; Based on the frequency and standard deviation of the initial finite element model and the measured frequency and standard deviation of the offshore platform, a distance objective function is constructed, and the mean and standard deviation of the correction parameters are first optimized and corrected based on the optimization target; The mean and standard deviation after the first optimization correction are brought into the Kriging model to calculate the first corrected finite element model frequency. Based on the measured frequency of the offshore platform and the corrected finite element model frequency, the KL divergence is used to construct the distribution objective function. The mean and standard deviation of the correction parameters are optimized for the second time based on the optimization objective. The mean and standard deviation of the correction parameters after the second optimization correction are substituted into the Kriging model to obtain the finite element model frequency after the second correction. The error between the finite element model frequency after the second correction and the measured frequency of the offshore platform is calculated. If the error meets the accuracy requirements, the corrected offshore platform finite element model is obtained based on the finite element model frequency after the second correction.

[0009] In some embodiments of the present invention, the step of selecting the parameter to be corrected includes: A sensitivity analysis is performed on the design parameters, and parameters with high sensitivity are selected as parameters to be corrected.

[0010] In some embodiments of the present invention, the design parameters of the offshore platform include structural parameters and material parameters.

[0011] In some embodiments of the present invention, the step of establishing the Kriging model includes: Each set of modified parameter vectors and corresponding structural frequencies are recorded as a set of sample data. The first 80% of the sample data were selected as the training set, and the remaining sample data were selected as the test set to establish the Kriging model.

[0012] In some embodiments of the present invention, after establishing the Kriging model, the following steps are also included: The effectiveness evaluation index determination coefficient R2 and relative root mean square error RMSE were used to verify the established Kriging model and set the model verification accuracy; If the accuracy of the Kriging model meets the model verification accuracy requirements, the established Kriging model is used to continue calculating the initial finite element model frequency. If the accuracy of the Kriging model does not meet the model verification accuracy requirements, the Latin hypercube sampling steps are repeated, and the corrected parameter matrix obtained by Latin hypercube sampling is re-substituted into the initial finite element model to re-establish the Kriging model until the accuracy of the Kriging model meets the model verification accuracy requirements.

[0013] In some embodiments of the present invention, the step of calculating the initial finite element frequency includes: Calculate the initial mean and standard deviation of each correction parameter, and Monte Carlo sampling is performed to obtain the finite element model parameter samples. ,in, For the The initial mean and standard deviation vector of the modified parameters obtained by Monte Carlo sampling; Substitute the parameter samples into the Kriging model to obtain the initial finite element model frequency .

[0014] In some embodiments of the present invention, the distance objective function includes a mean objective function and a standard deviation objective function, and the step of constructing the distance objective function includes: ; ; in: represents the mean objective function, represents the standard deviation objective function, represents the frequency order, Indicates The mean of the frequencies of the finite element model, Indicates The mean frequency of the actual structure of the first-order offshore platform, Indicates The mean standard deviation of the frequency of the finite element model of order, Indicates The standard deviation of the frequency of the actual structure of the first-order offshore platform, .

[0015] In some embodiments of the present invention, the method for optimizing and correcting the mean and standard deviation of the correction parameter based on the optimization target includes: Taking the parameter mean and standard deviation of the correction parameters optimized and corrected as the solution target, the distance objective function is substituted into the multi-objective optimization algorithm, the mean and standard deviation of the correction parameters are solved by the optimization algorithm, and the correction target threshold is set. When the output result of the mean objective function is less than the correction target threshold and the output result of the standard deviation objective function is less than the correction target threshold, the correction is stopped, and the mean and standard deviation at this time are used as the mean and standard deviation of the correction parameters after optimization and correction.

[0016] In some embodiments of the present invention, the method for constructing a distribution objective function includes:

[0017] in, is the mean frequency of the finite element model, is the actual frequency mean of the marine platform, is the frequency covariance matrix of the finite element model , I platform actual frequency covariance matrix.

[0018] Compared with the prior art, the step-by-step fast random model correction method for offshore platform structures proposed in the present invention has the following beneficial effects: A distribution model correction method is proposed. To solve the problem of poor correction effect of structural frequency distribution difference, the model correction is divided into two steps. The first correction process is for distance correction of the distance objective function, and the second correction process is based on distribution difference correction of the distribution objective function. Compared with the direct random correction method, the step-by-step random correction method has fewer effective iterations than the direct method, has high efficiency in random model correction, and has more stable correction effect on the mean and standard deviation of structural parameters. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0020] Figure 1 A flow chart of the step-by-step rapid random model correction method for an offshore platform structure provided by the present invention; Figure 2 This is a comparison chart of the effect of correcting the mean value of parameters for the direct method and the distribution method; Figure 3 This is a comparison chart of the effects of correcting the standard deviation of parameters between the direct method and the distribution method; Figure 4Comparison chart of error correction effect of the mean and standard deviation of the first five frequencies of the direct method and the distributed method under different working conditions; Figure 5 The effect diagram of the confidence ellipse-step random correction method for the finite element model correction frequency and the measured frequency; Figure 6 The results of the confidence ellipse-direct random correction method for the finite element model correction frequency and the measured frequency are shown in Figure 2. Figure 7 This is a comparison chart of the average effective number of iterations for each working condition; Figure 8 The figure shows the comparison of the mean error correction effect of the parameters of the direct and step-by-step model correction methods; (a) shows the comparison of the mean error correction effect of the parameters of the direct model correction method; (b) shows the comparison of the mean error correction effect of the parameters of the step-by-step model correction method; Fig. 9 Comparison diagram of the error correction effect of the standard deviation of parameters of direct and step-by-step model correction methods; (a) Comparison diagram of the error correction effect of the standard deviation of parameters of direct model correction method; (b) Comparison diagram of the error correction effect of the standard deviation of parameters of step-by-step model correction method. DETAILED DESCRIPTION

[0021] In order to make the technical problems, technical solutions and beneficial effects to be solved by the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0022] In the embodiments of the present application, prefixes such as "first" and "second" are used only to distinguish different description objects, and have no limiting effect on the position, order, priority, quantity or content of the described objects. The use of prefixes such as ordinal numbers to distinguish description objects in the embodiments of the present application does not constitute a limitation on the described objects. For the statement of the described objects, please refer to the description in the context of the claims or embodiments, and no unnecessary limitation should be constituted due to the use of such prefixes. In addition, in the description of the present embodiment, unless otherwise specified, the meaning of "multiple" is two or more.

[0023] The technical solution in the embodiment of the present application will be described below in conjunction with the drawings in the embodiment of the present application. In the description of the embodiment of the present application, unless otherwise specified, " / " means or, for example, A / B can mean A or B; "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships, for example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone.

[0024] In the several embodiments provided in the embodiments of the present application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0025] Based on this, the present invention proposes a step-by-step fast random model correction method for offshore platform structures. Figure 1 , specifically including the following steps: S1: initial finite element model establishment step of the offshore platform.

[0026] According to the design parameters of the offshore platform, an initial finite element model of the offshore platform is established, and each design parameter corresponds to a structural parameter or a size parameter of the offshore platform.

[0027] Among them, the design parameters of the offshore platform include structural parameters and material parameters. The structural parameters reflect the structural design characteristics of the offshore platform, and the material parameters reflect what parameters are used in the main structural part of the offshore platform.

[0028] S2: Correction parameter selection step.

[0029] Among the design parameters of the offshore platform, select the parameters to be modified , determine the maximum and minimum values ​​of each correction parameter .in, is an integer greater than 1. The maximum value of the parameter and the minimum value of the parameter are determined based on the design scheme and design requirements of the offshore platform.

[0030] The purpose of the step of determining the correction parameters is to extract some parameters from the numerous design parameters of the offshore platform for correction in order to correct the model of the offshore platform. Due to the large number of design parameters of the offshore platform, the efficiency of using all parameters to correct the model is low, and there may be a risk of large correction errors. In order to correct the initial finite element model according to the design parameters, in the step of selecting the parameters to be corrected, the sensitivity index is used to select the parameters to be corrected.

[0031] In the embodiment of the present invention, a sensitivity analysis is performed on the design parameters, and parameters with high sensitivity are selected as parameters to be corrected.

[0032] The analysis of parameter sensitivity is used to evaluate the sensitivity of the model output to parameter changes. High sensitivity means that a slight modification of the correction parameter can result in a more obvious output response; low sensitivity means that the output response does not change much after the correction coefficient is modified, and effective model correction cannot be achieved. The analysis method of parameter sensitivity can be combined with the specific reality of the offshore platform structure, and the local analysis method and global analysis method in the existing technology can be used. The local method is suitable for quickly evaluating the impact of parameters near specific points, while the global method is more suitable for comprehensively analyzing the role of parameters in the entire space. In practical applications, multiple methods can be combined to obtain more comprehensive results. I will not go into details.

[0033] S3: Parameter sampling step.

[0034] The maximum and minimum values ​​of each parameter to be corrected are calculated. Sampling times, get the modified parameter matrix , each correction parameter vector Include correction parameters.

[0035] In a specific embodiment, the maximum and minimum values ​​of each correction parameter are used to perform Sub-Latin hypercube sampling, get the modified parameter matrix ,in For the The modified parameter vectors obtained by sub-Latin hypercube sampling, each of which contains correction parameters. Among them, is an integer greater than 1.

[0036] In the specific implementation process, other sampling methods can also be used, which will not be described in detail.

[0037] S4: Steps to establish the Kriging model.

[0038] The Kriging model assumes that the data is generated by a Gaussian process. Its core idea is to use the spatial correlation between known data points to predict the value of unknown points. As a powerful interpolation and prediction tool, the Kriging model is particularly suitable for spatial data analysis and engineering optimization.

[0039] The correction parameter matrix obtained by Latin hypercube sampling is brought into the initial finite element model, the structural parameters or size parameters corresponding to the correction parameters in the correction parameter matrix are updated to obtain an updated finite element model, modal analysis or finite element analysis is performed on the updated finite element model, and the correction parameter vector is input into the finite element model to obtain the output The correction parameter vector corresponds to Structural frequency vector ; Establish a Kriging model based on the modified parameter vector and its corresponding structural frequency vector.

[0040] In some embodiments of the present invention, the step of establishing the Kriging model includes: recording each set of correction parameter vectors and corresponding structural frequencies as a set of sample data, and obtaining The first 80% of the sample data are selected as the training set, and the remaining sample data are selected as the test set to establish the Kriging model. The steps of establishing the Kriging model belong to the prior art and will not be repeated here.

[0041] In some embodiments of the present invention, after establishing the Kriging model, the following steps are further included: The effectiveness evaluation index determination coefficient R2 and relative root mean square error RMSE were used to verify the established Kriging model and set the model verification accuracy; If the accuracy of the Kriging model meets the model verification accuracy requirements, the established Kriging model is used to continue calculating the initial finite element model frequency. If the accuracy of the Kriging model does not meet the model verification accuracy requirements, the Latin hypercube sampling steps are repeated, and the corrected parameter matrix obtained by Latin hypercube sampling is re-substituted into the initial finite element model to re-establish the Kriging model until the accuracy of the Kriging model meets the model verification accuracy requirements.

[0042] S5: Initial finite element model frequency calculation step.

[0043] Calculate the initial finite element model frequency based on the modified parameters and Kriging model ;in, is an integer greater than 1.

[0044] Specifically, the initial mean and standard deviation of each correction parameter are calculated. conduct Monte Carlo sampling is performed to obtain the finite element model parameter samples. ,in, Indicates The initial mean of the order correction parameter, Indicates The standard deviation of the order correction parameter, For the The initial mean and standard deviation vector of the modified parameters obtained by Monte Carlo sampling; Substitute the parameter samples into the Kriging model to obtain the initial finite element model frequency .

[0045] S6: step of constructing distance objective function.

[0046] Using the measured frequencies of offshore platforms The initial finite element model frequency Construct a distance objective function, where the measured frequency of the offshore platform is obtained based on the actual working environment of the offshore platform.

[0047] The distance objective function includes a mean objective function and a standard deviation objective function. In the embodiment of the present invention, the Euclidean norm is used as the distance objective function. The Euclidean norm is an indicator to measure the distance between two vectors, and is usually used to calculate the similarity or difference between vectors. The Euclidean norm of the structural frequency mean and standard deviation is: ; ; in: represents the mean objective function, represents the standard deviation objective function, represents the frequency order, Indicates The mean frequency of the finite element model, Indicates The mean frequency of the actual structure of the first-order offshore platform, Indicates The mean standard deviation of the frequency of the finite element model of order, Indicates The standard deviation of the frequency of the actual structure of the first-order offshore platform. The steps for calculating the mean and standard deviation are well-known techniques and will not be described in detail.

[0048] S7: Optimize the distance objective function and make the first correction to the mean and standard deviation.

[0049] The first correction is based on the distance objective function correction.

[0050] The distance objective function is constructed using the measured frequency of the offshore platform and the frequency of the initial finite element model. The distance objective function includes a mean objective function and a standard deviation objective function. The mean and standard deviation of the correction parameters are optimized and corrected based on the optimization objective.

[0051] In some embodiments of the present invention, the method for optimizing and correcting the mean and standard deviation of the correction parameter based on the optimization target includes: Taking the parameter mean and standard deviation of the correction parameter optimized and corrected as the solution target, the distance objective function is substituted into the multi-objective optimization algorithm, and the mean and standard deviation of the correction parameter are solved by the optimization algorithm, and the correction target threshold is set. When the output result of the mean objective function is less than the correction target threshold, and the output result of the standard deviation objective function is less than the correction target threshold, the correction is stopped, and the mean and standard deviation at this time are used as the mean and standard deviation of the correction parameter after optimization and correction: .in, For the The first optimized corrected mean, For the The standard deviation after the first optimization correction.

[0052] S8: Construct distribution objective function.

[0053] The mean and standard deviation after the first correction are brought into the Kriging model to calculate the frequency of the finite element model after the first correction. Based on the measured frequency of the offshore platform and the corrected finite element model frequency, the KL divergence is used to construct the distribution objective function. The mean and standard deviation of the correction parameters are corrected for the second time based on the optimization objective.

[0054] Specifically, the first optimization correction parameter mean and standard deviation Substitute into the Kriging model to obtain the model frequency after distance correction ; Based on the measured frequency of the ocean platform Finite element model frequency after distance correction , KL divergence is used to construct the distribution objective function.

[0055] Compared with the Bhattacharyya distance indicator, the KL divergence (Kullback-Leibler divergence) indicator can better quantify the differences between small sample data distributions. and the actual structural frequency Calculation is performed to obtain the mean frequency of the finite element model: and the actual frequency mean of the offshore platform , finite element model frequency covariance matrix and the platform actual frequency covariance matrix , then the inverse KL divergence index between the finite element model and the actual structure frequency is: .

[0056] That is, as the distribution objective function.

[0057] S9: Optimize the distribution function and perform a second distance correction.

[0058] The second correction is based on the distribution objective function correction.

[0059] The mean and standard deviation after the first correction are brought into the Kriging model to calculate the frequency of the finite element model after the first correction. Based on the measured frequency of the offshore platform and the corrected finite element model frequency, the KL divergence is used to construct the distribution objective function. The mean and standard deviation of the correction parameters are corrected for the second time based on the optimization objective.

[0060] Taking the mean and standard deviation of the modified parameters as the solution target, the distribution objective function is substituted into the multi-objective optimization algorithm. The optimization algorithm input is the population size, number of iterations, fitness deviation value, and initial finite element model frequency. The optimization solution is continued. When the convergence condition is met, the solution is stopped to obtain the modified mean and standard deviation of the modified parameters. .in, For the The second optimized corrected mean, For the The standard deviation after the second optimization correction.

[0061] S10: Correction of precision error check.

[0062] The mean and standard deviation of the second optimized corrected parameters obtained in step S9 are Substitute the corrected finite element model frequency and calculate the error between the corrected finite element model frequency and the measured frequency. If the accuracy requirement is met, the corrected finite element model is obtained. If the accuracy requirement is not met, repeat steps S7 to S9 until the accuracy requirement is met.

[0063] The following takes the step-by-step random model correction of a jacket platform as an example to illustrate the effect of the model correction method provided by the present invention. In the embodiment of the present invention, a direct model correction method and a step-by-step model correction method are used to correct the model for the same jacket platform, and the model correction effects of the two methods are compared.

[0064] Firstly, 100 sets of modified parameter vectors were obtained by Monte Carlo sampling according to the mean and standard deviation of the modified parameters of the initial finite element model. The first five frequencies of the structure corresponding to each parameter vector were obtained as the initial finite element model frequency by substituting them into the finite element model. Subsequently, the Euclidean norm and KL divergence of the finite element model frequency and the measured frequency were constructed as distance correction index and distribution difference index, respectively; Then, the population size in the optimization algorithm is set to 200, the number of iterations is 100, and the fitness deviation value is 1×10 -6 , the input is the measured frequency, the mean and standard deviation of the modified parameters of the initial finite element model; Furthermore, the Euclidean norm of the finite element model frequency and the measured frequency is used to correct the distance between the two, and the distance index between the two frequency distributions is calculated. The distance index is less than 10 4 When , the distance correction is stopped, and the objective function is changed to KL divergence to continue the optimization solution; finally, the optimization algorithm outputs the mean and standard deviation of the correction parameters of the corrected finite element model.

[0065] Comparing the correction errors of the mean and standard deviation of the correction parameters by the direct method and the step-by-step method, the correction error of the mean of the correction parameters by the step-by-step method is no more than 0.01%, the correction error of the mean of the correction parameters by the direct method is between 0.009% and 0.015%, and the correction error of the standard deviation of the parameters by the step-by-step method is about 5.0%, which is smaller than the correction error of the standard deviation of the parameters by the direct method. This also shows that the correction effect of the step-by-step method is better than that of the direct method. In summary: the correction error of the mean and standard deviation of the structural parameters, refer to Figure 2 and Figure 3 ,It can be seen that the correction effects of the step-by-step method and the direct method are similar, and there is no obvious decrease in the correction error of the structural parameters.

[0066] refer to Figure 4 In order to study the corrected mean and standard deviation of the first five frequencies of the corrected finite element model, the first five frequency errors corrected by the direct method and the step-by-step method are visualized as a bar chart. It can be seen that the correction error of the direct method for the mean value of the structural frequency is 0.012%~0.018%, and the standard deviation correction error is about 6%~7%; the correction error of the step-by-step method for the mean value of the structural frequency has decreased significantly compared with the direct method, and its frequency mean correction error is 0.001%~0.008%, and the standard deviation correction error is 2%~2.5%. It can be concluded that the step-by-step method has a better correction effect on the structural correction parameters and frequency response, and has good correction stability.

[0067] In order to study the random model correction effect of the two correction strategies under the influence of different degrees of noise, small noise of 0.15% and 1% and large noise of 3% and 5% were added to the measured frequency, and the direct method and the step-by-step method were used to correct the random model. Both methods used the NSGA-Ⅱ optimization algorithm for iterative solution, and the parameter settings were the same as before, that is, the population size was 200, the number of iterations was 100, the fitness deviation value was 1×10-6, the optimization algorithm output results were the mean and standard deviation of the correction parameters, and the surrogate model was the active learning Kriging model. In order to avoid contingency, the two methods were used to perform 20 random model corrections for different noise levels, and the average value was taken to observe the correction effect of the mean and standard deviation of the structural correction parameters.

[0068] In order to observe the correction of the distribution difference between the finite element model frequency and the measured frequency by the step-by-step correction method, the confidence ellipses of the first two frequencies of the corrected finite element model and the measured frequency under the influence of 0.15% noise are plotted. It can be seen that the center points of the corrected frequency confidence ellipse and the measured frequency confidence ellipse are almost completely coincident, indicating that the correction effect of the mean frequency of the finite element model is good; the lengths of the major and minor axes of the two ellipses are similar, indicating that the correction effect of the frequency standard deviation is good. Compared with the direct correction effect, the tilt angles of the two ellipses corrected by the step-by-step method are smaller, and the overall fitting degree of the two ellipses has been significantly improved, which proves the superiority of the step-by-step method for the correction effect of frequency distribution differences.

[0069] refer to Figure 5 and Figure 6 The correction effect of the step-by-step random model correction method proposed in the present invention on the mean and standard deviation of the structural parameters is similar to that of the direct random model correction method, but the correction effect of the step-by-step method on the frequency distribution difference of the finite element model is very superior. The corrected finite element model frequency confidence ellipse is almost completely consistent with the measured frequency confidence ellipse, and the overall correction effect is better than that of the direct random model correction method.

[0070] The step-by-step method divides the direct method into distance correction and distribution difference correction, and changes the one-time optimization calculation of the direct method to two. If the number of iterations is too many, it will affect the efficiency of the structural random model correction. The setting conditions use the step-by-step method and the direct method to perform 20 model corrections respectively. The multi-objective optimization algorithm sets the population size to 200, and performs statistical analysis on the average number of iterations of the four working conditions. The number of iterations when the KL divergence value between the finite element model frequency and the measured frequency reaches the minimum value is the final effective number of iterations. The average number of effective iterations for each working condition is as follows: Figure 7 shown.

[0071] For the direct method, the average effective number of iterations for condition E1 is 9.6, which is the least among the four conditions. The average effective number of iterations for condition E2 is 10.1, and the average effective number of iterations for condition E3 is 10. The condition E4 has the most iterations, with an average effective number of iterations of 10.4.

[0072] For the step-by-step method, the average number of effective iterations of the four working conditions has decreased significantly compared with the direct method. The average number of effective iterations of working condition E1 is 5.1 times, the number of iterations of working condition E2 is 5 times, and the working condition with the most iterations is E3, with an average of 5.2 times. Working condition E4 has the least average number of effective iterations among the four working conditions, which is 4.9 times.

[0073] In summary: the step-by-step random model correction method proposed in the present invention has stronger optimization performance and is not easy to fall into the local optimal solution. Although the step-by-step method divides the optimization process into two steps, the number of iterations is significantly reduced compared with the direct method, indicating that the step-by-step method can be better applied to the correction of actual structural random models.

[0074] Through the previous study on the random model correction effect of the direct method and the step-by-step method, it was found that a small number of abnormal correction effects would occur during multiple tests. In order to be closer to the actual engineering application scenario, the finite element model frequency with 0.15% noise added to the E1 condition was selected as the measured frequency, and the direct method and the step-by-step method were used to perform 100 model corrections, respectively, and the mean and standard deviation errors of the corrected structural parameters were observed.

[0075] refer to Figure 8 , it can be seen that: for the mean value of the structural correction parameter, the fluctuation of the direct method in the test is very obvious, while the correction result of the step-by-step method in the test is relatively stable. At the same time, the stability of the model correction result can also be measured by the standard deviation of the mean correction error of the 100 test parameters. The standard deviations of the mean correction errors of the elastic modulus E, density D, and the outer diameter DLP of the lower half of the column by the direct method are 0.017%, 0.016%, and 0.011%, respectively, and the standard deviations of the mean correction errors of the elastic modulus E, density D, and the outer diameter DLP of the lower half of the column by the step-by-step method are 0.0034%, 0.004%, and 0.0026%, respectively. It can be seen that the step-by-step method has a more stable correction effect on the mean value of the structural parameters.

[0076] The corrected standard deviation of the structural parameters is plotted as Fig. 9 As shown. The fluctuation of the direct method in the test is still very obvious, and the correction result of the step-by-step method in the test is relatively stable. The standard deviation of the correction error of the mean value of the test parameters for 100 times is used to measure the stability of the model correction results. The standard deviations of the correction errors of the elastic modulus E, density D, and the outer diameter DLP of the lower half of the column by the direct method are 6.86%, 7.52%, and 8.27%, respectively. The standard deviations of the correction errors of the elastic modulus E, density D, and the outer diameter DLP of the lower half of the column by the step-by-step method are 1.456%, 1.51%, and 1.72%, respectively. It can be seen that the correction effect of the standard deviation of the structural parameters by the step-by-step method is more stable.

[0077] In summary: the step-by-step correction method proposed in the present invention has better stability than the traditional direct method.

[0078] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. It should be pointed out that for those skilled in the art, any modification, equivalent replacement and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention. Therefore, the protection scope of the patent application shall be based on the protection scope of the attached claims.

Claims

1. A step-by-step fast stochastic model correction method for offshore platform structures, characterized in that: The following steps are involved: According to the design parameters of the offshore platform, the initial finite element model of the offshore platform is established; Among the design parameters, select parameters to be corrected, and determining the maximum and minimum values ​​of each parameter to be corrected; The maximum and minimum values ​​of each parameter to be corrected are calculated. Sampling times, get the modified parameter matrix , each correction parameter vector Include Correction parameters; Substituting the modified parameter matrix into the initial finite element model to obtain an updated finite element model, performing modal analysis or finite element analysis on the updated finite element model to obtain a structural frequency vector corresponding to the modified parameter vector; A Kriging model is established based on the modified parameter vector and its corresponding structural frequency vector; Calculate the initial finite element model frequency based on the modified parameter vector and Kriging model; Based on the mean and standard deviation of the initial finite element model frequency and the mean and standard deviation of the measured frequency of the offshore platform, a distance objective function is constructed, and the mean and standard deviation of the correction parameters are corrected for the first time based on the optimization objective; The mean and standard deviation after the first correction are brought into the Kriging model to calculate the frequency of the finite element model after the first correction. Based on the measured frequency of the offshore platform and the corrected finite element model frequency, the KL divergence is used to construct the distribution objective function, and the mean and standard deviation of the correction parameters are corrected for the second time based on the optimization objective. The mean and standard deviation of the correction parameters after the second correction are substituted into the Kriging model to obtain the finite element model frequency after the second correction. The error between the finite element model frequency after the second correction and the measured frequency of the offshore platform is calculated. If the error meets the accuracy requirements, the corrected offshore platform finite element model is obtained based on the finite element model frequency after the second correction.

2. The step-by-step fast random model correction method for offshore platform structure according to claim 1, characterized in that: The step of selecting the parameter to be corrected comprises: A sensitivity analysis is performed on the design parameters, and parameters with high sensitivity are selected as parameters to be corrected.

3. The step-by-step fast random model correction method for offshore platform structure according to claim 1 or 2, characterized in that: The design parameters of the offshore platform include structural parameters and material parameters.

4. The step-by-step offshore platform structure rapid stochastic model correction method according to claim 1, characterized in that: The steps to establish the Kriging model include: Each set of correction parameter vectors and corresponding structural frequencies are recorded as a set of sample data. The first 80% of the sample data were selected as the training set, and the remaining sample data were selected as the test set to establish the Kriging model.

5. The step-by-step fast random model correction method for offshore platform structure according to claim 1 or 4, characterized in that: After establishing the Kriging model, the following steps are also included: The effectiveness evaluation index determination coefficient R2 and relative root mean square error RMSE were used to verify the established Kriging model and set the model verification accuracy; If the accuracy of the Kriging model meets the model verification accuracy requirements, the established Kriging model is used to continue calculating the initial finite element model frequency. If the accuracy of the Kriging model does not meet the model verification accuracy requirements, the Latin hypercube sampling steps are repeated, and the corrected parameter matrix obtained by Latin hypercube sampling is re-substituted into the initial finite element model to re-establish the Kriging model until the accuracy of the Kriging model meets the model verification accuracy requirements.

6. The step-by-step fast random model correction method for offshore platform structure according to claim 1, characterized in that: The steps to calculate the initial finite element frequencies include: Calculate the initial mean and standard deviation of each correction parameter, and Monte Carlo sampling is performed to obtain the finite element model parameter samples. ,in, For the The initial mean and standard deviation vector of the modified parameters obtained by Monte Carlo sampling; Substitute the parameter samples into the Kriging model to obtain the initial finite element model frequency .

7. The step-by-step fast random model correction method for offshore platform structure according to claim 1 or 6, characterized in that: The distance objective function includes a mean objective function and a standard deviation objective function, and the step of constructing the distance objective function includes: ； ； in: represents the mean objective function, represents the standard deviation objective function, represents the frequency order, Indicates The mean of the frequencies of the finite element model, Indicates The mean frequency of the actual structure of the first-order offshore platform, Indicates The mean standard deviation of the frequency of the finite element model of order, Indicates The standard deviation of the frequency of the actual structure of the first-order offshore platform, .

8. The step-by-step fast random model correction method for offshore platform structure according to claim 7, characterized in that: The method for optimizing and correcting the mean and standard deviation of the correction parameters based on the optimization target includes: Taking the optimization and correction of the parameter mean and standard deviation of the correction parameters as the solution goal, the distance objective function is substituted into the multi-objective optimization algorithm, the optimization algorithm is used to solve the mean and standard deviation of the correction parameters, and a correction target threshold is set. When the output result of the mean objective function is less than the correction target threshold, and the output result of the standard deviation objective function is less than the correction target threshold, the correction is stopped, and the mean and standard deviation at this time are used as the mean and standard deviation of the correction parameters after optimization and correction.

9. The step-by-step fast random model correction method for offshore platform structure according to claim 1, characterized in that: Methods for constructing distribution objective functions include: in, is the mean frequency of the finite element model, is the actual frequency mean of the marine platform, is the frequency covariance matrix of the finite element model , is the actual frequency covariance matrix of the platform.

Images