Method and device for correcting simulation parameters of ship lock structure, and electronic equipment

Abstract

The application provides a lock structure simulation parameter correction method and device and electronic equipment. The method comprises the following steps: obtaining historical displacement measured data of a lock structure to form standardized measured data; determining to-be-corrected parameters according to a lock structure simulation model; performing global search on the to-be-corrected parameters by using a covariance matrix self-adaptive evolution algorithm, obtaining distribution characteristic parameters with optimal matching degree between a displacement response prediction value of a model output and the standardized measured data by iteratively optimizing the correlation between adaptive learning parameters, and determining an initial parameter distribution of the to-be-corrected parameters; combining the to-be-corrected parameters and the displacement response prediction value to construct an augmented state space model; taking real-time lock displacement data as an observation value, and adopting a Kalman filtering algorithm to perform recursive assimilation on augmented state variables in the space model, so as to realize the update correction of the to-be-corrected parameters. The application can assimilate real-time lock displacement data and dynamically update and correct the to-be-corrected parameters in the lock structure model.

Description

Technical Field

[0001] This application relates to the field of analysis technology for the structural safety and operational stability of locks, and in particular to a method, device and electronic equipment for correcting simulation parameters of lock structures. Background Technology

[0002] In ship lock engineering, the system consisting of the lock chamber, lock bottom plate, and foundation is subjected to complex environments such as water pressure, seepage pressure, and temperature changes over a long period of time. The stress distribution and mechanical parameters (such as the elastic modulus) within the structure will continuously change with the increase of operating time. In order to ensure the anti-slip, anti-buoyancy, and overall structural safety of the ship lock under complex working conditions, it is crucial to acquire and correct these dynamically changing mechanical parameters in real time.

[0003] However, current parameter calibration methods still have some limitations in practical applications: First, existing parameter calibration processes are often discrete or manually triggered, mostly involving one-time adjustments within a specific time period, failing to achieve synchronous updates of parameters and monitoring data. Second, these methods mostly focus on the correction of local parameters, lacking a real-time feedback mechanism that can effectively transform measured data into updates to the overall structural parameters. This results in parameter updates often lagging behind the actual evolution of the structure, failing to reflect the continuous changes in parameters over operating time, and limiting the timeliness and accuracy of safety monitoring. Therefore, developing a method that can assimilate monitoring data in real time and dynamically update structural parameters is of great significance for achieving accurate assessment of the safety status of ship locks. Summary of the Invention

[0004] The purpose of this application is to provide a method, device and electronic equipment for correcting simulation parameters of a lock structure, which can assimilate real-time lock displacement data and dynamically update the parameters to be corrected in the lock structure model.

[0005] Firstly, this application provides a method for correcting simulation parameters of a lock structure. The method includes: acquiring historical measured displacement data of the lock structure and forming standardized measured data through preprocessing; determining the parameters to be corrected based on the lock structure simulation model; using a covariance matrix adaptive evolution algorithm to perform a global search on the parameters to be corrected, and obtaining the distribution feature parameters that best match the displacement response prediction value output by the lock structure simulation model with the standardized measured data through adaptive learning and iterative optimization, thereby determining the initial parameter distribution of the parameters to be corrected; constructing an augmented state space model based on the parameters to be corrected, the displacement response prediction value, and the initial parameter distribution; using real-time lock displacement data as observations, and employing a Kalman filter algorithm to recursively assimilate the augmented state variables in the augmented state space model to achieve the update and correction of the parameters to be corrected.

[0006] Furthermore, the steps described above for obtaining historical displacement measurement data of the lock structure and forming standardized measurement data through preprocessing include: generating standardized measurement data that meets the requirements of subsequent real-time assimilation through the following processing methods: initially screening the historical displacement measurement data to remove duplicate and invalid records; filling missing values ​​in the historical displacement measurement data using the adjacent mean imputation method based on distribution characteristics; calculating the arithmetic mean of multiple repeated observations of the same monitoring point on the same day in the historical displacement measurement data as the unique representative value for that day; and resampling and time-aligning the horizontal and vertical displacement data in the historical displacement measurement data based on a preset assimilation step size to ensure that different types of monitoring variables remain consistent on the time scale.

[0007] Furthermore, the steps described above, which utilize the covariance matrix adaptive evolution algorithm to perform a global search of the parameters to be corrected, adaptively learn the correlation between parameters and iteratively optimize to obtain the distribution characteristic parameters that best match the displacement response prediction values ​​output by the lock structure simulation model with the standardized measured data, and thus determine the initial parameter distribution of the parameters to be corrected, include: setting the initial state of the covariance matrix adaptive evolution algorithm to form a multidimensional normal distribution; the initial state includes: the initial mean vector, initial step size, and initial covariance matrix of the parameters to be corrected, set according to lock design experience values; for each iteration, the following sampling, evolution, filtering, and updating steps are performed: sampling from the multidimensional normal distribution to generate a specified number of individuals A sample set consisting of parameter vectors; each parameter vector simultaneously contains a combination of candidate values ​​for all parameters to be corrected; each sample is sequentially substituted into the lock structure simulation model for evolution to obtain the corresponding physical field response prediction value; based on the physical field response prediction value and the corresponding value in the standardized measured data, the objective function value corresponding to each sample is calculated; the objective function values ​​corresponding to each sample are sorted, and a specified number of high-quality samples are selected; the distribution characteristic parameters of the multidimensional normal distribution are updated using the high-quality samples; the distribution characteristic parameters include: mean vector, step size, and covariance matrix; until the convergence condition is met or the maximum number of iterations is reached, the initial parameter distribution of the parameters to be corrected is constructed based on the final distribution characteristic parameters of the multidimensional normal distribution.

[0008] Furthermore, the steps described above for constructing an augmented state-space model based on the parameters to be corrected, the predicted displacement response, and the initial parameter distribution include: constructing an augmented state vector using the parameters to be corrected and the predicted displacement response as internal state variables; constructing an initial state set including multiple augmented state vectors using the Monte Carlo sampling method according to the initial parameter distribution; and constructing an augmented state-space model based on the augmented state vectors, nonlinear state evolution functions, and process noise in the initial state set.

[0009] Furthermore, the above-mentioned steps, which use real-time lock displacement data as observations and employ the Kalman filter algorithm to recursively assimilate the augmented state variables in the augmented state space model to update and correct the parameters to be corrected, include: taking each vector in the current initial state set as the posterior augmented state vector of the previous time step, and performing the following prediction, statistics, gain calculation, and assimilation steps: based on the augmented state space model, performing vector prediction on the posterior augmented state vector of the previous time step and statistically analyzing the prediction results to determine the state prior deviation matrix and displacement prediction deviation matrix at the current time step; and calculating the state at the current time step based on the state prior deviation matrix and displacement prediction deviation matrix. The state-displacement cross-covariance is calculated; based on the displacement prediction deviation matrix at the current moment, the displacement prediction autocovariance at the current moment is calculated; based on the state-displacement cross-covariance, displacement prediction autocovariance, and the preset measurement error covariance matrix at the current moment, the Kalman gain matrix at the current moment is calculated; the real-time value corresponding to the vector at the current moment is extracted from the real-time lock displacement data; based on the Kalman gain matrix at the current moment, the displacement prediction deviation of the real-time value at the current moment is corrected in real time to realize the update and correction of the parameter to be corrected; the updated and corrected posterior augmented state vector is used as the posterior augmented state vector of the previous moment, and the prediction, statistics, gain calculation, and assimilation steps are continued.

[0010] Furthermore, the steps described above, based on the augmented state space model, to predict the posterior augmented state vector of the previous time step and statistically analyze the prediction results to determine the state prior deviation matrix and displacement prediction deviation matrix at the current time step, include: for each vector in the initial state set, predicting the posterior augmented state vector of the previous time step according to the augmented state space model to determine the prior augmented state vector at the current time step; extracting the displacement response from the prior augmented state vector at the current time step through the observation matrix to determine the displacement prediction value at the current time step; calculating the mean of the prior augmented state vectors at the current time step corresponding to multiple vectors in the initial state set to determine the augmented state prediction mean; calculating the mean of the displacement prediction value at the current time step corresponding to multiple vectors in the initial state set to determine the displacement prediction mean; calculating the state prior deviation matrix at the current time step based on the prior augmented state vectors and the augmented state prediction mean at the current time step corresponding to multiple vectors in the initial state set; and calculating the displacement prediction deviation matrix at the current time step based on the displacement prediction value and the displacement prediction mean at the current time step corresponding to multiple vectors in the initial state set.

[0011] Furthermore, the above-mentioned steps for real-time correction of the displacement prediction deviation of the real-time value based on the Kalman gain matrix at the current time to achieve the update and correction of the parameter to be corrected include: calculating the difference between the real-time value and the displacement prediction value corresponding to the vector; obtaining the product of the Kalman gain matrix at the current time and the difference; and obtaining the sum of the product and the prior augmented state vector at the current time to obtain the posterior augmented state vector at the current time.

[0012] Secondly, this application also provides a device for correcting simulation parameters of a lock structure. The device includes: a data preprocessing module for acquiring historical measured displacement data of the lock structure and forming standardized measured data through preprocessing; a parameter determination module for determining the parameters to be corrected based on the lock structure simulation model; an evolution iteration module for performing a global search on the parameters to be corrected using an adaptive evolution algorithm of the covariance matrix, and obtaining the distribution characteristic parameters that best match the displacement response prediction value output by the lock structure simulation model with the standardized measured data through adaptive learning and iterative optimization, so as to determine the initial parameter distribution of the parameters to be corrected; a model construction module for constructing an augmented state space model based on the parameters to be corrected, the displacement response prediction value, and the initial parameter distribution; and a recursive assimilation module for using real-time lock displacement data as observations and employing a Kalman filter algorithm to recursively assimilate the augmented state variables in the augmented state space model to achieve the update and correction of the parameters to be corrected.

[0013] Thirdly, this application also provides an electronic device, including a processor and a memory, wherein the memory stores computer-executable instructions that can be executed by the processor, and the processor executes the computer-executable instructions to implement the method described in the first aspect above.

[0014] Fourthly, this application also provides a computer-readable storage medium storing computer-executable instructions, which, when invoked and executed by a processor, cause the processor to implement the method described in the first aspect above.

[0015] The method, apparatus, and electronic equipment for correcting simulation parameters of a lock structure provided in this application first acquire historical measured displacement data of the lock structure and preprocess it to form standardized measured data. Then, the parameters to be corrected are determined based on the lock structure simulation model. Next, an adaptive covariance matrix evolution algorithm is used to perform a global search on the parameters to be corrected. By adaptively learning the correlation between parameters and iteratively optimizing, the distribution characteristic parameters that best match the displacement response prediction value output by the lock structure simulation model with the standardized measured data are obtained, thus determining the initial parameter distribution of the parameters to be corrected. Further, based on the parameters to be corrected, the displacement response prediction value, and the initial parameter distribution, an augmented state space model is constructed. Finally, real-time lock displacement data is used as observations, and a Kalman filter algorithm is employed to recursively assimilate the augmented state variables in the augmented state space model, thereby updating and correcting the parameters to be corrected. This application presents a method for correcting simulation parameters of a lock structure based on covariance evolution and filtering assimilation. It includes a real-time feedback mechanism that can effectively convert real-time measured data into updates of the overall structural parameters. This mechanism can simultaneously update the parameters to be corrected and the real-time measured data, i.e., it can assimilate real-time lock displacement data and dynamically update the parameters to be corrected in the lock structure model. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the specific embodiments of this application or the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0017] Figure 1 A flowchart illustrating a method for correcting simulation parameters of a ship lock structure, provided as an embodiment of this application; Figure 2 A schematic diagram illustrating the workflow of a method for correcting simulation parameters of a lock structure provided in this application embodiment; Figure 3 A structural block diagram of a device for correcting simulation parameters of a lock structure provided in an embodiment of this application; Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0018] The technical solutions of this application will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0019] The following two problems exist with existing lock parameter calibration methods: (1) Existing calibration processes are often discrete or manually triggered, and adjustments are mostly made once within a specific time period, making it impossible to achieve synchronous updates of parameters and monitoring data. (2) These methods mostly focus on the correction of local parameters and lack a real-time feedback mechanism that can effectively transform measured data into updates of overall structural parameters. This results in parameter updates often lagging behind the actual evolution of the structure, making it difficult to reflect the continuous changes in parameters over time, thus limiting the timeliness and accuracy of safety monitoring.

[0020] This application provides a method, apparatus, and electronic device for correcting simulation parameters of a lock structure, capable of assimilating real-time lock displacement data and dynamically updating the parameters to be corrected in the lock structure model. To facilitate understanding of this embodiment, a detailed description of the method for correcting simulation parameters of a lock structure disclosed in this application is provided first. Figure 1 A flowchart illustrating a method for correcting simulation parameters of a ship lock structure, provided in this application embodiment, is included. The method specifically comprises the following steps: Step S102: Obtain historical displacement measurement data of the lock structure and form standardized measurement data through preprocessing; Historical displacement measurement data includes: historical measurement data of horizontal and vertical displacement; the above preprocessing process may include: outlier removal, missing value correction and other methods to complete data cleaning, and may include unifying the time scale of the data to form standardized measurement data that meet the requirements of subsequent analysis.

[0021] Step S104: Determine the parameters to be corrected based on the lock structure simulation model; The aforementioned lock structure simulation model can be a parametric finite element simulation model. In this step, the core physical parameters involved in the lock structure simulation can be identified as parameters to be corrected based on engineering requirements and simulation objectives. Specifically, based on the constructed lock structure simulation model, key physical parameters can be screened and their basic ranges defined; for example, key parameters can be determined from parameters related to mechanical properties, thermal properties, seepage hydraulic properties, and media physical properties. The value range of each key parameter can then be clarified by combining engineering design specifications, basic design data, and model constraints.

[0022] Step S106: Use the covariance matrix adaptive evolution algorithm to perform a global search for the parameters to be corrected. By adaptively learning the correlation between parameters and iteratively optimizing, obtain the distribution feature parameters that make the displacement response prediction value output by the lock structure simulation model match the standardized measured data best, so as to determine the initial parameter distribution of the parameters to be corrected. This step includes algorithm initialization, iterative sampling and evaluation, adaptive distribution update, and convergence output, with the goal of providing an accurate initial parameter distribution for the subsequent real-time assimilation process. The specific implementation details will be provided later.

[0023] Step S108: Based on the parameters to be corrected, the predicted displacement response, and the initial parameter distribution, construct an augmented state-space model; In practice, the parameters to be corrected and the predicted values ​​of displacement response are used as internal state variables to construct an augmented state vector; based on the initial parameter distribution, the Monte Carlo sampling method is used to construct an initial state set including multiple augmented state vectors; based on the augmented state vectors, nonlinear state evolution functions and process noise in the initial state set, an augmented state space model is constructed.

[0024] Step S110: Using real-time lock displacement data as observations, the Kalman filter algorithm is used to recursively assimilate the augmented state variables in the augmented state space model, thereby updating and correcting the parameters to be corrected.

[0025] The aforementioned real-time lock displacement data includes both horizontal and vertical displacement data acquired in real time. This step, through an assimilation process based on a Kalman filter algorithm, allows for simultaneous updating and correction of the real-time data and parameters.

[0026] This application embodiment also provides another method for correcting simulation parameters of lock structures, which is implemented based on the above embodiment. This embodiment focuses on describing the data preprocessing process, the process of determining the initial parameter distribution of the parameters to be corrected based on the covariance matrix adaptive evolution algorithm, the process of constructing the augmented state space model, and the parameter correction process based on the Kalman filter algorithm.

[0027] Step S102 above, which involves obtaining historical displacement measurement data of the lock structure and preprocessing it to form standardized measurement data, specifically includes: Standardized measured data that meets the requirements for subsequent real-time assimilation can be generated through the following processing methods: (1) Conduct preliminary screening of historical displacement measurement data to remove duplicate and invalid records; The aforementioned historical displacement measurement data includes both horizontal and vertical displacements. The horizontal displacement is determined based on the local spatial coordinate system of the lock, where: the x-direction is defined as the downstream direction, i.e., parallel to the lock centerline and pointing downstream, used to characterize the longitudinal displacement of the lock chamber structure; the y-direction is defined as the cross-current direction, i.e. perpendicular to the lock centerline and pointing from the left bank to the right bank, used to characterize the lateral deformation of the lock wall structure.

[0028] (2) For missing values ​​in the historical displacement measurement data, the adjacent mean filling method is used to fill in the missing values ​​according to the distribution characteristics; (3) For the repeated observations of the same monitoring point on the same day in the historical displacement measurement data, the arithmetic mean is calculated as the unique characteristic value of that day; this processing method is used to eliminate random observation errors and reduce data redundancy.

[0029] (4) Based on the preset assimilation step size (e.g., in "days"), the horizontal and vertical displacement data in the historical displacement measurement data are resampled and time-aligned to ensure that different types of monitoring variables remain consistent on the time scale, thereby generating standardized measurement data that meet the requirements of subsequent real-time assimilation.

[0030] Further, step S106 above, which uses the covariance matrix adaptive evolution algorithm to perform a global search of the parameters to be corrected, and obtains the distribution characteristic parameters that best match the displacement response prediction values ​​output by the lock structure simulation model with the standardized measured data through adaptive learning of the correlation between the parameters and iterative optimization, in order to determine the initial parameter distribution of the parameters to be corrected, includes: (1) Set the initial state of the covariance matrix adaptive evolution algorithm to form a multidimensional normal distribution; the initial state includes: the initial mean vector, initial step size and initial covariance matrix of the parameters to be corrected set according to the empirical values ​​of lock design. (2) For each iteration, the following sampling, evolution, filtering, and update steps are performed: A sample set consisting of parameter vectors of a specified number of individuals is generated by sampling from a multidimensional normal distribution; each parameter vector contains a combination of candidate values ​​for all parameters to be corrected. (2.1) Substitute each sample into the lock structure simulation model in sequence to evolve and obtain the corresponding physical field response prediction value; (2.2) The objective function value for each sample is calculated based on the predicted physical field response value and the corresponding value in the standardized measured data; (2.3) Sort the samples according to the objective function value corresponding to each sample and select a specified number of high-quality samples; (2.4) Update the distribution characteristic parameters of the multidimensional normal distribution using high-quality samples; the distribution characteristic parameters include: mean vector, step size and covariance matrix; (3) Until the convergence condition is met or the maximum number of iterations is reached, the initial parameter distribution of the parameter to be corrected is constructed based on the distribution characteristic parameters of the final multidimensional normal distribution.

[0031] The specific implementation process of the above algorithm is described in detail below: 1. Algorithm initialization.

[0032] The initial state of the covariance matrix adaptive evolution algorithm is set. First, the initial mean of the parameters to be corrected is set based on empirical values ​​from lock design. The components are, in order: internal friction angle of the buried soil, shear friction coefficient, shear cohesion, uplift pressure reduction coefficient, distance from the upstream face to the drainage hole, elastic model of the dam body in the x / z direction, and thermal expansion coefficient in the x / z direction. Next, the initial step size is set. and the initial covariance matrix , Let it be the identity matrix, indicating that the parameters are initially independent. Furthermore, set the population size to... and initialize the step size evolution path vector. Covariance Evolutionary Path Vector The zero vector contains pre-defined weights for mean updates. By using coordinate transformation, the physical parameters of the lock (internal friction angle in the soil, shear friction coefficient, etc.) are mapped to a dimensionless space, eliminating the influence of parameters of different orders of magnitude on the search efficiency.

[0033] 2. Iterative sampling and evaluation.

[0034] In the g-th iteration: based on the mean of the current parameter to be corrected. Step length Covariance Matrix Defined multidimensional normal distribution In this process, a sample set consisting of parameter vectors from λ individuals is randomly generated. Each parameter vector contains a candidate value combination for all parameters to be corrected. Each sample is sequentially substituted into the lock structure simulation model for a forward modeling calculation to obtain the corresponding predicted physical field response value, specifically including downstream displacement, transverse displacement, and vertical displacement reflecting the lock deformation characteristics. The normalized root mean square error (NSE) is used as the objective function. By comparing the predicted physical field response value with the standardized measured displacement data determined in step S102, the objective function value corresponding to each sample is calculated. , as its fitness.

[0035] 3. Adaptive distribution update.

[0036] The samples are ranked according to fitness, and the optimal one is selected. For each sample, the parameter evolution is performed sequentially according to the following steps: Mean update: Calculate this The weighted average of the preferred samples is used to update the mean of the next generation distribution. ,in, This represents the individual ranked i in fitness among λ samples. By updating the mean, the search center is made to approach the parameter region with high fitness (i.e., the simulated displacement and the measured displacement are highly consistent).

[0037] Step size adaptive update: The step size evolution path is updated by using the offset of the current generation mean of the lock's physical parameters relative to the previous generation mean. : ; Then calculate the next iteration step size. ,in, The step size evolution time constant, The damping coefficient is... The effective population size is given by the formula The calculations show that this step aims to dynamically adjust the search span in the parameter space based on the sensitivity of the lock parameters to displacement response.

[0038] The covariance matrix is ​​adaptively updated using the evolutionary path information of the current generation and historical best samples. This allows for the updating of the covariance matrix. Its core lies in automatically identifying and capturing the coupling relationships between various physical parameters of the lock, thereby achieving precise adaptive adjustment of the search distribution shape.

[0039] Covariance evolution path update: ,in, This is the time constant of the covariance matrix evolution path, used to control the attenuation weight of historical search direction information; As a correction factor, when рσ is reasonable, =1, when рσ is too large =0. The algorithm's hyperparameters... It is automatically generated using a preset empirical formula based on parameter dimension n, without the need for manual intervention or adjustment.

[0040] Covariance matrix update: The update formula combines "rank-μ update" and "rank-1 update": ; in, and These are the covariance matrices of the current generation and the next generation, respectively, with dimensions n×n (where n is the number of parameters to be optimized), representing the coupling relationship between parameters and the shape of the search distribution. The offset vector of the i-th preferred sample .

[0041] , The learning rate controls the weights of the evolutionary path term and the population distribution term, respectively. ; ; in, and These represent the rank-1 and rank-μ update learning rates, respectively. .

[0042] 4. Convergence and Output.

[0043] When the iteration reaches the preset maximum number of algebras, or the step size of the parameter distribution. The algorithm is considered convergent when the change in mean is less than a set threshold. This is based on the distribution characteristic parameters of the final multidimensional normal distribution. The initial parameter distribution of the parameters to be corrected is constructed, which is the initial parameter distribution of the parameters to be corrected in the EnKF algorithm. This distribution characterizes the optimal estimation center of the mechanical parameters of the lock structure and the uncertainty of the correlation between parameters under the historical best-fit state, providing a physically constrained initial state field for subsequent dynamic assimilation based on real-time monitoring data.

[0044] Further, step S108 above, the step of constructing an augmented state-space model based on the parameters to be corrected, the predicted displacement response values, and the initial parameter distribution, includes: (1) Using the parameters to be corrected and the predicted values ​​of displacement response as internal state variables, an augmented state vector is constructed; Using the key physical parameter vector of the lock Displacement response prediction values ​​(including horizontal displacement) and Vertical displacement ) is used as an internal state variable to construct an augmented state vector X.

[0045] ; in, This represents the vector of key physical parameters of the lock, i.e., the vector corresponding to the parameters to be corrected. The horizontal displacement is in the x-direction; The displacement is the horizontal displacement in the y-direction; This represents the vertical settlement displacement.

[0046] (2) Based on the initial parameter distribution, the Monte Carlo sampling method is used to construct an initial state set including multiple augmented state vectors; Based on the initial parameter distribution output in step S106 The Monte Carlo sampling method is used to construct N set members to form the initial state set. Each set member Each contains a set of initial values ​​of parameters to be corrected that satisfy statistical characteristics, and together with the initial displacement response calculated by the simulation model, they constitute the initial state space for subsequent recursive assimilation.

[0047] (3) Based on the augmented state vectors, nonlinear state evolution functions, and process noise in the initial state set, an augmented state space model is constructed. The augmented state space model is as follows: ; in, Let represent the posterior augmented state vector of the i-th set member at time t (corresponding to the previous time). This is process noise, usually assumed to be zero-mean white noise; It is a nonlinear state evolution function (implemented by the lock structure simulation model) used to characterize the physical relationship between parameters and displacement; Let t+1 be the prior augmented state vector of the i-th set member (i.e., vector) at time t+1 (i.e., the current time).

[0048] Further, in step S110 above, the real-time lock displacement data is used as the observation value, and the Kalman filter algorithm is used to recursively assimilate the augmented state variables in the augmented state space model to realize the update and correction of the parameters to be corrected. This includes: Using each vector in the current initial state set as the posterior augmented state vector of the previous time step, perform the following prediction, statistics, gain calculation, and assimilation steps: (1) Based on the augmented state space model, the posterior augmented state vector of the previous time step is predicted by vector and the prediction results are statistically analyzed to determine the state prior deviation matrix and displacement prediction deviation matrix at the current time step. (1.1) For each vector in the initial state set, the posterior augmented state vector of the previous time step is predicted according to the augmented state space model to determine the prior augmented state vector of the current time step; the displacement response is extracted from the prior augmented state vector of the current time step through the observation matrix to determine the displacement prediction value of the current time step; the corresponding calculation formula includes the augmented state space model mentioned above and the following observation equation: ; in, Let be the predicted displacement value corresponding to the i-th set member at time t+1; The observation matrix is ​​used to extract the displacement response from the prior augmented state vector of the i-th set member at time t+1. ; in, for A zero matrix of dimension 1 The number of parameters to be corrected; for A 3D identity matrix, where m corresponds to the displacement response ( ( ) dimension.

[0049] (1.2) Calculate the mean of the augmented state vectors corresponding to the current time step of the multiple vectors in the initial state set to determine the mean of the augmented state prediction; calculate the mean of the displacement prediction based on the displacement prediction values ​​corresponding to the current time step of the multiple vectors in the initial state set to determine the mean of the displacement prediction; the corresponding formulas are as follows: ; ; Where N represents the number of vectors in the initial state set; This represents the mean of the augmented state prediction at time t+1; This represents the predicted mean displacement at time t+1.

[0050] (1.3) Calculate the state prior bias matrix for the current time based on the prior augmented state vectors and the mean of augmented state predictions corresponding to multiple vectors in the initial state set at the current time; calculate the displacement prediction bias matrix for the current time based on the displacement prediction values ​​and the mean of displacement predictions corresponding to multiple vectors in the initial state set at the current time. The corresponding formulas are as follows: ; ; in, This represents the state prior deviation matrix at time t+1; This represents the displacement prediction deviation matrix at time t+1.

[0051] (2) Calculate the state-displacement cross-covariance at the current moment based on the state prior deviation matrix and displacement prediction deviation matrix at the current moment; calculate the displacement prediction autocovariance at the current moment based on the displacement prediction deviation matrix at the current moment; the corresponding formulas are as follows: ; ; in, This represents the state-displacement cross-covariance at the current moment; This represents the displacement prediction autocovariance at the current moment.

[0052] (3) Calculate the Kalman gain matrix at the current time based on the state-displacement cross-covariance, displacement prediction autocovariance and the preset measurement error covariance matrix; ; Where R is the observation error covariance matrix, and its components are usually set according to the sensor calibration accuracy or factory standard deviation. Let represent the Kalman gain matrix at time t+1.

[0053] (4) Extract the real-time value of the vector corresponding to the current moment from the real-time lock displacement data; based on the Kalman gain matrix at the current moment, correct the displacement prediction deviation of the real-time value at the current moment in real time, so as to realize the update and correction of the parameter to be corrected. Specifically, the difference between the real-time value and the predicted displacement value corresponding to the vector is calculated; the product of the Kalman gain matrix at the current time and the difference is obtained; the sum of the product and the prior augmented state vector at the current time is calculated to obtain the posterior augmented state vector at the current time. The corresponding formulas are as follows: ; in, This represents the a priori unbounded augmented state vector, based on information at time t and the process model, predicting the state at time t+1. This prediction only considers the system's own evolutionary laws and does not incorporate new observational data at time t+1. This represents the integration of new observation data at time t+1. Then, the optimal estimate of the system state is obtained, i.e., the posterior augmented state vector at time t+1. The updated set. This is the posterior augmented state set at time t+1, whose mean is the optimal estimate of the system state at that time, and serves as the initial condition for the next time step.

[0054] The above steps are the state update and assimilation output steps, which are achieved by incorporating the measured displacement at time t+1. (including downstream displacement observations) Cross-flow displacement observations and vertical displacement observations ), utilizing Kalman gain Real-time correction of horizontal and vertical displacement prediction biases caused by water pressure, seepage, or temperature fluctuations. By augmenting the covariance information in the state vector, the displacement observation residuals are mapped to the parameter space, thereby achieving synchronous and linked correction of sensitive parameters such as the dam's elastic modulus and uplift pressure reduction factor.

[0055] (5) The updated and corrected posterior augmented state vector is used as the posterior augmented state vector of the previous time step, and the prediction, statistics, gain calculation and assimilation steps are continued.

[0056] The method for correcting the simulation parameters of the lock structure provided in this application realizes real-time correction of the assimilated monitoring data, which is beneficial to improving the prediction accuracy of the lock structure simulation model.

[0057] Here is another specific example: Step 1: First, based on the simulation model of a ship lock structure in Central my country, the parameter to be optimized is the internal friction angle of the buried soil. Coefficient of shear friction Shear cohesion Lifting pressure reduction factor Distance from upstream face to drainage hole , elastic modulus of dam body in x / z direction and coefficient of thermal expansion in the x / z direction and .

[0058] Step Two: Obtain historical measured data for the safety monitoring of the lock structure, including horizontal and vertical displacements. First, the raw data is initially screened to remove duplicate and invalid records. Second, for missing values ​​in the dataset, the adjacent mean method is used to fill in the gaps based on the evolution of the data over time. Further, for multiple observations of the same monitoring point within the same day, the daily arithmetic mean is calculated to obtain the daily observation value corresponding to the assimilation step size. Finally, using "day" as the time unit, the horizontal and vertical displacements are aligned to a unified time node, generating a standardized measured dataset that meets the requirements of subsequent real-time assimilation.

[0059] Step 3: Based on the standardized measured dataset, the Covariance Matrix Adaptive Evolution Algorithm (CMA-ES) is used to perform a global optimization search for the parameters to be corrected. By adaptively learning the correlation between parameters and iterating, the parameter estimates that best match the calculated response of the finite element model with the measured values ​​are obtained, and the initial probability distribution for the subsequent assimilation process is constructed based on these estimates. The initial search range for each parameter is set as shown in Table 1 below: Table 1

[0060] First, the search range of the parameters to be corrected is preset based on their physical background, and the median of this range is taken as the initial mean vector to ensure that the algorithm evolves from the geometric center of the parameter space. To eliminate the influence of different parameter magnitudes on optimization efficiency, this embodiment maps each parameter to its search range during the initialization phase. The normalized space. Within this normalized space, the initial global evolution step size is... The step size is set to 0.25, serving as a baseline factor for search intensity, and its value is determined with reference to the average level of the normalized spatial span. During the initialization phase, the global evolution step size is set to 0.25, which, as a baseline factor for search intensity, must be determined with reference to the average level of the search spatial span for each parameter. For multiple parameters with different physical scales, this step does not require setting a separate step size for each dimension. Instead, it utilizes the coordinate transformation and scale scaling mechanisms within the CMA-ES algorithm to automatically compensate and correct the actual search step size for each dimension parameter based on individual fitness feedback during iteration, thereby effectively eliminating the impact of different parameter magnitudes on global optimization efficiency.

[0061] For multiple parameters with different physical scales (such as the elastic modulus and permeability of materials with huge numerical differences), this step utilizes the coordinate transformation and scale scaling mechanism inside the CMA-ES algorithm, combined with the step-size evolution path. Evolutionary path of covariance The adaptive update automatically compensates and corrects the actual search step size of each dimension parameter based on individual fitness feedback during the iteration process, thereby effectively eliminating the impact of different parameter magnitudes on the global optimization efficiency.

[0062] Subsequently, the maximum number of iterations of the preset algorithm was used as the stopping criterion, and the optimization objective function was defined as a transformed form of the Nash efficiency coefficient (NSE) to quantitatively evaluate the degree of matching between the finite element simulation values ​​and the measured displacement sequences. The calculation formula is as follows: ; in, Let i be the simulated value of the i-th displacement. For the corresponding i-th displacement observation, This represents the mean of the displacement observations.

[0063] As the iteration process progresses, the algorithm continuously adjusts the covariance matrix through sampling, evaluation, and selection in each generation of the population, gradually converging the search space towards an increasing NSE value. When the maximum number of iterations is reached, the algorithm outputs the optimal parameter combination and its statistical distribution characteristics. Through the aforementioned global optimization search, an initial distribution that achieves the best fit between the model response and the measured data is obtained.

[0064] As the iteration process progresses, the algorithm continuously adjusts the mean through sampling, evaluation, and selection in each generation of the population. Step length and covariance matrix This process gradually converges the search space towards an increasing NSE value. When the maximum number of iterations is reached, the algorithm outputs the optimal parameter combination and its statistical distribution characteristics. Through this global optimization search, an initial distribution that best fits the model response to the measured data is obtained.

[0065] Step 4: Select the key lock parameter vector and the displacement response calculated by the simulation model as internal state variables, and define them together as the augmented state vector X of the augmented state space model. Use the optimization result of the covariance matrix adaptive evolution algorithm as the initial distribution of the lock parameters; select the measured horizontal displacement and settlement displacement as external observable variables z.

[0066] By analyzing the mapping relationship between internal state variables and external observable variables, the following state-space model and observation equations are constructed: ; ; in, For structural simulation model, This is the observation matrix for extracting observations from simulation results. In this model, the parameter part evolves over time by augmenting state variables, and the parameters are dynamically corrected through real-time feedback of displacement observations.

[0067] Process noise Modeled as zero-mean Gaussian white noise, with covariance matrices as follows: ,Right now: ; Assuming that the sensor's measured data contains independent Gaussian observation noise, the corresponding observation noise covariance matrix is: Based on experience, the observed noise covariance will be... Set it to 10-20% of the standard deviation of the observations. Construct a diagonal matrix accordingly. .

[0068] Step 5: This step achieves a smooth transition from the CMA-ES global search space to the EnKF dynamic assimilation space. To ensure that the subsequent Kalman update process conforms to the Gaussian distribution assumption, the following differential transformation logic is executed: Parameter space mapping: Based on the parameter types, the Logit transformation is used to map the mean of the physical space output by CMA-ES. With covariance A forward transformation to a dimensionless computational space yields the mapped statistical characteristics. and .

[0069] Within the aforementioned dimensionless computational space, with a preset set size N, an initial state vector set containing N members is generated based on Monte Carlo sampling. , where each parameter vector : ; Substituting the parameter components into the mechanical model of the lock structure, the initial physical response quantities (horizontal displacement and vertical displacement) are calculated, thus forming the initial unbounded augmented state vector set of EnKF. ,in .

[0070] Step Six: Data Assimilation and Update, Integrating Real-Time Observations Including downstream displacement observations Cross-flow displacement observations and vertical displacement observations Parameter assimilation is performed using an ensemble Kalman filter algorithm.

[0071] Since the assimilation algorithm operates in an unbounded space, the parameter set of the unbounded space must be restored to the physical space before calling the structural simulation model. Extract the members of the posterior unbounded augmented state set updated at time t. Parameter components in The physical parameters are restored using the inverse logit transform. : ; L and U are the preset lower and upper bounds for the search of key physical parameters.

[0072] The restored physical parameters The input is fed into the established mechanical model of the lock structure, and the predicted displacement response at time t+1 is obtained through numerical calculation. Subsequently, the prior unbounded augmented state vector at time t+1 is constructed: ; In subsequent calculations, the predicted displacement response and the prior unbounded augmented state vector satisfy the following matrix extraction relationship: ; in, For the observation matrix, extract the displacement response from the augmented vector: ; in, for A zero matrix of dimension 1 The number of parameters to be corrected is 9 in this embodiment; for A 3D identity matrix, where m corresponds to the displacement response ( The dimension of ) is 3 in this embodiment.

[0073] Returning to the dimensionless computational space, calculate the mean of the augmented state prediction. and displacement prediction mean This serves as a priori estimate of the system state at that moment: ; ; Calculate the state prior deviation matrix and the displacement prediction deviation matrix: ; ; State update. Calculate the state-displacement cross-covariance and displacement prediction autocovariance: ; ; Based on the calculated state-displacement cross-covariance and displacement prediction autocovariance Calculate the Kalman gain matrix: ; in, The state-observation cross-covariance matrix, Let R be the displacement prediction autocovariance matrix. Let R be the observation error covariance matrix. Since the observations at each measuring point are independent, R is constructed as a diagonal matrix, with its diagonal elements... The standard deviation is determined based on the calibration accuracy of each sensor in a real engineering environment. Specifically, let the standard deviation of the j-th observation component be... ,but In this lock calibration scenario, the standard deviation of each observed component is taken. For the corresponding measured value 15% of the absolute value.

[0074] By correcting each predicted member using Kalman gain, the analytical (posterior) state is obtained: ; in, The prior state represents the prediction of the system state (including parameters to be corrected and physical quantities) at time t+1 based on the information at time t and the structural simulation model. This prediction only considers the system's own evolution law and does not incorporate new observation data at time t+1. The posterior state represents the fusion of new observation data from time t+1. Then, the optimal estimate of the system state. The updated set. This is the posterior augmented state set at time t+1, and its mean is the optimal estimate of the system state at that time. After inverse transformation, the output is the optimal identification value of the physical parameters, which is used as the initial condition for the cyclic assimilation at the next time step.

[0075] After completing the state update at time t+1, if real-time observation data at time t+2 is accessed, the prediction and correction process described in step six is ​​repeated to ultimately achieve cyclic assimilation and dynamic parameter tracking.

[0076] Based on the above method embodiments, this application also provides a device for correcting simulation parameters of a lock structure, see [link to relevant documentation]. Figure 3 As shown, the device includes: a data preprocessing module 302, used to acquire historical displacement measurement data of the lock structure and form standardized measurement data through preprocessing; a parameter determination module 304, used to determine the parameters to be corrected based on the lock structure simulation model; an evolution iteration module 306, used to perform a global search on the parameters to be corrected using an adaptive evolution algorithm of the covariance matrix, and to obtain the distribution feature parameters that best match the displacement response prediction value output by the lock structure simulation model with the standardized measurement data through adaptive learning of the correlation between parameters and iterative optimization, so as to determine the initial parameter distribution of the parameters to be corrected; a model construction module 308, used to construct an augmented state space model based on the parameters to be corrected, the displacement response prediction value, and the initial parameter distribution; and a recursive assimilation module 310, used to recursively assimilate the augmented state variables in the augmented state space model using real-time lock displacement data as observations and a Kalman filter algorithm, so as to update and correct the parameters to be corrected.

[0077] Furthermore, the aforementioned data preprocessing module 302 is used to generate standardized measured data that meets the requirements of subsequent real-time assimilation through the following processing methods: preliminary screening of historical displacement measured data to remove duplicate and invalid records; for missing values ​​in historical displacement measured data, the adjacent mean filling method is used to fill in the missing values ​​according to the distribution characteristics; for multiple repeated observations of the same monitoring point on the same day in historical displacement measured data, the arithmetic mean is calculated as the unique characteristic value of that day; based on the preset assimilation step size, the horizontal displacement data and vertical displacement data in historical displacement measured data are resampled and time-aligned to ensure that different types of monitoring variables remain consistent on the time scale.

[0078] Furthermore, the aforementioned evolution iteration module 306 is used to set the initial state of the covariance matrix adaptive evolution algorithm, forming a multidimensional normal distribution. The initial state includes: the initial mean vector, initial step size, and initial covariance matrix of the parameters to be corrected, set according to the empirical values ​​of the lock design. For each iteration, the following sampling, evolution, filtering, and updating steps are performed: sampling from the multidimensional normal distribution to generate a sample set consisting of parameter vectors of a specified number of individuals; each parameter vector simultaneously contains candidate value combinations of all parameters to be corrected; each sample is sequentially substituted into the lock structure simulation model for evolution to obtain the corresponding physical field response prediction value; the objective function value corresponding to each sample is calculated based on the physical field response prediction value and the corresponding value in the standardized measured data; the objective function value corresponding to each sample is sorted according to the objective function value corresponding to each sample, and a specified number of high-quality samples are selected; the distribution characteristic parameters of the multidimensional normal distribution are updated using the high-quality samples; the distribution characteristic parameters include: mean vector, step size, and covariance matrix; until the convergence condition is met or the maximum number of iterations is reached, the initial parameter distribution of the parameters to be corrected is constructed based on the final distribution characteristic parameters of the multidimensional normal distribution.

[0079] Furthermore, the aforementioned model construction module 308 is used to construct an augmented state vector using the parameters to be corrected and the predicted displacement response as internal state variables; to construct an initial state set including multiple augmented state vectors using the Monte Carlo sampling method according to the initial parameter distribution; and to construct an augmented state space model based on the augmented state vectors, nonlinear state evolution functions, and process noise in the initial state set.

[0080] Furthermore, the aforementioned recursive assimilation module 310 is used to treat each vector in the current initial state set as the posterior augmented state vector of the previous time step, and perform the following prediction, statistics, gain calculation, and assimilation steps: based on the augmented state space model, it performs vector prediction on the posterior augmented state vector of the previous time step and performs statistics on the prediction results to determine the state prior bias matrix and displacement prediction bias matrix of the current time step; based on the state prior bias matrix and displacement prediction bias matrix of the current time step, it calculates the state-displacement cross-covariance of the current time step; based on the displacement prediction bias matrix of the current time step, it calculates the current... The displacement prediction autocovariance of the previous time step is calculated; based on the state-displacement cross-covariance, displacement prediction autocovariance, and the preset measurement error covariance matrix, the Kalman gain matrix of the current time step is calculated; the real-time value corresponding to the vector at the current time step is extracted from the real-time lock displacement data; based on the Kalman gain matrix of the current time step, the displacement prediction deviation of the real-time value at the current time step is corrected in real time to realize the update and correction of the parameter to be corrected; the updated and corrected posterior augmented state vector is used as the posterior augmented state vector of the previous time step, and the prediction, statistics, gain calculation, and assimilation steps are continued.

[0081] Furthermore, the aforementioned recursive assimilation module 310 is used to predict the posterior augmented state vector of the previous time step for each vector in the initial state set according to the augmented state space model, thereby determining the prior augmented state vector of the current time step; extract the displacement response from the prior augmented state vector of the current time step through the observation matrix, thereby determining the displacement prediction value of the current time step; calculate the mean of the prior augmented state vectors of the current time step corresponding to multiple vectors in the initial state set, thereby determining the augmented state prediction mean; calculate the mean of the displacement prediction value of the current time step based on the displacement prediction value of the current time step corresponding to multiple vectors in the initial state set, thereby determining the displacement prediction mean; calculate the state prior deviation matrix of the current time step based on the prior augmented state vectors and the augmented state prediction mean of the current time step corresponding to multiple vectors in the initial state set, thereby calculating the displacement prediction deviation matrix of the current time step based on the displacement prediction value and the displacement prediction mean of the current time step corresponding to multiple vectors in the initial state set, thereby calculating the displacement prediction deviation matrix of the current time step.

[0082] Furthermore, the aforementioned recursive assimilation module 310 is used to calculate the difference between the real-time value and the displacement prediction value corresponding to the vector; to obtain the product of the Kalman gain matrix at the current time and the difference; and to obtain the sum of the product and the prior augmented state vector at the current time to obtain the posterior augmented state vector at the current time.

[0083] The device provided in this application embodiment has the same implementation principle and technical effect as the aforementioned method embodiment. For the sake of brevity, any parts of the device embodiment not mentioned can be referred to the corresponding content in the aforementioned method embodiment.

[0084] This application also provides an electronic device, such as... Figure 4 The diagram shows the structure of the electronic device, which includes a processor 41 and a memory 40. The memory 40 stores computer-executable instructions that can be executed by the processor 41, and the processor 41 executes the computer-executable instructions to implement the above-described method.

[0085] exist Figure 4 In the illustrated embodiment, the electronic device further includes a bus 42 and a communication interface 43, wherein the processor 41, the communication interface 43, and the memory 40 are connected via the bus 42.

[0086] The memory 40 may include high-speed random access memory (RAM) and may also include non-volatile memory, such as at least one disk storage device. Communication between this system network element and at least one other network element is achieved through at least one communication interface 43 (which can be wired or wireless), such as the Internet, wide area network, local area network, metropolitan area network, etc. The bus 42 may be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus, etc. The bus 42 can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 4 The symbol is represented by a single double-headed arrow, but this does not mean that there is only one bus or one type of bus.

[0087] Processor 41 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuitry in the hardware of processor 41 or by instructions in software form. Processor 41 can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in the embodiments of this application can be directly implemented by a hardware decoding processor, or implemented by a combination of hardware and software modules in the decoding processor. The software modules can reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The storage medium is located in the memory, and the processor 41 reads the information in the memory and, in conjunction with its hardware, completes the steps of the method described in the foregoing embodiment.

[0088] This application also provides a computer-readable storage medium storing computer-executable instructions. When the computer-executable instructions are called and executed by a processor, the computer-executable instructions cause the processor to implement the above-described method. For specific implementation details, please refer to the foregoing method embodiments, which will not be repeated here.

[0089] The computer program products of the methods, apparatus, and electronic devices provided in the embodiments of this application include a computer-readable storage medium storing program code. The instructions included in the program code can be used to execute the methods described in the preceding method embodiments. For specific implementations, please refer to the method embodiments, which will not be repeated here.

[0090] Unless otherwise specifically stated, the relative steps, numerical expressions, and values ​​of the components and steps described in these embodiments do not limit the scope of this application.

[0091] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a processor-executable, non-volatile, computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0092] In the description of this application, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0093] Finally, it should be noted that the above-described embodiments are merely specific implementations of this application, used to illustrate the technical solutions of this application, and not to limit them. The protection scope of this application is not limited thereto. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments, or make equivalent substitutions for some of the technical features, within the technical scope disclosed in this application. Such modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be covered within the protection scope of this application. Therefore, the protection scope of this application should be determined by the protection scope of the claims.

Claims

1. A method of calibrating a simulation parameter of a ship lock structure, characterized by, The method includes: Historical displacement measurement data of the lock structure was obtained and standardized measurement data was generated through preprocessing. The parameters to be corrected are determined based on the simulation model of the lock structure. The covariance matrix adaptive evolution algorithm is used to perform a global search on the parameters to be corrected. By adaptively learning the correlation between parameters and iteratively optimizing, the distribution characteristic parameters that best match the displacement response prediction values ​​output by the lock structure simulation model with the standardized measured data are obtained. This determines the initial parameter distribution of the parameters to be corrected, including: setting the initial state of the covariance matrix adaptive evolution algorithm to form a multidimensional normal distribution; the initial state includes: the initial mean vector, initial step size, and initial covariance matrix of the parameters to be corrected, set according to lock design experience values; for each iteration, the following sampling, evolution, filtering, and updating steps are performed: sampling from the multidimensional normal distribution to generate a sample consisting of parameter vectors of a specified number of individuals. This set; each parameter vector simultaneously contains a combination of candidate values ​​for all the parameters to be corrected; each sample is sequentially substituted into the simulation model of the lock structure for evolution to obtain the corresponding physical field response prediction value; based on the physical field response prediction value and the corresponding value in the standardized measured data, the objective function value corresponding to each sample is calculated; the objective function values ​​corresponding to each sample are sorted, and a specified number of high-quality samples are selected; the distribution characteristic parameters of the multidimensional normal distribution are updated using the high-quality samples; the distribution characteristic parameters include: mean vector, step size, and covariance matrix; until the convergence condition is met or the maximum number of iterations is reached, the initial parameter distribution of the parameters to be corrected is constructed based on the final distribution characteristic parameters of the multidimensional normal distribution; Based on the parameters to be corrected, the predicted displacement response, and the initial parameter distribution, an augmented state-space model is constructed, including: constructing an augmented state vector using the parameters to be corrected and the predicted displacement response as internal state variables; constructing an initial state set including multiple augmented state vectors using Monte Carlo sampling according to the initial parameter distribution; and constructing an augmented state-space model based on the augmented state vectors, nonlinear state evolution functions, and process noise in the initial state set. Using real-time lock displacement data as observations, a Kalman filter algorithm is employed to recursively assimilate the augmented state variables in the augmented state space model, thereby updating and correcting the parameters to be corrected. This includes: using each vector in the current initial state set as the posterior augmented state vector of the previous time step, and performing the following prediction, statistics, gain calculation, and assimilation steps: based on the augmented state space model, performing vector prediction on the posterior augmented state vector of the previous time step and statistically analyzing the prediction results to determine the current time step's state prior deviation matrix and displacement prediction deviation matrix; calculating the current time step's state-displacement cross-covariance based on the current time step's state prior deviation matrix and displacement prediction deviation matrix; and... Based on the displacement prediction deviation matrix at the current moment, calculate the displacement prediction autocovariance at the current moment; based on the state-displacement cross-covariance, the displacement prediction autocovariance, and the preset measurement error covariance matrix at the current moment, calculate the Kalman gain matrix at the current moment; extract the real-time value corresponding to the vector at the current moment from the real-time lock displacement data; based on the Kalman gain matrix at the current moment, correct the displacement prediction deviation of the real-time value at the current moment in real time to update and correct the parameter to be corrected; use the updated and corrected posterior augmented state vector as the posterior augmented state vector of the previous moment, and continue to execute the prediction, statistics, gain calculation, and assimilation steps.

2. The method of claim 1, wherein, The steps for obtaining historical displacement measurement data of the lock structure and generating standardized measurement data through preprocessing include: Standardized measured data that meets the requirements for subsequent real-time assimilation can be generated through the following processing methods: The historical displacement measurement data were initially screened to remove duplicate and invalid records; For missing values ​​in the historical displacement measurement data, the adjacent mean filling method is used to fill in the missing values ​​based on the distribution characteristics. For the historical displacement measurement data, the arithmetic mean of multiple repeated observations of the same monitoring point on the same day is calculated as the unique characteristic value of that day; Based on a preset assimilation step size, the horizontal and vertical displacement data in the historical displacement measurement data are resampled and time-aligned to ensure that different types of monitoring variables remain consistent on the time scale.

3. The method according to claim 1, characterized in that, Based on the augmented state-space model, the steps of performing vector prediction on the posterior augmented state vector of the previous time step and statistically analyzing the prediction results to determine the state prior deviation matrix and displacement prediction deviation matrix at the current time step include: For each vector in the initial state set, the posterior augmented state vector of the previous time step is predicted according to the augmented state space model to determine the prior augmented state vector of the current time step; the displacement response is extracted from the prior augmented state vector of the current time step through the observation matrix to determine the displacement prediction value of the current time step. The mean of the augmented state prediction is determined by calculating the mean of the prior augmented state vectors corresponding to the current time of the multiple vectors in the initial state set; the mean of the displacement prediction is determined by calculating the mean of the displacement prediction values ​​corresponding to the current time of the multiple vectors in the initial state set. Based on the prior augmented state vectors corresponding to the multiple vectors in the initial state set at the current time and the mean of the augmented state predictions, calculate the state prior deviation matrix at the current time; based on the displacement prediction values ​​corresponding to the multiple vectors in the initial state set at the current time and the mean of the displacement predictions, calculate the displacement prediction deviation matrix at the current time.

4. The method according to claim 1, characterized in that, Based on the Kalman gain matrix at the current moment, the displacement prediction deviation of the real-time value at the current moment is corrected in real time to realize the update and correction of the parameter to be corrected, including: Calculate the difference between the real-time value and the predicted displacement value corresponding to the vector; Calculate the product of the Kalman gain matrix at the current time and the difference; The sum of the product and the prior augmented state vector at the current time is obtained to get the posterior augmented state vector at the current time.

5. A device for correcting simulation parameters of a ship lock structure, characterized in that, The device includes: The data preprocessing module is used to acquire historical displacement measurement data of the lock structure and to generate standardized measurement data through preprocessing. The parameter determination module is used to determine the parameters to be corrected based on the simulation model of the lock structure. The evolution iteration module is used to perform a global search on the parameters to be corrected using an adaptive evolution algorithm of the covariance matrix. Through adaptive learning of the correlation between parameters and iterative optimization, it obtains the distribution characteristic parameters that best match the displacement response prediction values ​​output by the lock structure simulation model with the standardized measured data, thereby determining the initial parameter distribution of the parameters to be corrected. This includes: setting the initial state of the adaptive evolution algorithm of the covariance matrix, forming a multidimensional normal distribution; the initial state includes: the initial mean vector, initial step size, and initial covariance matrix of the parameters to be corrected, set according to lock design experience values; for each iteration, the following sampling, evolution, filtering, and updating steps are performed: sampling from the multidimensional normal distribution to generate parameter vectors including a specified number of individuals. A sample set composed of parameters is used; each parameter vector simultaneously contains a combination of candidate values ​​for all parameters to be corrected; each sample is sequentially substituted into the simulation model of the lock structure for evolution to obtain the corresponding physical field response prediction value; based on the physical field response prediction value and the corresponding value in the standardized measured data, the objective function value corresponding to each sample is calculated; the objective function values ​​corresponding to each sample are sorted, and a specified number of high-quality samples are selected; the distribution characteristic parameters of the multidimensional normal distribution are updated using the high-quality samples; the distribution characteristic parameters include: mean vector, step size, and covariance matrix; until the convergence condition is met or the maximum number of iterations is reached, the initial parameter distribution of the parameters to be corrected is constructed based on the final distribution characteristic parameters of the multidimensional normal distribution; The model building module is used to construct an augmented state-space model based on the parameters to be corrected, the predicted displacement response values, and the initial parameter distribution. This includes: constructing augmented state vectors using the parameters to be corrected and the predicted displacement response values ​​as internal state variables; constructing an initial state set including multiple augmented state vectors using Monte Carlo sampling based on the initial parameter distribution; and constructing an augmented state-space model based on the augmented state vectors, nonlinear state evolution functions, and process noise in the initial state set. The recursive assimilation module uses real-time lock displacement data as observations and employs a Kalman filter algorithm to recursively assimilate the augmented state variables in the augmented state space model, thereby updating and correcting the parameters to be corrected. This includes: using each vector in the current initial state set as the posterior augmented state vector of the previous time step, and performing the following prediction, statistics, gain calculation, and assimilation steps: based on the augmented state space model, performing vector prediction on the posterior augmented state vector of the previous time step and statistically analyzing the prediction results to determine the state prior deviation matrix and displacement prediction deviation matrix at the current time step; and calculating the state-displacement mutual... Covariance; Calculate the displacement prediction autocovariance at the current moment based on the displacement prediction deviation matrix at the current moment; Calculate the Kalman gain matrix at the current moment based on the state-displacement cross-covariance, the displacement prediction autocovariance, and the preset measurement error covariance matrix; Extract the real-time value corresponding to the vector at the current moment from the real-time lock displacement data; Based on the Kalman gain matrix at the current moment, perform real-time correction on the displacement prediction deviation of the real-time value at the current moment to update and correct the parameter to be corrected; Use the updated and corrected posterior augmented state vector as the posterior augmented state vector at the previous moment, and continue to execute the prediction, statistics, gain calculation, and assimilation steps.

6. An electronic device, characterized in that, The method includes a processor and a memory, the memory storing computer-executable instructions executable by the processor, the processor executing the computer-executable instructions to implement the method of any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions that, when invoked and executed by a processor, cause the processor to perform the method described in any one of claims 1 to 4.

Images