Roadbed and pavement test and data collection system under mobile load

Abstract

The application discloses a kind of roadbed pavement test and data acquisition system under mobile load, comprising: collecting high-precision road surface roughness data, and inhibiting noise by space-time convolution variation auto-encoding network decomposition feature.Based on normal stiffness and morphological gradient, dynamic grid division is used to adjust sensor density, and multi-modal data is fused using Bayesian weight. Adaptive signal details are extracted using multi-scale transformation, and filter is designed to enhance damage signal and optimize structural entropy. A multi-field coupled grid control equation is constructed to realize dynamic topology optimization reconstruction. A damage evolution state space is established, and a prediction model is trained through a multi-stage reinforcement learning framework combined with experience replay. A digital twin mapping model is constructed, integrating Bayesian optimization and reinforcement learning to establish a virtual-real feedback channel for dynamic decision optimization. The application not only provides innovative improvements to traditional test methods and data acquisition systems, but also accurately simulates and analyzes complex nonlinear dynamic behavior.

Description

Technical Field

[0001] This invention relates to the field of roadbed and pavement testing technology, and in particular to a roadbed and pavement testing and data acquisition system under moving loads. Background Technology

[0002] With the rapid development of modern transportation, the vehicle loads on infrastructure such as highways and railways are becoming increasingly complex and variable. Especially under high-speed and heavy-load conditions, the subgrade and pavement structures are subjected to complex dynamic loads for extended periods, facing severe challenges to safety and durability. Different types of vehicles, with varying speeds, tire contact characteristics, and load distribution patterns, act on the pavement, causing stress redistribution and deformation accumulation in the subgrade and pavement materials, thus affecting their long-term service performance. Particularly in infrastructure such as highways, airport runways, and heavy-haul railways, the dynamic response induced by moving loads has a crucial impact on pavement lifespan and driving safety.

[0003] Traditional roadbed and pavement testing methods primarily rely on static load tests or simplified dynamic loading tests, which struggle to realistically reproduce the complex conditions of actual traffic environments and cannot comprehensively assess the dynamic characteristics of roadbeds and pavements under moving loads. Furthermore, current data acquisition systems still have limitations in spatial resolution, temporal synchronization, and multimodal fusion capabilities, resulting in insufficient accuracy in monitoring pavement deformation, stress response, and damage evolution. Therefore, there is an urgent need to develop a testing and data acquisition system capable of simulating real traffic loads, sensing roadbed and pavement conditions in real time, and accurately acquiring multidimensional data to comprehensively reveal the dynamic behavior of roadbeds and pavements under moving loads.

[0004] It is worth noting that roadbeds and pavements under moving loads may exhibit highly nonlinear dynamic characteristics, such as local buckling, overall instability, and fatigue cumulative damage. These nonlinear dynamic behaviors are influenced by various factors, including material properties, structural parameters, environmental factors, and load type. Their evolution mechanisms are complex and exhibit significant spatiotemporal inhomogeneity. Currently, related research still lacks systematic theories and methods, and the ability to quantitatively analyze and predict nonlinear dynamic behaviors is relatively limited. Therefore, how to accurately analyze and predict the nonlinear dynamic behavior of roadbeds and pavements under moving loads to improve the safety and durability of structures is a crucial technical problem that urgently needs to be solved. Summary of the Invention

[0005] To address the above problems, this invention provides a roadbed and pavement testing and data acquisition system under moving loads.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0007] A system for testing and acquiring data on subgrade and pavement under moving loads, including:

[0008] Moving load simulation module: Collects high-precision road surface unevenness data, and uses a spatiotemporal convolutional variational autoencoder network for feature decomposition and noise suppression, and then optimizes the load application path based on adaptive deformation gradient analysis;

[0009] Multimodal sensor data fusion module: Based on the normal stiffness distribution and road surface morphology gradient, the sensor deployment density is dynamically adjusted by adaptive meshing, multimodal sensor data is fused using Bayesian dynamic weight allocation, and noise is suppressed by neighborhood entropy constraints;

[0010] Nonlinear signal feature enhancement module: It uses multi-scale transformation combined with adaptive scale selection to extract signal detail components, enhances the damaged signal and suppresses noise through adaptive filter, and then optimizes the enhanced signal based on structural information entropy constraints to improve the identifiability of the damaged signal.

[0011] Dynamics Solving Module: Constructs adaptive mesh control equations driven by multi-field coupling, adjusts mesh density based on the coupling effects of stress field, velocity field, acceleration field and damage field, and reconstructs mesh using dynamic topology optimization method;

[0012] Damage prediction module: Establish a state space based on damage evolution, construct a reinforcement learning framework with a multi-stage learning strategy, and achieve adaptive optimization of the damage prediction strategy by training the model through experience playback and long-term discount factor.

[0013] Digital Twin Decision Module: Constructs a digital twin virtual-real mapping model, establishes a dynamic feedback channel between the physical system and the virtual simulation system, and dynamically adjusts the optimization parameters through an intelligent optimization feedback mechanism combined with reinforcement learning and Bayesian optimization to build an optimization control model based on twin decision-making and achieve dynamic optimization decision-making.

[0014] Furthermore: the moving load simulation module includes:

[0015] By utilizing multi-sensor fusion technology, laser point cloud scanning, inertial measurement units, and ultra-wideband radar work together to collect road surface height point cloud data, attitude changes of vehicle-mounted simulation devices, and non-uniform morphological interference information caused by changes in roadbed materials. The data is then fused to establish a continuous, seamless, and high-precision three-dimensional road surface morphology model.

[0016] A spatiotemporal convolutional variational autoencoder network is used to perform multi-layer feature decomposition on the collected data, and an adaptive noise suppression mechanism is used to improve data reliability, thereby providing high-precision road morphology data input for subsequent adaptive feedback adjustment.

[0017] Furthermore: the multimodal sensing data fusion module includes:

[0018] Adaptive mesh generation is adopted, and the spatial distribution density of the sensors is dynamically adjusted according to the normal stiffness gradient and the road surface morphology gradient to optimize the sensor layout, making the sensors denser in high-stress areas and reducing the sensor density in low-stress areas.

[0019] Based on Bayesian dynamic weight allocation, the data fusion weights are adaptively adjusted according to the signal-to-noise ratio and confidence level of different sensors to ensure the accuracy of the fused data;

[0020] By utilizing the neighborhood entropy constraint method, noise suppression is performed from both the temporal and spatial domains, thereby improving data quality.

[0021] Furthermore: the nonlinear signal feature enhancement module includes:

[0022] A multi-scale transform with adaptive scale selection is adopted. The signal is decomposed by wavelet transform, the energy distribution of wavelet coefficients is calculated, and the optimal scale is selected to preserve the characteristics of minor damage.

[0023] Signal enhancement based on adaptive filters dynamically adjusts filtering parameters by calculating local signal-to-noise ratio and filter gain function, enhancing damaged signals and suppressing irrelevant noise;

[0024] By utilizing signal optimization based on structural information entropy constraints, the local information entropy of the enhanced signal is calculated, and the signal is further optimized by constraining the weight function with information entropy.

[0025] Furthermore: the dynamics solution module includes:

[0026] An adaptive mesh control equation driven by multi-field coupling is constructed, which integrates stress field gradient, velocity field abrupt change, acceleration field rate of change and damage field characteristics to form the driving force for mesh adjustment;

[0027] Based on adaptive density function adjustment, the local mesh size is dynamically determined according to multi-field coupled control equations to achieve mesh refinement in stress concentration and damage-prone areas, as well as mesh sparsification in low-gradient areas.

[0028] Dynamic topology optimization is introduced, and the cell shape is adjusted through a mesh quality evaluation function to optimize the mesh topology, ensuring that the overall mesh quality is not degraded while local meshes are adjusted.

[0029] Furthermore: the damage prediction module includes:

[0030] A state space based on damage evolution is constructed, state variables are defined by combining stress, velocity, acceleration and damage characteristics, a damage evolution state vector is constructed, and a damage accumulation objective function is defined to optimize long-term prediction capability.

[0031] A reinforcement learning framework based on a multi-stage learning strategy is constructed, which divides the damage evolution stage into the initial micro-damage stage, the plastic damage development stage, and the fatigue failure stage. Corresponding learning strategies are designed for each stage, and the learning strategies are dynamically adjusted through a multi-stage strategy switching function.

[0032] We employ experience replay and long-term discount factor for reinforcement learning training, construct an experience pool, update the policy using the long-term discount factor, and optimize the damage prediction model parameters based on the policy gradient.

[0033] Furthermore: the digital twin decision-making module includes:

[0034] A digital twin virtual-real mapping model is constructed. Through a high-dimensional state mapping function, the stress, displacement, damage and environmental state of the physical system are dynamically synchronized with the virtual simulation system. The accuracy of the twin mapping is optimized based on an adaptive filter to ensure high-dimensional synchronization between the physical system and the virtual system.

[0035] An optimization feedback mechanism is constructed, an optimization objective function is defined, a reinforcement learning optimization strategy is combined, optimization parameters are dynamically adjusted, and Bayesian optimization method is used to achieve adaptive adjustment of parameters, ensuring the adaptability of decision optimization;

[0036] An optimal control model based on twin decision-making is constructed, a twin decision-making control strategy is defined, and model predictive control is adopted to optimize the control strategy and achieve dynamic optimal control.

[0037] The technological advancements achieved by this invention compared to existing technologies are as follows:

[0038] This invention employs an innovative intelligent moving load simulation algorithm (AFMLS) based on an adaptive feedback mechanism, which can simulate the spatial variability of vehicle loads and the interaction between the roadbed and tires in real time, solving the problem of fixed load paths and load distributions in traditional testing methods. This advantage enables the system to more accurately reproduce complex dynamic load conditions in real traffic environments, providing more realistic load inputs for analyzing the nonlinear dynamic behavior of roadbeds and pavements.

[0039] The system utilizes a hierarchical adaptive density optimization multimodal sensor data fusion algorithm to automatically adjust the sensor deployment density in different areas to cope with dynamic changes in complex stress fields. This technology enables the system to efficiently collect data from sensors of different locations and types, improving data quality and accuracy. This is crucial for understanding and predicting the dynamic behavior of roadbeds and pavements, especially under complex loads, allowing for real-time reflection of multi-dimensional states such as stress, strain, displacement, and damage.

[0040] The system employs a nonlinear signal feature enhancement module based on multi-scale transformation, which can extract low-frequency trend information and high-frequency local features under complex loads, enhancing the identifiability of minute damage signals. This solves the problem that existing methods easily lose minute damage information when processing the nonlinear behavior of roadbeds and pavements, providing more accurate signal processing capabilities for the analysis of nonlinear dynamic behavior.

[0041] By introducing a dynamic solution module based on multi-field coupling and adaptive mesh adjustment, this invention can automatically adjust the computational mesh in stress concentration regions, optimizing computational efficiency and reducing errors. This is particularly important for complex nonlinear dynamic problems, especially in the analysis of instability and buckling phenomena in roadbeds and pavements, where it can effectively reduce computational errors caused by uneven mesh partitioning and improve prediction accuracy.

[0042] The system integrates a damage prediction module based on multi-stage reinforcement learning, which can accurately predict the long-term damage evolution trend of roadbeds and pavements under moving loads through a dynamic optimization mechanism of reinforcement learning. This advantage is particularly suitable for long-term operational infrastructure, providing a scientific basis for pavement maintenance and repair under constantly changing traffic loads, and reducing the risk of sudden damage and destruction.

[0043] By combining a digital twin decision-making module with intelligent optimization feedback, this invention can adjust optimization decisions based on real-time data and feedback under different environmental conditions. This innovative feature solves the problem that traditional optimization methods are difficult to adapt to complex engineering environments in real time, providing flexible and efficient decision support. In particular, under complex traffic flow and load changes, it can optimize the maintenance strategy of roadbed and pavement, and improve structural safety and durability.

[0044] In summary, this invention not only provides a revolutionary improvement to traditional testing methods and data acquisition systems, but also enables accurate simulation and analysis of complex nonlinear dynamic behaviors. Through high-precision load simulation, intelligent data fusion, nonlinear feature extraction, efficient solution, and long-term damage prediction, this invention provides strong technical support for improving the safety, reliability, and durability of roadbed and pavement structures, and solves common problems in traditional methods such as insufficient accuracy, response lag, and prediction errors. Attached Figure Description

[0045] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof.

[0046] In the attached diagram:

[0047] Figure 1 This is a system structure diagram of the present invention. Detailed Implementation

[0048] The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of the present invention will now be described with reference to the accompanying drawings.

[0049] like Figure 1 As shown, this invention discloses a roadbed and pavement testing and data acquisition system under moving loads, comprising:

[0050] Moving load simulation module: Collects high-precision road surface unevenness data, and uses a spatiotemporal convolutional variational autoencoder network for feature decomposition and noise suppression, and then optimizes the load application path based on adaptive deformation gradient analysis;

[0051] Multimodal sensor data fusion module: Based on the normal stiffness distribution and road surface morphology gradient, the sensor deployment density is dynamically adjusted by adaptive meshing, multimodal sensor data is fused using Bayesian dynamic weight allocation, and noise is suppressed by neighborhood entropy constraints;

[0052] Nonlinear signal feature enhancement module: It uses multi-scale transformation combined with adaptive scale selection to extract signal detail components, enhances the damaged signal and suppresses noise through adaptive filter, and then optimizes the enhanced signal based on structural information entropy constraints to improve the identifiability of the damaged signal.

[0053] Dynamics Solving Module: Constructs adaptive mesh control equations driven by multi-field coupling, adjusts mesh density based on the coupling effects of stress field, velocity field, acceleration field and damage field, and reconstructs mesh using dynamic topology optimization method;

[0054] Damage prediction module: Establish a state space based on damage evolution, construct a reinforcement learning framework with a multi-stage learning strategy, and achieve adaptive optimization of the damage prediction strategy by training the model through experience playback and long-term discount factor.

[0055] Digital Twin Decision Module: Constructs a digital twin virtual-real mapping model, establishes a dynamic feedback channel between the physical system and the virtual simulation system, and dynamically adjusts the optimization parameters through an intelligent optimization feedback mechanism combined with reinforcement learning and Bayesian optimization to build an optimization control model based on twin decision-making and achieve dynamic optimization decision-making.

[0056] Specifically, the moving load simulation module based on environment-aware neurons (ESN) includes:

[0057] In the moving load simulation module, the Environment Sensing Neuron (ESN) is responsible for collecting the unevenness information of the road surface in real time and eliminating random noise through a dynamic filtering mechanism, thereby providing high-precision road morphology data input for subsequent adaptive feedback adjustment.

[0058] Module 1.1: High-precision road surface unevenness acquisition through multi-sensor fusion

[0059] Since traditional single laser or inertial sensors are insufficient to meet the requirements for sensing high-frequency, minute deformations, this invention employs laser point cloud scanning combined with an inertial measurement unit (IMU) and ultra-wideband radar (UWB-Radar) for data fusion to ensure an optimal combination of spatial accuracy, temporal synchronization, and noise suppression. The core objective of the data acquisition is to establish a continuous, seamless, and high-precision three-dimensional road surface morphology model as the basic input data for intelligent load simulation.

[0060] The specific implementation process includes:

[0061] 1. Set up N laser scanning sensors (LiDAR) evenly distributed at the bottom of the moving load simulation system to collect road surface height point cloud data P. lidar (x, y, z, t):

[0062] P lidar (x, y, z, t) = f lidar (λ, θ, d)

[0063] Where λ is the laser wavelength, θ is the scanning angle, and d is the echo ranging data.

[0064] 2. An inertial measurement unit (IMU) is used to calculate the attitude changes of the vehicle-mounted simulation device and obtain the gravitational acceleration 'a'. g and angular velocity ω g And combined with gyroscope data G gyro (t) Perform tilt compensation:

[0065]

[0066] 3. Combine ultra-wideband radar (UWB-Radar) for deep structure detection to obtain information on non-uniform morphological interference caused by changes in roadbed materials. uwb (x, y, t), and correct the laser point cloud data:

[0067] P corr (x, y, t) = P lidar (x, y, t) + S uwb (x, y, t)·w uwb

[0068] Among them, w uwb These are the weighting coefficients for the radar signal.

[0069] Module 1.2: Multilayer Feature Decomposition and Noise Suppression Based on Spatiotemporal Convolutional Variational Autoencoder (SCVAE)

[0070] Because road surface roughness data contains short-term high-frequency noise (vehicle vibration, environmental interference) and long-term topographic trend information (road construction errors, local damage), a single low-pass filter cannot effectively separate the real shape from the noise signal.

[0071] This invention employs a Spatio-Temporal Convolutional Variational Autoencoder (SCVAE) network to perform multi-level feature decomposition on the collected data and improves data reliability through an adaptive noise suppression mechanism.

[0072] The specific implementation process includes:

[0073] 1. Set the input spatiotemporal data matrix:

[0074] X input (x, y, t) = {P} corr (x, y, t), R corr (t)}

[0075] Among them, P corr (x, y, t) is the fused road point cloud data, R corr (t) represents the gyroscope data after attitude correction.

[0076] 2. Local road surface morphology features are extracted using a Spatiotemporal Convolutional Network (STCN):

[0077] H conv =σ(W stcn *X input +b)

[0078] Among them, W stcn σ is the spatiotemporal convolution kernel, and σ is the nonlinear activation function.

[0079] 3. High-dimensional feature reconstruction is performed using a variational autoencoder (VAE) to generate low-noise road surface morphology estimates:

[0080] Z vae ~q φ (Z|H conv )

[0081] Among them, Z vae For the latent space representation, q φ This is the encoder network for VAE.

[0082] 4. Optimization using KL divergence loss:

[0083] L kl =D KL (q φ (Z)||p(Z))

[0084] Where p(Z) is a standard normal distribution.

[0085] 5. Calculate the final denoised 3D road surface morphology:

[0086]

[0087] Module 1.3: Load Path Optimization Based on Adaptive Deformation Gradient Analysis (AGDA)

[0088] Since the normal stiffness of the tire-road contact area changes significantly during vehicle operation, this invention uses Adaptive Gradient Deformation Analysis (AGDA) to dynamically adjust the load application path and ensure that the contact state between the tire and the road surface is as close as possible to the actual working condition.

[0089] The specific implementation process includes:

[0090] 1. Calculate the normal stiffness gradient of the road surface:

[0091]

[0092] 2. Calculate the stress concentration factor in the load-bearing area:

[0093] S stress (x, y) = ∫ A G stiff (x′,y′)dx′dy′

[0094] 3. Employing a gradient descent optimization method, the load application path is dynamically adjusted to ensure uniform stress on the tires:

[0095]

[0096] Where α is the optimization step size.

[0097] Final output:

[0098] P filtered (x, y, t): 3D road surface morphology data after filtering and optimization.

[0099] P opt (x, y): Optimization results of load application path based on deformation gradient analysis.

[0100] G stiff (x, y): The calculated normal stiffness distribution, used for subsequent load adjustments.

[0101] This module uses key technologies such as multi-sensor data fusion, high-dimensional feature decomposition, and deformation gradient optimization to acquire high-precision road surface morphology data in real time and optimize the load application path, providing reliable input data for subsequent adaptive feedback adjustments.

[0102] Specifically, the multimodal sensing data fusion module based on hierarchical adaptive density optimization includes:

[0103] After the moving load simulation module completes the real-time acquisition and dynamic filtering of road surface roughness data based on the environment-aware neuron (ESN), this embodiment has obtained high-precision three-dimensional road surface morphology data P after noise reduction and optimization. filtered (x, y, t) and the corresponding normal stiffness distribution G stiff (x, y), these data provide the necessary input information for the multimodal sensing data fusion module.

[0104] In this module, this embodiment designs a hierarchical adaptive density optimization strategy to dynamically adjust the deployment density of multimodal sensors based on the stress distribution and morphological change characteristics of different regions, thereby improving the accuracy and robustness of data fusion.

[0105] Module 2.1: Partition Density Optimization Based on Adaptive Mesh Generation (AGP-DO)

[0106] Because the stress distribution of the roadbed-pavement system is highly non-uniform, if a fixed density of sensors is used, it will lead to a waste of sensor resources in low-stress areas and insufficient sensor data in high-stress concentration areas.

[0107] This invention employs Adaptive Grid Partitioning (AGP), based on the normal stiffness gradient G. stiff (x, y) and road surface morphology gradient By dynamically adjusting the spatial distribution density of sensors, a denser sensor deployment is achieved in high-stress areas, while the sensor density is reduced in low-stress areas, thereby improving measurement accuracy and resource utilization efficiency.

[0108] The specific implementation process includes:

[0109] 1. Calculate the rate of change of normal stiffness to identify stress concentration regions:

[0110]

[0111] 2. Calculate the pavement morphological gradient modulus to identify regions of abrupt geometric changes:

[0112]

[0113] 3. Define the sensor density optimization function, where D s (x, y) represents the optimal sensor deployment density at a given coordinate location:

[0114]

[0115] Among them, D min and D max These represent the minimum and maximum sensor deployment densities, respectively.

[0116] 4. Employ a dynamic mesh refinement strategy:

[0117] If D s (x, y) > D threshold This refines the grid and increases the sensor density in that area.

[0118] If D s (x, y) <D threshold If so, the grid is merged to reduce the sensor density in that area.

[0119] Module 2.2: Multimodal Data Fusion Based on Bayesian Dynamic Weight Allocation (BDW-Fusion)

[0120] Since different sensors (such as strain gauges, accelerometers, fiber Bragg gratings, ultrasonic ranging, etc.) have different measurement accuracies and noise characteristics, the core challenge of this module is how to dynamically adjust the weights of different data sources according to the actual working conditions to ensure the accuracy of data fusion.

[0121] This invention employs Bayesian Dynamic Weighting (BDW), which adaptively adjusts the data fusion weights of different sensors based on the signal-to-noise ratio (SNR) and confidence level of real-time measurement data.

[0122] The specific implementation process includes:

[0123] 1. Set up data sets for different sensors:

[0124] S = {S1, S2, ..., S} N}

[0125] Among them, S i This represents the measurement data from the i-th sensor.

[0126] 2. Calculate the signal-to-noise ratio (SNR) for each sensor:

[0127]

[0128] 3. Calculate the uncertainty of the measurement data:

[0129]

[0130] Where λ is the adjustment coefficient, SNR th This is the signal-to-noise ratio threshold.

[0131] 4. Calculate the Bayesian dynamic weight assignment:

[0132]

[0133] 5. Perform weighted fusion to obtain the final optimized data:

[0134]

[0135] Module 2.3: Noise Suppression Based on Neighborhood Entropy Constraints (NEC-Denoise)

[0136] Since random noise inevitably exists in sensor data, this embodiment adopts the Neighborhood Entropy Constraint (NEC) method to suppress noise from both spatiotemporal perspectives and improve data quality.

[0137] The specific implementation process includes:

[0138] 1. Calculate the local neighborhood entropy of the data:

[0139]

[0140] Where, p i This represents the probability distribution of data points within this neighborhood.

[0141] 2. Define the noise suppression weighting function:

[0142]

[0143] 3. Perform noise suppression processing:

[0144] S denoise (x, y) = W denoise (x, y)·S fused (x, y)

[0145] Final output:

[0146] D s (x, y): The dynamically optimized sensor deployment density distribution matrix.

[0147] S fused (x, y, t): Fusion data based on Bayesian dynamic weight allocation.

[0148] Sdenoise (x, y, t): High-precision sensor data optimized by neighborhood entropy constraint noise reduction.

[0149] This module achieves dynamic optimization of sensor deployment density through adaptive mesh generation, Bayesian dynamic weight allocation, and neighborhood entropy-constrained noise suppression techniques, thereby improving the accuracy and robustness of data fusion and laying a solid foundation for subsequent nonlinear signal feature enhancement and dynamic solution.

[0150] Specifically, the nonlinear signal feature enhancement module based on multi-scale transformation includes:

[0151] After the multimodal sensing data fusion module completes the hierarchical adaptive density-optimized multimodal sensing data fusion algorithm, this embodiment has obtained high-precision sensing data S after noise reduction optimization. denoise The data (x, y, t) contains the dynamic response information of the subgrade and pavement under moving loads. However, due to the complexity of the subgrade-pavement system under nonlinear dynamics, many subtle damage signals are masked by background noise or lost due to the limitations of traditional filters. Therefore, the core objective of this module is to enhance the characteristics of nonlinear signals using a multi-scale transform domain adaptive filter, thereby improving the identifiability of subtle damage signals.

[0152] Module 3.1: Multiscale Transform Based on Adaptive Scale Selection (MST-ASS)

[0153] Since damage signals typically manifest as local high-frequency abnormal fluctuations, while the low-frequency components of structural response are mainly affected by overall stiffness and damping characteristics, single-scale signal analysis methods cannot simultaneously capture low-frequency trends and high-frequency local features.

[0154] This invention employs Multi-Scale Transform (MST) to extract detailed components of the signal at different scales, and introduces an Adaptive Scale Selection (ASS) mechanism to ensure the selection of the optimal scale, thereby preserving the features of minor damage.

[0155] The specific implementation process includes:

[0156] 1. Wavelet transform is used to decompose signal S denoise (x, y):

[0157]

[0158] in:

[0159] W j (x, y, t) represents the wavelet detail coefficients (high-frequency components) of the j-th layer, reflecting damage characteristic information;

[0160] R J (x, y, t) are the wavelet approximation coefficients (low-frequency components) of the J-th layer, representing the overall trend.

[0161] 2. Calculate the energy distribution of the wavelet coefficients to screen for the optimal scale:

[0162]

[0163] Among them, E j This represents the proportion of wavelet energy at the j-th scale.

[0164] 3. Define the adaptive scaling weight function:

[0165]

[0166] Where, α j The weight representing the importance of this scale.

[0167] 4. Select the optimal scale j * :

[0168]

[0169] The detail factor corresponding to this scale This will be used for subsequent feature enhancement processing.

[0170] Module 3.2: Signal Enhancement Based on Adaptive Filters (AFSE)

[0171] Since minute damage signals are often affected by environmental noise, directly extracting high-frequency components may lead to misjudgment. Therefore, an adaptive filter (AFSE) is needed to enhance the damage signal and suppress irrelevant noise.

[0172] This filter dynamically adjusts its filtering parameters through statistical analysis of local time-frequency characteristics, thereby ensuring that it can effectively extract damage signals under different operating conditions.

[0173] The specific implementation process includes:

[0174] 1. Calculate the local signal-to-noise ratio (L-SNR):

[0175]

[0176] in:

[0177] This is the local average amplitude;

[0178] σ local(x, y, t) represents the local standard deviation.

[0179] 2. Define the filter gain function:

[0180]

[0181] in:

[0182] β is an adjustment parameter;

[0183] SNR th This is the signal-to-noise ratio threshold.

[0184] 3. Enhanced damage signals:

[0185]

[0186] This module can effectively suppress noise in low signal-to-noise ratio regions while enhancing the identifiability of damaged signals.

[0187] Module 3.3: Signal Optimization Based on Structural Information Entropy Constraints (SIE-0)

[0188] Since damage signals often exhibit local abrupt changes, the enhanced signal can be further optimized by calculating the structural information entropy of the signal, making it clearer and more identifiable.

[0189] The specific implementation process includes:

[0190] 1. Calculate the local information entropy of the enhanced signal:

[0191]

[0192] Where, p i This represents the normalized probability distribution of the signal amplitude.

[0193] 2. Define the information entropy constraint weight function:

[0194]

[0195] 3. Optimize and enhance the signal:

[0196]

[0197] Final output:

[0198] S enhanced (x, y, t): The enhanced signal after multi-scale transformation, adaptive filtering and structural information entropy optimization. This signal can more clearly reflect the characteristics of minor damage and suppress background noise.

[0199] This module employs techniques such as multi-scale transformation, adaptive filtering, and structural information entropy optimization to simultaneously preserve both low-frequency trend information and high-frequency local damage characteristics of the signal, significantly improving the identifiability of minute damage signals and providing high-quality data input for subsequent nonlinear dynamic behavior identification and damage prediction.

[0200] Specifically, the dynamic solution module based on multi-field coupling adaptive mesh adjustment includes:

[0201] Following the nonlinear signal feature enhancement module: a nonlinear signal feature enhancement module based on multi-scale transformation, this embodiment has obtained the dynamic response signal S after nonlinear feature enhancement. enhanced The equation (x, y, t) contains information on the nonlinear dynamic behavior of the subgrade-pavement system under moving loads. However, traditional dynamic solution methods typically employ fixed mesh generation, which leads to large errors in stress concentration regions and low computational efficiency in low stress gradient regions. Therefore, the core objective of this module is to introduce a multi-field coupled adaptive mesh adjustment method, enabling the mesh generation to adapt to changes in dynamic characteristics in real time, thereby ensuring an optimal balance between computational accuracy and efficiency.

[0202] Module 4.1: Constructing Adaptive Mesh Control Equations Driven by Multi-Field Coupling (MAC-CF)

[0203] When the roadbed-pavement system is subjected to moving loads, nonlinear coupling effects occur between the stress field, velocity field, acceleration field, and damage field. Therefore, it is difficult to guarantee the global calculation accuracy by relying solely on information from a single physical field to adjust the mesh generation.

[0204] This invention integrates stress field gradient, velocity field abrupt change, acceleration field rate of change, and damage field characteristics into a mesh adjustment driving force to construct a multi-field coupled control equation (MAC-CF).

[0205] The specific implementation process includes:

[0206] 1. Define multi-field variables:

[0207] Stress field: σ(x, y, t)

[0208] Velocity field: v(x, y, t)

[0209] Acceleration field: a(x, y, t)

[0210] Damage field (calculated by the nonlinear signal feature enhancement module): D(x, y, t)

[0211] 2. Calculate the stress gradient field:

[0212]

[0213] Among them, G σ (x, y, t) reflects the degree of drastic change in local stress.

[0214] 3. Calculate the region of abrupt velocity change:

[0215]

[0216] This indicator is used to identify the area affected by inertial forces under moving loads.

[0217] 4. Calculate the rate of change of acceleration:

[0218]

[0219] This parameter measures drastic changes in the acceleration field, reflecting the region of impact loading or nonlinear dynamic phenomena (such as buckling and instability).

[0220] 5. Calculate the damage field gradient:

[0221]

[0222] This feature is used to identify potential damage propagation areas, enabling the mesh to be automatically refined in areas of concentrated damage.

[0223] 6. Construct the multi-field coupled control equations (MAC-CF):

[0224] G MAC (x, y, t) = w σ G σ (x, y, t) + w v G σ G v (x, y, t) + w a G a (x, y, t) + w D G D (x, y, t)

[0225] in:

[0226] w σ w v w a w D The weighting coefficients for different physical fields can be determined through experimental data or numerical optimization.

[0227] G MAC (x, y, t) serves as the driving force for adaptive mesh adjustment, controlling the fineness of mesh generation.

[0228] Module 4.2: Adaptive Density Function-Based Mesh Adjustment (ADF-MG)

[0229] Traditional mesh generation methods typically employ uniform or single-variable-based densification strategies. However, in this problem, the mesh density in different regions should be determined according to the multi-field governing equations G. MAC (x, y, t) are dynamically adjusted.

[0230] Therefore, this embodiment uses an adaptive density function (ADF) to determine the local mesh size.

[0231] The specific implementation process includes:

[0232] 1. Define the mesh density function:

[0233]

[0234] in:

[0235] ρ min and ρ max These are the minimum and maximum grid densities, respectively.

[0236] and These are the global minimum and maximum multi-field coupling control values, respectively;

[0237] This formula ensures that the mesh is automatically refined in regions of stress concentration, significant damage, abrupt velocity changes, and drastic acceleration changes, while the mesh density is reduced in low gradient regions.

[0238] 2. Calculate the mesh size distribution:

[0239]

[0240] This formula determines the local mesh size h(x, y, t). The smaller the value, the denser the mesh, and the larger the value, the sparser the mesh.

[0241] Module 4.3: Dynamic Topology Optimization-Based Mesh Reconstruction (DTO-MG)

[0242] Traditional finite element mesh adjustment methods mainly rely on fixed rules or manual adjustment, which makes it difficult to respond in real time to dynamically changing load and stress distributions.

[0243] This invention introduces the Dynamic Topology Optimization (DTO) method, which ensures that the overall mesh quality is not degraded while adjusting the local mesh.

[0244] The specific implementation process includes:

[0245] 1. Adjusting element shape based on mesh quality evaluation function:

[0246]

[0247] in:

[0248] A is the area of ​​the triangular unit;

[0249] l i Let be the side length of the triangular unit;

[0250] Q(x, y, t) is between 0 and 1. The closer the value is to 1, the better the mesh quality.

[0251] 2. Mesh Reconstruction Strategy:

[0252] If Q(x,y,t) th (Set a threshold) to locally optimize the cell shape to reduce distortion;

[0253] If the mesh is overly refined and the computation time exceeds the limit, the density function ρ(x, y, t) can be adjusted by incorporating error estimation to reduce the computational burden.

[0254] Final output:

[0255] h(x, y, t): Adaptively adjusted mesh size distribution;

[0256] The optimized mesh topology ensures computational accuracy while improving computational efficiency.

[0257] Minimizing the dynamic solution error provides more accurate input data for subsequent nonlinear behavior modeling.

[0258] This module employs an adaptive mesh adjustment strategy driven by multi-field coupling to automatically refine the mesh in regions characterized by high stress gradients, abrupt velocity changes, acceleration variations, and concentrated damage, while adaptively reducing mesh density in low gradient regions. This ensures an optimal balance between computational accuracy and efficiency, providing a solid numerical foundation for subsequent nonlinear dynamic behavior identification and prediction.

[0259] Specifically, the damage prediction module based on multi-stage reinforcement learning includes:

[0260] Following the dynamics solution module—specifically, the dynamics solution module based on multi-field coupling and adaptive mesh adjustment—this embodiment has obtained high-precision dynamic stress field, velocity field, acceleration field, and damage field distribution data, and ensured computational efficiency through an adaptive mesh optimization strategy. However, traditional damage prediction methods typically rely on static machine learning models or regression analysis based on finite samples, making it difficult to accurately simulate the damage accumulation and evolution trends under long-term loading.

[0261] ​The core objective of this module is to introduce the Multi-Stage Reinforcement Learning (MSRL) method, enabling damage prediction models to dynamically optimize their learning strategies at different stages, thereby improving the accuracy of long-term predictions.

[0262] Module 5.1: Constructing a Damage Evolution State Space (ESS)

[0263] Traditional reinforcement learning methods face the problems of high-dimensional state space sparsity and decision path instability in complex dynamic systems, making it difficult to effectively learn long-term damage evolution patterns.

[0264] The specific implementation process includes: This module introduces the Damage Evolution State Space (DESS) and combines stress, velocity, acceleration and damage characteristics to construct a state representation of reinforcement learning in order to optimize long-term prediction capabilities.

[0265] 1. Define the damage evolution state variables:

[0266] Stress state: S σ (t) = {σ(x, y, t)}

[0267] Speed ​​status: S v (t) = {v(x, y, t)}

[0268] Acceleration state: S a (t) = {a(x, y, t)}

[0269] Damage status: S D (t) = {D(x, y, t)}

[0270] 2. Construct the damage evolution state vector:

[0271] S(t)=[S σ (t), S v (t), S a (t), S D (t)]

[0272] Where S(t) represents the system damage state at the current time step t.

[0273] 3. Define the damage accumulation objective function:

[0274]

[0275] in:

[0276] R(t) represents the reinforcement learning reward function representing the degree of damage accumulation;

[0277] α σ α v α a α D These are the weighting coefficients for each physical field;

[0278] This objective function ensures that reinforcement learning can accurately capture the cumulative effect of damage, thereby optimizing long-term predictive capabilities.

[0279] Module 5.2: Constructing a Reinforcement Learning Framework (MSRL) Based on a Multi-Stage Learning Strategy

[0280] Traditional reinforcement learning methods employ a single policy update mechanism, which is difficult to adapt to the evolutionary characteristics of different damage stages.

[0281] This invention employs a phased optimization strategy, dynamically adjusting the learning strategy at different stages to improve the model's long-term predictive ability.

[0282] The specific implementation process includes:

[0283] 1. Divide the damage evolution stages:

[0284] Initial micro-damage phase (Elastic Deformation Phase): Damage variable D(x, y, t) changes slowly and is mainly controlled by low-frequency loads.

[0285] Plastic Damage Phase: The damage variable DD(x, y, t) begins to grow rapidly, and the system enters a nonlinear damage accumulation process.

[0286] Fatigue Failure Phase: The damage variable D(x, y, t) approaches the threshold D. crit The structure has entered a critical failure state.

[0287] 2. Construct a multi-stage policy switching function:

[0288]

[0289] in:

[0290] π1(t) is the learning strategy for the micro-damage stage, which mainly optimizes the linear damage prediction model;

[0291] π2(t) represents the learning strategy for the plastic damage development stage, and deep reinforcement learning is used to optimize the damage prediction capability.

[0292] π3(t) is the learning strategy for the fatigue failure stage, which optimizes the accuracy of failure prediction and ensures accurate identification of critical damage states.

[0293] Module 5.3: Reinforcement Learning Training Based on Experience Replay and Long-Term Discount Factor

[0294] Because damage evolution is time-dependent, traditional reinforcement learning methods struggle to effectively utilize long-term historical data for training.

[0295] This invention introduces experience replay and long-term discount factor to ensure that the damage prediction model can capture the cumulative damage effect under long-term loading.

[0296] The specific implementation process includes:

[0297] 1. Build an experience pool:

[0298]

[0299] in:

[0300] S(t) represents the current damage state;

[0301] A(t) represents the prediction strategy adopted by the current reinforcement learning model;

[0302] R(t) is the damage accumulation objective function at the current time step;

[0303] S(t+1) represents the damage state at the next time step.

[0304] 2. Adopt a long-term discount factor update strategy:

[0305]

[0306] in:

[0307] γ is the long-term discount factor (usually set to 0.9-0.99) to ensure that the model can focus on long-term damage effects;

[0308] Q(S t A t ) is the Q-learning value function.

[0309] 3. Damage prediction model based on policy gradient optimization:

[0310]

[0311] in:

[0312] θ represents the parameters of the reinforcement learning model;

[0313] η is the learning rate;

[0314] This optimization strategy ensures that reinforcement learning can dynamically adjust the prediction strategy, thereby improving the accuracy of damage prediction.

[0315] Final output:

[0316] D pred (x, y, t): Long-term damage prediction results;

[0317] A damage evolution model optimized by reinforcement learning ensures long-term prediction accuracy;

[0318] The adaptive strategy optimization mechanism enables the model to automatically adjust its learning strategy at different stages.

[0319] This module achieves adaptive optimization of the damage prediction strategy through a multi-stage reinforcement learning method, enabling the damage evolution model to dynamically adjust the learning strategy, improve long-term prediction accuracy, and provide more accurate and reliable data support for nonlinear dynamic analysis, health monitoring, and life prediction of subgrade-pavement systems.

[0320] Specifically, the digital twin decision-making module based on intelligent optimization feedback includes:

[0321] In the damage prediction module: This embodiment has established a long-term damage prediction model and optimized the prediction accuracy of damage evolution through a multi-stage reinforcement learning strategy. However, traditional optimization methods lack a dynamic feedback mechanism, making it difficult to adjust the optimization strategy in real time to adapt to complex engineering environments, leading to decision lag or inability to adapt to unexpected situations. Therefore, this module introduces an Intelligent Optimization Feedback (IOF) mechanism to construct a Digital Twin (DT) decision framework, achieving dynamic decision optimization based on real-time perception, intelligent feedback, and adaptive optimization.

[0322] Module 6.1: Constructing a Digital Twin Mapping (DTM) Model

[0323] Traditional digital twin methods are mostly based on static models or predefined parameter optimization, which makes it difficult to adapt to the complex nonlinear characteristics under dynamic loads.

[0324] This module introduces a high-dimensional state mapping function to establish a dynamic feedback channel between the physical system (PS) and the virtual simulation system (VS), ensuring the real-time performance and accuracy of the twin model.

[0325] The specific implementation process includes:

[0326] 1. Define the state variables of the physical system:

[0327] Stress state: S σ (t)

[0328] Displacement state: S u (t)

[0329] Damage status: S D (t)

[0330] Environmental status: S E (t)

[0331] 2. Define the digital twin mapping function:

[0332]

[0333] in:

[0334] S σ (t), S u (t), S D (t), S E (t) represents the real-time measurement value of the physical system;

[0335] The calculated value for the digital twin system;

[0336] This mapping function ensures high-dimensional synchronization between the physical and virtual systems.

[0337] 3. Optimizing Siamese Mapping Accuracy Based on Adaptive Filters:

[0338]

[0339] in:

[0340] λ is an adaptive weighting factor that ensures the twin system can quickly correct errors;

[0341] This optimization strategy can effectively improve the dynamic adaptability of digital twin models.

[0342] Module 6.2: Building an Intelligent Optimization Feedback Mechanism (IOF)

[0343] Traditional optimization methods are based on fixed rules or single-step optimization strategies, lacking real-time feedback capabilities and making it difficult to adapt to complex environmental changes.

[0344] This module constructs a real-time optimization adjustment strategy based on reinforcement learning through the Intelligent Optimization Feedback (I0F) mechanism to ensure the adaptability of decision optimization.

[0345] The specific implementation process includes:

[0346] 1. Define the optimization objective function:

[0347]

[0348] in:

[0349] J opt To optimize the objective;

[0350] Θ represents the set of optimization parameters;

[0351] ω σ ω D These are the weighting factors for each state variable;

[0352] The objective function ensures that the optimization strategy can adjust the damage prediction bias and stress response error in real time.

[0353] 2. Constructing reinforcement learning optimization strategies:

[0354]

[0355] in:

[0356] π IOF (t) represents the reinforcement learning strategy for intelligent optimization feedback;

[0357] R IOF (t) represents the immediate reward for optimizing the strategy;

[0358] This optimization strategy improves system adaptability by dynamically adjusting optimization parameters.

[0359] 3. Dynamically adjust optimization parameters based on Bayesian optimization:

[0360]

[0361] in:

[0362] By using Bayesian optimization methods, we can ensure that the optimization parameters are adaptively adjusted at different damage stages, thereby improving the accuracy of decision-making.

[0363] Module 6.3: Constructing an Optimal Control Model Based on Twin Decision Making

[0364] Traditional decision-making methods employ offline optimization strategies, making it difficult to achieve real-time optimization.

[0365] This module achieves dynamic optimization control through digital twins, reinforcement learning, and intelligent feedback mechanisms.

[0366] The specific implementation process includes:

[0367] 1. Define the twin decision control strategy:

[0368]

[0369] in:

[0370] u DT (t) represents the control variables for twin decision-making;

[0371] K DT The optimal gain matrix for the twin system;

[0372] This control strategy can effectively reduce the error between the twin system and the physical system.

[0373] 2. Optimize the control strategy based on MPC (Model Predictive Control):

[0374]

[0375] in:

[0376] By employing the MPC method, we can ensure that the control strategy can be adaptively adjusted at different stages, thereby optimizing the overall decision-making effect.

[0377] Final output:

[0378] u DT (t): Control decisions based on digital twin optimization;

[0379] The Intelligent Optimization Feedback (IOF) mechanism ensures that the optimization strategy can be adjusted in real time.

[0380] An optimization control model based on reinforcement learning improves decision-making accuracy and adaptability.

[0381] This module constructs a high-precision dynamic optimization decision-making system through a digital twin decision-making module based on intelligent optimization feedback. This ensures that the damage prediction model can adjust the optimization strategy in real time to adapt to complex environmental changes and improve the long-term operational safety and durability of the load-subgrade-pavement system.

[0382] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the claims of the present invention.

Claims

1. A system for testing and data collection of roadbed and pavement under moving load, characterized in that, The method comprises the following steps: Mobile load simulation module: using multi-sensor fusion technology, through the cooperative work of laser point cloud scanning, inertial measurement unit and ultra-wideband radar, collecting road height point cloud data, vehicle load simulation device attitude change and non-uniform form interference information caused by roadbed material change, and carrying out fusion processing on the data, establishing a continuous, seamless and high-precision three-dimensional road surface form model; Adopting a spatio-temporal convolutional variational auto-encoding network to perform multi-layer feature decomposition on the collected data, and improving the data reliability through an adaptive noise suppression mechanism, thereby providing high-precision road surface form data input for subsequent adaptive feedback adjustment; Multi-modal sensor data fusion module: according to the normal stiffness distribution and the road surface form gradient, the sensor distribution density is dynamically adjusted by adaptive grid division, the multi-modal sensor data is fused by using Bayesian dynamic weight allocation, and noise suppression is performed through neighborhood entropy constraint; Nonlinear signal feature enhancement module: multi-scale transformation is used in combination with adaptive scale selection to extract signal detail components, an adaptive filter is used to enhance damage signals and suppress noise, and the enhanced signals are optimized based on structural information entropy constraint to improve the recognizability of damage signals; Dynamics solving module: an adaptive grid control equation driven by multi-field coupling is constructed, the grid density is adjusted according to the coupling effect of stress field, velocity field, acceleration field and damage field, and the grid is reconstructed by using dynamic topology optimization method; Damage prediction module: a state space based on damage evolution is constructed, state variables are defined in combination with stress, velocity, acceleration and damage characteristics, a damage evolution state vector is constructed, and a damage accumulation objective function is defined to optimize the long-term prediction capability; A reinforcement learning framework based on multi-stage learning strategy is constructed, the damage evolution stages are divided, including initial micro-damage stage, plastic damage development stage and fatigue failure stage, corresponding learning strategies are designed respectively, and the learning strategies are dynamically adjusted by using multi-stage strategy switching function; Empirical playback and long-term discount factor are used for reinforcement learning training, an experience pool is constructed, the strategy is updated by using long-term discount factor, and the damage prediction model parameters are optimized based on strategy gradient; Digital twin decision module: a digital twin virtual-real mapping model is constructed, a dynamic feedback channel between the physical system and the virtual simulation system is established, an intelligent optimization feedback mechanism is used in combination with reinforcement learning and Bayesian optimization to dynamically adjust and optimize parameters, an optimization control model based on twin decision is constructed, and dynamic optimization decision is realized.

2. The system for testing and data acquisition of road bed and pavement under moving load according to claim 1, characterized in that, The multi-modal sensor data fusion module comprises: Adaptive grid division is used to dynamically adjust the spatial distribution density of sensors according to the normal stiffness gradient and the road surface form gradient, and the sensor arrangement is optimized, so that the sensor density in the high stress area is higher and the sensor density in the low stress area is lower; Based on Bayesian dynamic weight allocation, the data fusion weight is adaptively adjusted according to the signal-to-noise ratio and confidence of different sensors to ensure the accuracy of the fused data; The neighborhood entropy constraint method is used to suppress noise from the time and space domains to improve data quality.

3. The system for testing and data acquisition of road bed and pavement under moving load according to claim 2, characterized in that, The nonlinear signal feature enhancement module comprises: Adopting the multi-scale transform of adaptive scale selection, the signal is decomposed by wavelet transform, the energy distribution of wavelet coefficients is calculated, and the optimal scale is screened to retain the characteristics of micro-damage. Based on the signal enhancement of adaptive filter, the filter parameters are dynamically adjusted through local signal-to-noise ratio calculation and filter gain function to enhance the damage signal and suppress irrelevant noise. Signal optimization based on structural information entropy constraint is used to calculate the local information entropy of the enhanced signal and further optimize the signal through the information entropy constraint weight function.

4. The system for testing and data collection of road bed and pavement under moving load according to claim 3, characterized in that, The kinetic solving module comprises: An adaptive mesh control equation driven by multi-field coupling is constructed to form the driving force of mesh adjustment by comprehensively considering the stress field gradient, velocity field mutation, acceleration field change rate and damage field characteristics; Based on the adaptive density function, the local mesh size is dynamically determined according to the multi-field coupling control equation to realize the mesh refinement in the stress concentration and damage significant area and the mesh sparsification in the low gradient area; Dynamic topology optimization is introduced to adjust the unit shape through the mesh quality evaluation function, optimize the mesh topology structure and ensure that the overall mesh quality is not deteriorated while the local mesh is adjusted.

5. The system for testing and data collection of road bed and pavement under moving load according to claim 4, characterized in that, The digital twin decision module comprises: A digital twin virtual-real mapping model is constructed to dynamically synchronize the stress, displacement, damage and environmental state of the physical system with the virtual simulation system through the high-dimensional state mapping function, and the mapping precision is optimized based on the adaptive filter to ensure the high-dimensional synchronization of the physical system and the virtual system; An optimization feedback mechanism is constructed, the optimization objective function is defined, the optimization parameters are dynamically adjusted by combining the reinforcement learning optimization strategy, and the adaptive adjustment of the parameters is realized through the Bayesian optimization method to ensure the adaptability of the decision optimization; An optimization control model based on twin decision is constructed, the twin decision control strategy is defined, the model predictive control optimization control strategy is adopted, and the dynamic optimization control is realized.

Images