A soft ground uneven settlement self-adaptive compensation control method

Abstract

This invention discloses an adaptive compensation control method for uneven settlement in soft strata, belonging to the field of settlement control in geotechnical engineering. The method involves collecting distributed micro-strain field and pore water pressure gradient data at the bottom of the foundation in the soft strata and inputting them into a stratum rheological evolution prediction model. Multi-scale spatiotemporal features are extracted using a spatiotemporal attention coding module to generate a stratum creep trend vector. The equivalent modulus correction coefficient is calculated using a virtual stiffness reconstruction algorithm to generate a dynamic stress wave excitation parameter set. Coherent stress beams are emitted using microarray actuators embedded in the bearing layer interface, and a dynamic stress-locking zone is constructed through constructive interference to form a virtual rigid support. The excitation parameters are corrected through a dual-loop adaptive feedback control mechanism combined with a heuristic optimization algorithm until the micro-strain field returns to uniformity. Simultaneously, energy dissipation monitoring and safety mechanisms are added to mitigate the risk of soil liquefaction. This invention achieves active suppression and adaptive compensation for uneven settlement in soft strata, improving the accuracy of settlement control.

Description

Technical Field

[0001] This invention relates to the field of settlement control technology in geotechnical engineering, specifically to an adaptive compensation control method for uneven settlement in soft strata. Background Technology

[0002] Due to the inherent characteristics of soft soil, such as strong cohesive plasticity, easy accumulation of pore water pressure, and poor structural stability, it is prone to uneven settlement under the influence of factors such as building foundation loads, groundwater seepage, and external environmental disturbances. This has become a key technical problem that urgently needs to be solved in the field of geotechnical engineering. Uneven settlement will cause additional stress and deformation in the foundation structure, which may lead to cracking of building walls and damage to components, or even foundation instability and overall structural tilting, seriously threatening the safety and service life of the project. Traditional treatment methods are mainly divided into two categories: passive protection and active reinforcement. Passive protection resists settlement deformation by increasing the stiffness of the foundation and setting settlement joints, but it suffers from high material consumption, high cost, and poor adaptability. Active reinforcement improves soil properties through grouting, deep mixing, etc., but its construction cycle is long, it causes great disturbance to the strata, and it is difficult to achieve real-time dynamic control of the settlement process. Faced with soft soil with complex rheological characteristics, it is difficult to achieve the ideal settlement control effect. Meanwhile, traditional methods rely heavily on single-point monitoring data for settlement assessment, which cannot accurately capture the spatiotemporal evolution of the micro-strain field in the strata. They are prone to failure of compensation measures due to prediction lag and are difficult to adapt to the dynamic changes in creep settlement of soft strata.

[0003] With the development of intelligent and precise geotechnical engineering, settlement prediction and adaptive compensation technologies based on monitoring data have gradually become research hotspots, and various sensor monitoring and intelligent algorithms have begun to be applied to the field of ground settlement control. However, the settlement evolution of soft strata is a complex process involving the coupling of multiple physical fields, including the dynamic interaction of various factors such as soil microstrain, pore water pressure, and rheological properties. Existing numerical simulation and machine learning models struggle to incorporate physical constraints such as soil constitutive relations and fluid dynamics equations into the prediction process, resulting in weak model generalization ability, low prediction accuracy, and an inability to accurately predict ground creep trends. Furthermore, existing settlement compensation technologies often employ single stiffness enhancement methods, lacking the ability to dynamically reconstruct the equivalent modulus of the soil in the target compensation area, making it difficult to adjust compensation strategies in real time according to settlement evolution characteristics. In addition, feedback control mechanisms are mostly single-loop controls, lacking sufficient ability to suppress high-frequency disturbances, and the parameter adjustment efficiency of optimization algorithms is low, failing to achieve rapid convergence of microstrain field homogenization, thus restricting the accuracy and timeliness of uneven settlement control in soft strata.

[0004] Chinese patent (publication number CN108661187B) discloses an adaptive support for uneven foundation settlement and its working method. This technology uses an adaptive support consisting of movable tie rods, an inner cylinder, an outer cylinder, and springs, installed between the frame column base and the rigid foundation. The automatic elongation of the support structure eliminates uneven settlement of the frame structure due to the foundation. Compared to traditional passive protection methods, it is simple in structure, easy to maintain, and can achieve adaptive compensation for settlement, effectively reducing the impact of foundation deformation on the superstructure. However, this technology only passively adapts to the connection between the superstructure and the foundation, without actively reinforcing the soil itself. It cannot fundamentally suppress creep settlement in weak strata and can only adapt to small-amplitude settlement deformations. Its compensation capacity is insufficient when facing large-gradient, dynamically evolving uneven settlement. Furthermore, it lacks real-time monitoring and parameter control of the settlement process, limiting its adaptability and accuracy. Chinese patent (publication number CN111156034A) discloses a pre-embedded grouting unit and a settlement control system and method for soft strata. This technology reinforces soft strata through grouting using pre-embedded grouting units, achieving active control of strata settlement. Compared to the aforementioned support compensation technology, it improves the overall stiffness of soft strata from the perspective of soil improvement, thus solving to some extent the problem that passive adaptation cannot suppress strata creep. However, this technology relies on the solidification effect of the grouting material and is a static reinforcement method. It cannot adjust the reinforcement range and intensity in real time according to the dynamic evolution of strata settlement. Furthermore, the grouting process is prone to causing local disturbances in the strata, which may lead to a sudden increase in pore water pressure and pose a risk of soil liquefaction. At the same time, it lacks real-time feedback and parameter correction of the reinforcement effect, making it difficult to achieve adaptive and precise compensation for settlement. Summary of the Invention

[0005] Based on the above-mentioned technical problems, this application discloses an adaptive compensation control method for uneven settlement in soft strata, specifically including:

[0006] Data on distributed micro-strain field and pore water pressure gradient at the bottom of the foundation of the weak stratum are obtained, and the data are input into a stratum rheological evolution prediction model constructed based on a physical information neural network.

[0007] Based on the spatiotemporal attention encoding module in the formation rheological evolution prediction model, multi-scale spatiotemporal feature extraction is performed on the micro-strain field data to generate a formation creep trend vector;

[0008] By using the virtual stiffness reconstruction algorithm embedded in the formation rheological evolution prediction model, the equivalent modulus correction coefficient of the target compensation area is calculated based on the formation creep trend vector, and a dynamic stress wave excitation parameter set is generated.

[0009] Using a micro-array actuator embedded in the foundation bearing layer interface, a coherent stress beam with a specific frequency, phase and amplitude is emitted according to the dynamic stress wave excitation parameter set. The dynamic stress locking zone is instantaneously constructed in the soft soil layer by utilizing the constructive interference principle of waves, thereby increasing the equivalent dynamic elastic modulus of the soil in the target area to form a virtual rigid support effect.

[0010] A dual-loop adaptive feedback control mechanism is adopted to monitor the residual of the micro-strain field under the action of the dynamic stress locking zone in real time. The dynamic stress wave excitation parameter set is corrected online through a heuristic optimization algorithm until the micro-strain field distribution is restored to a uniform state, thereby realizing the active suppression and adaptive compensation of uneven settlement of soft strata.

[0011] Preferably, the formation rheological evolution prediction model is constructed using a physical information neural network, specifically as follows:

[0012] The Navier-Stokes equations and soil constitutive relations are embedded as physical constraints into the loss function of a neural network to construct a physical information loss term. The physical information loss item The calculation formula is: ,in, Indicates the first Micro-strain at each sampling point Indicates time, Indicates dependence on porosity The diffusion coefficient tensor, Indicates pore water pressure, Represents the soil viscoplastic rheological operator; by minimizing the data-driven loss term and the physical information loss term. The weighted sum of these factors is used to train the formation rheological evolution prediction model.

[0013] Preferably, the spatiotemporal attention encoding module is specifically used for:

[0014] The input micro-strain field data is decomposed in the spatial domain and sliced ​​in the time domain using a sliding window to generate spatial feature maps and time series segments.

[0015] Calculate the association weight matrix between grid nodes in the spatial feature map using a multi-head self-attention mechanism. And use causal convolutional networks to compute the temporal dependency weight vectors of time series segments. ;

[0016] The correlation weight matrix With time-dependent weight vector Kronecker product operation is performed to generate spatiotemporal joint attention mask, and the original micro-strain field data is weighted and fused to output a formation creep trend vector containing high-order spatiotemporal coupling features.

[0017] Preferably, the specific steps for calculating the equivalent modulus correction coefficient using the virtual stiffness reconstruction algorithm include:

[0018] Based on the formation creep trend vector, the main subsidence axis with the largest subsidence rate gradient is identified; several sub-compensation units are divided along the main subsidence axis, and the target equivalent dynamic elastic modulus is calculated for each sub-compensation unit. The target equivalent dynamic elastic modulus The calculation formula is: ,in, The initial static elastic modulus distribution, The sedimentation rate gradient vector. This is the stiffness enhancement gain coefficient. Gradient-sensitive threshold, The rheological compensation factor decays over time; based on the target equivalent dynamic elastic modulus. The difference between the measured equivalent modulus and the actual equivalent modulus is used to invert the equivalent modulus correction factor to obtain the required equivalent modulus correction factor.

[0019] Preferably, the dynamic stress wave excitation parameter set includes a center frequency. Initial phase and amplitude envelope ;

[0020] The center frequency The determination method is as follows: analyze the particle size distribution curve and pore structure characteristics of the weak strata, and calculate the characteristic resonance frequency of the soil skeleton. The center frequency Set to slightly below the characteristic resonant frequency The sub-resonance frequency band, namely ,in This is the sub-resonance frequency adjustment coefficient, used to prevent soil liquefaction and maximize the interparticle frictional interlocking effect.

[0021] Preferably, the process of the microarray actuator emitting a coherent stress beam includes:

[0022] The microarray actuator is divided into several phased array sub-apertures, each sub-aperture containing... A separate piezoelectric ceramic actuator unit;

[0023] Based on the dynamic stress wave excitation parameter set, a delay control is applied to each actuator unit so that the spherical waves emitted by all sub-apertures undergo constructive interference at the focal point of the target compensation region.

[0024] The synthetic wave field intensity at the focal point The formula is: ,in, For the first The amplitude of each unit For the first Phase offset of each unit For wavenumber vectors, For the first The distance from each unit to the focal point; through dynamic adjustment The dynamic stress-locking band is controlled to move in shape and position in three-dimensional space.

[0025] Preferably, the formation mechanism of the dynamic stress-locking band is as follows:

[0026] Under the action of the coherent stress beam, the soft soil particles generate high-frequency micro-amplitude vibrations, resulting in periodic fluctuations in the normal stress at the particle contact surface. By utilizing the non-monotonic friction law in particle mechanics, the effective normal stress between particles is increased during the compression half-cycle of the vibration cycle, and the contact is maintained without separation through the inertial effect during the tension half-cycle.

[0027] Asymmetric contact mechanics behavior leads to a significant increase in the tangent modulus of the soil macroscopically, thereby inducing a macroscopically ordered force chain network reinforcement zone in the microscopically disordered soil structure, forming a dynamic stress-locked zone.

[0028] Preferably, the dual-loop adaptive feedback control mechanism includes an inner loop velocity control loop and an outer loop displacement shape control loop;

[0029] The inner loop speed control loop uses micro-strain rate As a feedback quantity, a sliding mode variable structure controller is used to quickly suppress the sudden change in strain rate caused by high-frequency disturbances;

[0030] The outer ring displacement mode control loop uses the spatial uniformity index of the micro-strain field distribution. As the feedback quantity, a model predictive controller is used to continuously optimize the dynamic stress wave excitation parameter set in the future time domain; the spatial uniformity index Defined as the Gini coefficient of the micro-strain field gradient, it is used to quantify the degree of non-uniformity in settlement.

[0031] Preferably, the heuristic optimization algorithm is an improved behavioral particle swarm optimization algorithm; in the behavioral particle swarm optimization algorithm, the particle position update equation introduces a potential well shrinkage factor. Similar to the random observation state collapse mechanism; the particle position update formula is: ,in, For the first During the nth iteration The particle in the first The parameter value of the dimension. It is a random weighted average of the individual's historical best position and the global best position. The location of the population centroid. for Uniformly distributed random numbers within the interval; the frequency bandwidth and phase delay matrix of the dynamic stress wave excitation parameter set are dynamically adjusted through the position update formula to minimize the norm of the micro-strain field residual.

[0032] Preferably, the method further includes an energy dissipation monitoring step:

[0033] The ratio of the acoustic power density radiated by the microarray actuator to the soil to the mechanical power density absorbed by the soil is calculated in real time, and the energy coupling efficiency coefficient is defined. When the rate of increase in pore water pressure in a local area exceeds a preset threshold, and the energy coupling efficiency coefficient... When an abnormally steep drop occurs, it is determined that the soil is about to enter the critical liquefaction state; a safety mechanism is immediately triggered, forcibly switching the frequency of the coherent stress beam to the non-resonant dissipation frequency band, and gradually reducing the amplitude envelope. The beam is reduced to zero, and the reverse phase cancellation beam is activated to quickly dissipate the formed dynamic stress locking zone and prevent the foundation from becoming unstable.

[0034] Compared with the prior art, the technical solution of this application has the following technical effects:

[0035] This invention accurately acquires micro-strain field and pore water pressure gradient data at the bottom of soft soil foundations. Based on a prediction model constructed using a physical information neural network, and combined with a spatiotemporal attention encoding module, it effectively extracts multi-scale spatiotemporal features and accurately generates a formation creep trend vector. Simultaneously, through an embedded virtual stiffness reconstruction algorithm, it accurately calculates the equivalent modulus correction coefficient, generating a suitable dynamic stress wave excitation parameter set, providing precise parameter support for subsequent settlement compensation.

[0036] This invention utilizes a microarray actuator embedded in the foundation bearing layer interface to emit coherent stress beams based on a dynamic stress wave excitation parameter set. By employing the principle of wave constructive interference, a dynamic stress locking zone is instantaneously constructed within the soft soil layer. This effectively enhances the equivalent dynamic elastic modulus of the soil in the target area, forming a stable virtual rigid support effect and achieving effective suppression of uneven settlement from the soil itself.

[0037] This invention employs a dual-loop adaptive feedback control mechanism, which can monitor the residual of the micro-strain field under the action of dynamic stress locking band in real time. Combined with an improved heuristic optimization algorithm, it can quickly correct the dynamic stress wave excitation parameter set online, realize the rapid suppression of micro-strain rate mutations and the precise optimization of settlement space uniformity, until the micro-strain field distribution is restored to uniformity, thus achieving adaptive compensation for settlement.

[0038] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the preferred embodiments of this application are described in detail below with reference to the accompanying drawings.

[0039] The above and other objects, advantages and features of this application will become more apparent to those skilled in the art from the following detailed description of specific embodiments in conjunction with the accompanying drawings. Attached Figure Description

[0040] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the 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 based on these drawings without creative effort. In all drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0041] Based on the description of the figures and their corresponding technical content in the document, the titles of the figures are as follows:

[0042] Figure 1 A schematic diagram of the overall process of the adaptive compensation control method for uneven settlement in soft strata;

[0043] Figure 2 Schematic diagram of the architecture of a formation rheological evolution prediction model based on physical information neural networks;

[0044] Figure 3 A schematic diagram of the multi-scale spatiotemporal feature extraction process of the spatiotemporal attention coding module;

[0045] Figure 4 Schematic diagram of the overall architecture of dual-loop adaptive feedback control and heuristic parameter optimization;

[0046] Figure 5 Spatiotemporal evolution characteristics of microstrain field in weak strata and cloud map of compensation effect;

[0047] Figure 6 A comparison of the spatiotemporal variations of maximum settlement and settlement rate under different settlement control methods;

[0048] Figure 7 Schematic diagram of the three-dimensional morphology and spatial distribution characteristics of dynamic stress-locking bands;

[0049] Figure 8 Spatial distribution cloud map of equivalent dynamic elastic modulus of soil in the target area of ​​weak strata;

[0050] Figure 9 Spatiotemporal evolution coupling analysis curves of multiple core indicators in the settlement compensation process. Detailed Implementation

[0051] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. In the following description, specific details such as specific configurations and components are provided merely to help fully understand the embodiments of this application. Therefore, those skilled in the art should understand that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this application. In addition, for clarity and brevity, descriptions of known functions and structures are omitted in the embodiments.

[0052] It should be understood that the phrase "an embodiment" or "this embodiment" throughout the specification means that a specific feature, structure, or characteristic related to the embodiment is included in at least one embodiment of this application. Therefore, "an embodiment" or "this embodiment" appearing throughout the specification does not necessarily refer to the same embodiment. Furthermore, these specific features, structures, or characteristics can be combined in any suitable manner in one or more embodiments.

[0053] Furthermore, reference numerals and / or letters may be repeated in different examples within this application. Such repetition is for the purpose of simplification and clarity and does not in itself indicate a relationship between the various embodiments and / or settings discussed.

[0054] In this article, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can mean: A exists alone, B exists alone, and A and B exist simultaneously. The term " / and" in this article describes another type of relationship between related objects, indicating that two relationships can exist. For example, A / and B can mean: A exists alone, and A and B exist alone. In addition, the character " / " in this article generally indicates that the related objects before and after it are in an "or" relationship.

[0055] In this article, the term "at least one" is merely a description of the relationship between related objects, indicating that there can be three relationships. For example, "at least one of A and B" can mean: A exists alone, A and B exist simultaneously, or B exists alone.

[0056] It should also be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion.

[0057] Example 1

[0058] This embodiment mainly describes an adaptive compensation control method for uneven settlement in soft strata, such as... Figure 1 As shown, it specifically includes:

[0059] Data on distributed micro-strain field and pore water pressure gradient at the bottom of the foundation of the weak stratum are obtained, and the data are input into a stratum rheological evolution prediction model constructed based on a physical information neural network.

[0060] Based on the spatiotemporal attention encoding module in the formation rheological evolution prediction model, multi-scale spatiotemporal feature extraction is performed on the micro-strain field data to generate a formation creep trend vector;

[0061] By using the virtual stiffness reconstruction algorithm embedded in the formation rheological evolution prediction model, the equivalent modulus correction coefficient of the target compensation area is calculated based on the formation creep trend vector, and a dynamic stress wave excitation parameter set is generated.

[0062] Using a micro-array actuator embedded in the foundation bearing layer interface, a coherent stress beam with a specific frequency, phase and amplitude is emitted according to the dynamic stress wave excitation parameter set. The dynamic stress locking zone is instantaneously constructed in the soft soil layer by utilizing the constructive interference principle of waves, thereby increasing the equivalent dynamic elastic modulus of the soil in the target area to form a virtual rigid support effect.

[0063] A dual-loop adaptive feedback control mechanism is adopted to monitor the residual of the micro-strain field under the action of the dynamic stress locking zone in real time. The dynamic stress wave excitation parameter set is corrected online through a heuristic optimization algorithm until the micro-strain field distribution is restored to a uniform state, thereby realizing the active suppression and adaptive compensation of uneven settlement of soft strata.

[0064] Furthermore, a distributed fiber optic sensing array and a pore water pressure sensing array are deployed at the bottom of the foundation in a three-dimensional spatial grid pattern. The fiber optic sensing array is arranged with a spatial grid spacing of 0.5m × 0.5m to achieve the acquisition of distributed micro-strain field data over the entire foundation bottom. The acquired data is in three-dimensional spatial coordinates. With time Quadruple functions The micro-strain value of each spatial grid node at different times is represented; the pore water pressure sensing array is arranged with a spatial grid spacing of 1m×1m, and the pore water pressure value of each node is collected by the pore water pressure sensor. And based on the spatial difference method, the pore water pressure gradient data were calculated. The system completes the full-domain, continuous, and dynamic acquisition of multiphysics field data. After standardization, all acquired data is divided into training and testing sets in an 8:2 ratio and input into a geological rheological evolution prediction model based on Physical Information Neural Network (PINN).

[0065] like Figure 2 As shown, the formation rheological evolution prediction model is based on a deep fully connected neural network. The input layer dimension is the feature dimension of the collected data, including 7 features: micro-strain, pore water pressure, spatial components of pore water pressure gradient, and spatiotemporal coordinates. There are 6 hidden layers, with the number of neurons in each layer distributed in a gradient of 256-512-1024-1024-512-256. The output layer contains features related to formation creep trends, with the dimension being the preset feature vector dimension. The core of the model construction is to embed the physical laws of geotechnical engineering as constraints into the loss function. Specifically, the Navier-Stokes equations and the viscoplastic constitutive relation of soil are used as physical constraints to construct a physical information loss term. The formula is: ,in, This represents the total number of sampling points in the training set. For the first Measured microstrain values ​​at each sampling point For the first The rate of change of microstrain over time at each sampling point is determined by the central difference method. Calculations show that This refers to the data acquisition time step. For depends on porosity The second-order diffusion coefficient tensor, ,in They are respectively The diffusion coefficient in the directional direction is obtained by fitting the soil porosity. ( , (These are inherent parameters of the soil). The tensor correction divergence term for pore water pressure is expanded as follows: ; The soil viscoplastic rheological operator is characterized using the Bingham viscoplastic model. ,in For soil shear modulus, The viscosity coefficient of the soil. The yield stress of the soil was determined by laboratory geotechnical tests. It is the square of the Euclidean second norm, used to quantify the deviation between the predicted value and the derived value of the micro-strain time change rate.

[0066] The model's total loss function For data-driven loss terms With physical information loss item The weighted sum, i.e. ,in For the weighting coefficients, satisfying Based on the requirements for data fitting accuracy and conformity with physical laws, it is set as follows: , Data-driven loss term The mean squared error loss is calculated using the following formula: ,in For the model to the first The model is trained using the Adam optimizer with a learning rate of 1e-4 and 10,000 iterations. The validation set loss is calculated every 100 iterations, and training stops when the validation set loss no longer decreases after 500 consecutive iterations. A test set is used to validate the model's performance and ensure its prediction accuracy. This yields a formation rheological evolution prediction model that combines data fitting ability with conformity to physical laws, and can accurately characterize the spatiotemporal evolution of microstrain field and pore water pressure field in weak formations.

[0067] Furthermore, such as Figure 3 As shown, a spatiotemporal attention coding module is embedded in the hidden layer of the stratigraphic rheological evolution prediction model. This module is the core feature extraction unit of the model, using standardized micro-strain field data. Using spatial domain decomposition, temporal domain slicing, spatiotemporal attention weight calculation, and weighted fusion as input, multi-scale spatiotemporal feature extraction is completed through four steps to generate a fixed-dimensional stratigraphic creep trend vector. The specific execution process is as follows:

[0068] Multi-scale spatial domain decomposition: decomposing three-dimensional micro-strain field data The grid was divided into three scale layers based on spatial scale: fine scale (original grid size 0.5m × 0.5m), medium scale (grid size 1.0m × 1.0m), and coarse scale (grid size 2.0m × 2.0m). Spatial feature maps at each scale were obtained through spatial downsampling. , , ,in The feature maps are respectively in The dimension of direction, and , , This enables the extraction of multi-scale spatial features of micro-strain fields;

[0069] Time-domain sliding window slicing: micro-strain time series for each spatial grid node Perform sliding window slicing and set the window length. Step length Generate several consecutive time series segments ( , (Total number of time segments), and to avoid future information leakage, the time series is causally truncated to ensure that each time segment contains only micro-strain data from historical moments;

[0070] Spatiotemporal attention weight calculation: For the spatial feature map at each scale, a multi-head self-attention mechanism is used (number of heads...) Calculate the spatial correlation weight matrix for the spatial feature map. ( Its spatial correlation weight matrix ( The element of the total number of spatial grid nodes at this scale is calculated as follows: ,in The first The query vector and key vector of each spatial node are obtained from the feature map through a linear transformation. ( (as the key vector dimension); for time series segments Using causal convolutional networks (kernel size) Temporal features are extracted using a step size of 1 and padding of 1, and temporal dependency weight vectors are obtained by mapping through a single-layer perceptron. Its elements Characterizing the first The temporal contribution of each time step to the current moment; then the spatial correlation weight matrix is ​​used. With time-dependent weight vector Perform Kronecker product operation to generate spatiotemporal joint attention mask. The dimension of the mask is The element value represents the characteristic importance of each spatiotemporal node;

[0071] Feature weighted fusion and vector generation: utilizing spatiotemporal joint attention masks The original micro-strain field data were weighted and fused to calculate the result. ( (The spatial index corresponds to the spatiotemporal node), and the fused multi-scale spatiotemporal features are concatenated to obtain the fused feature matrix. ( This represents the total number of spatial nodes. (This represents the total number of time steps). Finally, through global average pooling and linear transformation, the fused feature matrix is ​​mapped to a fixed dimension. The one-dimensional eigenvector, i.e., the formation creep trend vector. Each element of this vector represents a high-order spatiotemporal coupling feature of the micro-strain field, fully encompassing the spatial evolution law and temporal development trend of creep and subsidence in soft strata.

[0072] Furthermore, a virtual stiffness reconstruction algorithm is embedded in the formation rheological evolution prediction model. This algorithm uses the formation creep trend vector... Using the soil's physical and mechanical parameters as the core input, and combining these parameters, the process involves five steps: settlement characteristic identification, sub-compensation unit division, target equivalent dynamic elastic modulus calculation, equivalent modulus correction coefficient calculation, and excitation parameter set generation. This completes the mapping from creep trend to stress wave excitation parameters. The specific calculation process is as follows:

[0073] Settlement feature identification: vectorizing the formation creep trend The data is input into a multilayer perceptron, and the settlement rate distribution of the soft strata is obtained through feature decoupling. The settlement rate gradient vector was calculated based on the spatial gradient. Its modulus Characterizing the spatial variation of settlement rate; identifying the global maximum value of the modulus. The corresponding spatial curve is the main settlement axis. This axis represents the core evolution direction of uneven subsidence in soft strata and is a key area for subsequent compensation control.

[0074] Sub-compensation unit division: along the main settlement axis By spatial length By dividing the data into equidistant parts, several interconnected three-dimensional sub-compensation units are obtained. The spatial range of each sub-compensation unit is centered on the axis node, extending outwards... A cubic region extending 0.5m in each direction covers the main settlement axis and the surrounding settlement influence area. Each sub-compensation unit serves as an independent stiffness compensation and stress wave excitation unit.

[0075] Calculation of the target equivalent dynamic elastic modulus: for each sub-compensation unit According to its spatial coordinates With time Calculate the target equivalent dynamic elastic modulus The formula is:

[0076]

[0077] in, The initial static elastic modulus distribution of the sub-compensation unit was determined by a combination of indoor geotechnical tests and field plate load tests, and the coordinates are shown in the spatial coordinates. Two-dimensional functions; The settling rate gradient magnitude of the sub-compensation unit; The stiffness enhancement gain coefficient is set according to the soil viscoplasticity level; for soft plastic soil, it is taken as... Flowing plastic soil ; The gradient-sensitive threshold is determined by the global mean of the sedimentation rate gradient. ; The rheological compensation factor decays over time and is characterized by an exponential decay model. ,in The rheological attenuation coefficient is determined by soil creep tests, and its value ranges from [value range missing]. ; It is a natural exponential calculation used to achieve a nonlinear mapping between the settlement rate gradient and the stiffness enhancement magnitude, so that the larger the settlement gradient, the greater the stiffness enhancement magnitude in the region.

[0078] Equivalent modulus correction coefficient calculation: First, the measured equivalent dynamic elastic modulus of each sub-compensation unit is measured on-site using the stress wave reflection method. Calculate the difference between the target value and the measured value. Then, the equivalent modulus correction factor is defined. The ratio of the target value to the measured value, i.e. This coefficient is a dimensionless quantity, representing the stiffness increase factor required for each sub-compensation unit. This indicates a need to increase soil stiffness. When the soil stiffness meets the requirements, This indicates that no stiffness enhancement is required.

[0079] Generation of dynamic stress wave excitation parameter set: using equivalent modulus correction coefficient Based on the core criteria, and combining the particle size distribution curve, pore structure characteristics, and physical parameters of the soil skeleton in the soft strata, a dynamic stress wave excitation parameter set is calculated and generated. This parameter set consists of the core control parameters of the microarray actuator, including the center frequency. Initial phase Amplitude envelope The three core parameters are calculated as follows:

[0080] Center frequency First, the effective particle size is calculated using the soil particle size distribution curve. Average particle size Combined with porosity Calculate the characteristic resonant frequencies of the soil skeleton The formula is ,in The shear wave velocity in the soil was determined by in-situ wave velocity tests; subsequently, the center frequency was set to the sub-resonance band. ,in This is the subresonant frequency adjustment coefficient, with a value of [value missing]. To prevent soil liquefaction and maximize the interparticle frictional interlocking effect;

[0081] initial phase Based on the spatial location of the sub-compensation units, a phased array phase design method is adopted to focus the stress wave in the target compensation area, with the initial phase... Distribute evenly according to spatial angles;

[0082] Amplitude envelope : with equivalent modulus correction factor Linear correlation, the formula is ,in The base amplitude is determined by the power parameters of the microarray actuator. For the amplitude modulation frequency, take .

[0083] The generated dynamic stress wave excitation parameter set Each sub-compensation unit is matched with a unique excitation parameter to achieve spatially differentiated control of stress wave excitation.

[0084] Furthermore, a microarray actuator is embedded at a predetermined location at the interface between the bearing layer and the underlying weak stratum (the boundary between the bearing layer and the underlying weak stratum). This actuator is a piezoelectric ceramic array actuation device, composed of several phased array sub-apertures, each sub-aperture being [missing information - likely a size value]. rectangular array ( All are positive integers, set according to the size of the compensation area. or Each array unit is an independent piezoelectric ceramic actuator, which can achieve independent control of frequency, phase and amplitude. The embedding depth and position of the microarray actuator correspond to the sub-compensation unit to ensure that the stress wave can act accurately on the target compensation area.

[0085] The microarray actuator is based on the dynamic stress wave excitation parameter set The core of transmitting a coherent stress wave beam is to achieve constructive interference of waves and spatial focusing of stress wave energy. The specific transmission process is as follows: First, the dynamic stress wave excitation parameter set is... The process is broken down to each piezoelectric ceramic actuator unit, and an independent delay control is applied to each unit. ( ; The delay time is determined by the spatial distance between the target focus and the element. ,in For the first The spatial distance from each unit to the target focus The longitudinal wave velocity in the soil is given by the delay control mechanism, which ensures that the spherical stress waves emitted by all actuator units are superimposed in phase at the focal point of the target compensation region, achieving constructive interference. The combined wave field intensity at the focal point is... The calculation formula is: ,in, For the first The real-time amplitude of each actuator unit is defined by the amplitude envelope. Allocated according to unit weight , For unit weights; The imaginary unit, ; This is the time phase term of the stress wave, characterizing the change of phase over time; For the first The phase offset of each unit is determined by the initial phase. With delay control jointly determined, ; For wavenumber vectors, ,in The wavelength of the stress wave. , is the unit vector in the direction of stress wave propagation; This is the spatial phase term of the stress wave, characterizing the change of phase with spatial distance; Modulo-square operation is used to quantize the wave field energy intensity at the focal point. During constructive interference, the synthesized wave field intensity is the sum of the squares of the wave field intensities of each element, thus achieving focused amplification of energy.

[0086] By focusing stress wave energy, a dynamic stress-locked zone is instantaneously constructed within the soft soil layer. The formation of this zone relies on the coupling effect between the microscopic contact mechanics of soil particles and their macroscopic mechanical properties. Its core formation mechanism is as follows: under the action of a coherent stress wave beam, the soft soil particles generate high-frequency micro-amplitude vibrations (the vibration frequency is the same as the stress wave center frequency). Consistent, vibration amplitude is ), causing the normal stress at the particle contact surface Periodic fluctuations occur. ,in For the static normal stress at particle contact, The amplitude of the normal stress fluctuation; using the non-monotonic friction law in particulate matter mechanics, during the compression half-cycle of the vibration period ( The normal stress at the particle contact surface increases, the effective normal stress between particles increases, and the friction coefficient increases; during the tensile half-cycle of the vibration cycle ( The soil maintains particle contact and prevents separation through the inertial effect of soil particles, thus avoiding a sudden drop in effective normal stress. This asymmetric particle contact mechanical behavior increases the macroscopic tangential modulus of the soil. The increase is significant, and the increase is positively correlated with the stress wave field strength. In the micro-disordered soil particle structure, a macro-ordered force chain network reinforcement zone is induced along the stress wave propagation direction. This reinforcement zone is the dynamic stress locking zone.

[0087] The shape and position of the dynamic stress-locking band can be precisely controlled by adjusting the phase offset of each actuator unit in real time. This alters the focal position and energy distribution of the constructive interference of stress waves, enabling the locking zone to translate, expand, and contract in three-dimensional space, adapting to the dynamic evolution characteristics of uneven subsidence in soft strata. Simultaneously, the stiffness and thickness of the locking zone are determined by the intensity of the synthesized wave field. The greater the wave field intensity, the higher the equivalent dynamic elastic modulus and the greater the thickness of the locking band, resulting in a stronger virtual rigid support effect. This can directly offset the creep deformation of weak strata and suppress the development of uneven settlement from the soil itself.

[0088] Furthermore, such as Figure 4 As shown, a dual-loop adaptive feedback control mechanism is employed to monitor and control the effect of the dynamic stress-locking band in real time. This mechanism consists of an inner loop velocity control loop and an outer loop displacement shape control loop. The two loops work independently yet collaboratively. The inner loop rapidly suppresses high-frequency disturbances, while the outer loop precisely optimizes the spatial shape of the micro-strain field, working together to control the micro-strain field residuals. As the core monitoring indicator, ,in The micro-strain threshold under uniform settlement is determined by engineering design requirements, and the norm of the micro-strain field residuals is... Characterizing the overall unevenness of settlement, the control target is to make until the micro-strain field distribution returns to a uniform state.

[0089] Inner loop speed control loop: using micro-strain rate As the core feedback quantity, the micro-strain rate is the first derivative of the micro-strain with respect to time. The values ​​are calculated from real-time acquired micro-strain field data using the central difference method. This loop employs a sliding mode variable structure controller as its core control unit; the sliding mode surface function is first defined. ,in The setpoint for the micro-strain rate is determined by the allowable settlement rate in the engineering project; subsequently, the sliding mode control law is designed. ,in To control the gain, It is a sign function; through sliding mode variable structure control, the micro-strain rate is rapidly tracked to the set value. It can quickly suppress micro-strain rate changes caused by external environmental disturbances (such as groundwater seepage and sudden load changes) within 50ms, ensuring the dynamic stability of the micro-strain field and providing stable monitoring data for outer loop control.

[0090] Outer loop displacement shape control loop: based on the spatial uniformity index of micro-strain field distribution As the core feedback quantity, the spatial uniformity index is defined as the Gini coefficient of the micro-strain field gradient, used to quantify the spatial non-uniformity of settlement. Its calculation process is as follows: First, calculate the magnitude sequence of the micro-strain field gradient. Sort the sequence in ascending order as follows ( Then the spatial uniformity index is calculated. The formula is: ,in, , A value closer to 0 indicates a more uniform distribution of the micro-strain field, while a value closer to 1 indicates a more severe degree of settlement non-uniformity. This loop uses a Model Predictive Controller (MPC) as its core control unit, and sets the prediction time domain... Control Time Domain Using a stratigraphic rheological evolution prediction model as the prediction model, the future is predicted. The micro-strain field distribution at each time step; with spatial uniformity index Minimization is the optimization goal; continuous optimization for the future. The dynamic stress wave excitation parameter set at each time step generates the optimal control sequence, enabling precise control of the spatial morphology of the micro-strain field.

[0091] The dual-loop adaptive feedback control mechanism acquires micro-strain field residuals in real time. This is used as the optimization objective of the heuristic optimization algorithm, and the algorithm optimizes the dynamic stress wave excitation parameter set. Online iterative correction is performed to adjust the norm of the micro-strain field residuals. The potential shrinkage factor continues to decrease. This application employs an improved Behavioral Particle Swarm Optimization (BPSO) algorithm, which introduces a potential well shrinkage factor based on the traditional particle swarm optimization algorithm. By incorporating a stochastic observation state collapse mechanism, the problem of traditional algorithms easily getting trapped in local optima is solved, achieving a global optimal search for the excitation parameters. The specific execution process is as follows:

[0092] Particle swarm initialization: setting the dynamic stress wave excitation parameter set The parameter is used as the particle dimension, particle dimension Consistent with the number of parameters (center frequency, initial phase, amplitude envelope are each 1-dimensional, total dimension) ); Set particle swarm size Randomly generate initial particle swarm positions With speed ,in The lower and upper limits of the values ​​for each parameter are determined by the performance parameters of the microarray actuator. For particle serial numbers, For parameter dimensions.

[0093] Fitness function calculation: The fitness function is designed with minimizing the norm of the micro-strain field residuals as the fitness objective. ,in It should be a very small positive number to avoid the denominator being 0; the larger the fitness function value, the better the effect of the parameter set corresponding to the particle and the lower the degree of non-uniformity of the micro-strain field.

[0094] Particle position update: Introducing a potential well shrinkage factor With the random observation state collapse mechanism, the particle position update formula is:

[0095]

[0096] in, This represents the number of iterations. For the first During the nth iteration The particle in the first The parameter value of the dimension; For the first During the nth iteration The particle in the first The optimal position of the mixed dimension is the individual's historical optimal position. with the global optimal position random weighted average, , These are random weighting coefficients; The potential well contraction factor has a value of [value missing]. This is used to control the search range of particles, which decreases linearly with the number of iterations, thus achieving a gradual transition from global search to local search. For the first The position of the population centroid at the next iteration is calculated as follows: ; For the first During the nth iteration The particle in the first The current position of the dimension; for Uniformly distributed random numbers within the interval are generated by a random number generator; For logarithmic search terms, random step size search of particles is implemented; This is a random observation state collapse mechanism, generated by random numbers. The size is determined when Time to take ,when Time to take This increases the diversity of the particle swarm.

[0097] Optimal position update: Calculate the first position update. The fitness function value of each particle after the next iteration ,like Then update the individual's historical best position. If the fitness of an individual particle's historical best position is greater than its fitness of the global best position, then update the global best position. .

[0098] Iteration Termination and Parameter Output: Set the maximum number of iterations When the number of iterations reaches Or the fitness function value tends to stabilize (the rate of change of fitness over 20 consecutive iterations). When the iteration terminates, the global optimal position is determined. As the optimal set of dynamic stress wave excitation parameters, it is output to the microarray actuator to achieve online parameter correction.

[0099] The improved behavioral particle swarm optimization algorithm, through the above process, optimizes the frequency bandwidth of the dynamic stress wave excitation parameter set. With phase delay matrix Dynamic adjustments are made to ensure that the coherent stress beam emitted by the microarray actuator continuously adapts to the subsidence evolution characteristics of the soft strata. The dual-loop adaptive feedback control mechanism and optimization algorithm form a closed-loop regulation. This process is executed cyclically until the micro-strain field distribution returns to a uniform state, thereby achieving active suppression and adaptive compensation of uneven subsidence in the soft strata.

[0100] Throughout the closed-loop control process, energy dissipation monitoring is performed synchronously to calculate the acoustic power density radiated from the microarray actuators to the soil in real time. With the density of mechanical power absorbed by the soil Define the energy coupling efficiency coefficient. This coefficient characterizes the efficiency of stress wave energy transfer to the soil under normal working conditions. Stable at Meanwhile, the rate of increase in pore water pressure in each sub-compensation unit is monitored in real time. Set a threshold for the rate of increase of pore water pressure. Energy coupling efficiency coefficient threshold When a local area is detected to meet the requirements and When the soil is determined to be about to enter the critical liquefaction state, the safety protection mechanism is immediately triggered:

[0101] Forced to change the center frequency of the coherent stress beam Switch to the non-resonance dissipation frequency band. To avoid resonance between stress waves and the soil skeleton;

[0102] The amplitude envelope is gradually reduced according to the exponential decay law. To zero, This reduces the energy input of stress waves.

[0103] Activating the reverse phase cancellation beam controls the microarray actuator to emit a stress beam with the opposite phase and the same amplitude as the original stress wave, causing the original stress wave and the cancellation beam to interfere destructively in the soil, quickly dissipating the formed dynamic stress locking zone, fundamentally preventing foundation instability, and ensuring the safety and reliability of the entire compensation control process.

[0104] This detailed description of the implementation describes how a controllable virtual rigid support is formed by constructing a dynamic stress-locking band, thereby suppressing uneven settlement at its source. Dual-loop feedback and an improved optimization algorithm enable real-time and accurate correction of compensation parameters, quickly restoring the micro-strain field to uniformity and achieving active suppression and adaptive compensation of settlement. At the same time, energy monitoring and safety mechanisms can effectively avoid the risks of soil liquefaction and foundation instability, thus improving the overall stability of soft soil foundations and the accuracy, timeliness, and safety of settlement control.

[0105] Based on Example 1, this example uses a bridge foundation in a soft coastal silty clay soil stratum as the test object. The initial static elastic modulus of the soil in this area is... The soil has a compressibility of 3.5-5.6 MPa, a porosity φ of 46%-51%, a natural water content of 50%-61%, and a compressibility coefficient of 1.8-2.5 MPa⁻¹, classifying it as a soft stratum with high rheological properties. The test area was defined as 20m × 15m, with a foundation bearing layer interface depth of 8.5m. A total of 32 sets of distributed fiber optic strain sensors, 24 sets of pore water pressure sensors, 8 sets of wave velocity testing probes, and 4 sets of real-time soil modulus testing devices were deployed throughout the area. The sensor sampling frequency was 10Hz, and continuous monitoring was performed. The verification process involved multi-physics field data acquisition, training and deployment of the stratum rheological evolution prediction model, installation and debugging of the microarray actuator, dynamic compensation control, and effect monitoring and analysis. Existing methods selected included pre-embedded grouting reinforcement (CK1), adaptive support compensation (CK2), and shallow stress wave reinforcement (CK3). Three mainstream technologies were compared with the method (SY) of this application. Each experimental group used the same monitoring equipment, data standards and test conditions. The microarray actuator was arranged with an 8×8 phased array aperture. The power of the unit piezoelectric ceramic actuator was 50W. The center frequency adjustment range was 50~500Hz. The parameters of the dual-loop adaptive feedback control mechanism and the improved particle swarm optimization algorithm were all calibrated indoors and then deployed on site. The energy coupling efficiency coefficient threshold was 0.4 and the pore water pressure rise rate threshold was 0.2kPa / min.

[0106] The model training phase was based on multiphysics data from the first 10 days of the experiment. The training and test sets were divided in an 8:2 ratio. After 8500 iterations with the Adam optimizer, the validation set loss tended to stabilize. The model's coefficient of determination for predicting microstrain fields was... The model achieved a mean square error (RMSE) of 11.8 με and a mean absolute error (MAE) of 0.028 mm / (m·d) for predicting the settlement rate gradient, meeting the requirements for dynamic compensation control in the field. After deployment, the model accurately identified the main settlement axis of the test area, located 7.2 m from the left longitudinally. The maximum settlement rate gradient at the axis was 0.17 mm / (m·d). Based on this, 22 three-dimensional sub-compensation units were divided, and the dynamic stress wave excitation parameters were calculated to have a center frequency of 175225 Hz, an amplitude envelope of 0.7-1.6 V, and an initial phase of 0-2π uniformly distributed. After the microarray actuator was embedded and debugged, the results showed that the uniformity of the synthesized wave field intensity at the focal point of the target compensation area was 93.1%, the dynamic stress locking band formation time was ≤0.4 s, and the locking band thickness could be precisely adjusted within the range of 0.3-1.3 m.

[0107] like Figure 5 As shown, a series of cloud maps depicting the spatiotemporal evolution of the microstrain field in the weak strata of the test area were obtained. Using a 20m × 15m plane of the test area as the base, the color gradient from light blue to dark red represents the microstrain value from 0 to 100 με, with a color scale accuracy to 0.1 με. The cloud maps are arranged chronologically at four key nodes: before the experiment, day 7, day 15, and day 30. Each time node's cloud map is separated by dashed lines and shares the same color scale system. The main settlement axis (thick black solid line), the boundary of the sub-compensation unit (thin white dashed line), the location of the microarray actuator (red solid dots), and the area covered by the dynamic stress-locking zone (yellow semi-transparent mask) are also marked. As can be visually observed from the figure, the micro-strain field distribution was extremely uneven before the experiment, with the micro-strain value reaching 90.2 με at the main settlement axis and a gradient difference in the surrounding area. After dynamic compensation using the method of this application, the micro-strain value at the main settlement axis decreased to 45.6 με on the 7th day, and the uniformity of micro-strain in the area covered by the locking band was significantly improved. On the 15th day, the micro-strain value across the entire region was controlled within 30 με. On the 30th day, the micro-strain value across the entire region stabilized below 20 με, with only 10.8 με at the main settlement axis. The spatial distribution of the micro-strain field became completely uniform, and the area covered by the locking band highly overlapped with the area of ​​uniform micro-strain, clearly demonstrating the precise temporal and spatial control effect of the method of this application on the micro-strain field.

[0108] This validation performed comprehensive monitoring of the 30-day core settlement control indicators of the four methods. The data, after standardization, are shown in Table 1, which presents the monitoring results of the core settlement control indicators for the four methods in soft strata. For the maximum settlement, For regional settlement difference, This represents the spatial uniformity index of the micro-strain field. The settling time of the micro-strain field. The average equivalent dynamic elastic modulus of the soil in the target area. To ensure overall safety, all indicators are averaged over 30 days of monitoring data, rounded to three significant figures. Table 1 clearly shows that the method presented in this application is comprehensively superior to the three existing technologies in terms of core settlement control indicators. Compared to CK1, CK2, and CK3, the levels decreased by 62.8%, 54.1%, and 47.6%, respectively. Compared to the other three, the figures decreased by 77.2%, 70.3%, and 63.5% respectively. The value is reduced to 0.07, which is 86.5%, 81.6% and 76.9% lower than the other three, respectively. It can achieve micro-strain field stability within 4.0 days, which is 73.3%, 66.7% and 60.0% shorter than the other three, respectively. At the same time, it achieves a significant increase in the equivalent dynamic elastic modulus of the soil and the safety factor is improved by more than 85% compared with the existing technology. This fully demonstrates the significant technical advantages of the method in terms of settlement control accuracy, uniformity and stiffness improvement.

[0109] Table 1. Monitoring results of core indicators for subsidence control in soft strata using four methods.

[0110] Experimental group (mm) (mm) (d) (MPa) CK1-Grouting Reinforcement Method 32.9 9.2 0.52 15.0 8.3 1.51 CK2-Adaptive Support Method 26.8 7.1 0.40 12.0 7.2 1.48 CK3-Shallow Stress Wave Reinforcement Method 23.5 5.7 0.30 10.0 10.6 1.65 SY - Application Method 12.2 2.1 0.07 4.0 18.9 2.88

[0111] like Figure 6 The maximum settlement and settlement rate are shown as two-dimensional spatiotemporal variations. Figure 6 The curves correspond to four methods: CK1, CK2, CK3, and SY. The solid blue line represents the curve of maximum settlement over time, and the dashed red line represents the curve of settlement rate over time. The inflection point of settlement rate (black triangle mark), the stable time node of micro-strain field (vertical black dashed line), and the settlement stability threshold (horizontal gray dashed line) are also marked. All curves have been smoothed by cubic spline interpolation, and the data point marking accuracy is 0.1 mm and 0.01 mm / d. A comparison of the four sub-figures in the attached diagram clearly shows that the maximum settlement of methods CK1, CK2, and CK3 exhibits an approximately linear growth trend, with no obvious inflection point in the settlement rate. The settlement rate remains above 0.5 mm / d for 30 days without any stabilization trend. In contrast, the maximum settlement of the method in this application shows a slow increase in the first four days of the test, then stabilizes completely after four days. The settlement rate rapidly decreases from an initial 0.95 mm / d to below 0.05 mm / d, and after the stabilization point, the settlement shows no further increase, with the settlement rate approaching zero. This clearly demonstrates that the method in this application can quickly achieve active suppression and permanent stabilization of settlement in soft strata.

[0112] To further verify the dynamic control performance, energy utilization efficiency, and safety protection capability of the method proposed in this application, continuous monitoring of the dynamic response and safety-related indicators of the four methods was conducted. Table 2 shows the monitoring results of the dynamic control and safety protection performance indicators of the four methods. ε represents the residual norm of the micro-strain field, which characterizes the degree of settlement non-uniformity; the smaller the value, the better the uniformity. The average energy coupling efficiency coefficient characterizes the efficiency of stress wave energy transfer to the soil. The probability of soil liquefaction risk; The average response time for online parameter correction characterizes the real-time performance of dynamic control. To determine the maximum increase in pore water pressure throughout the experiment; The 30-day settlement rebound rate. As shown in Table 2, the method of this application... With an energy coupling efficiency of only 17.8 με, far lower than the three existing technologies, the energy coupling efficiency coefficient reaches 0.76, which is more than 43% higher than the existing technologies. The probability of liquefaction risk is 0 throughout the test. The parameter correction response time is only 32ms, which is more than 82% shorter than the existing technologies. Moreover, the increase in pore water pressure is minimal and the settlement rebound rate is 0. This fully demonstrates that the method of this application has excellent dynamic control real-time performance, energy utilization efficiency and safety protection capabilities. It can fundamentally avoid the risk of soil liquefaction and achieve permanent suppression of settlement.

[0113] Table 2. Monitoring Results of Dynamic Control and Safety Protection Performance Indicators of Four Methods

[0114] Experimental group (with) (%) (ms) (kPa) (%) CK1-Grouting Reinforcement Method 96.5 0.53 8.5 260 12.8 1.2 CK2-Adaptive Support Method 73.2 0.49 6.9 205 8.6 0.8 CK3-Shallow Stress Wave Reinforcement Method 58.7 0.58 4.2 160 10.3 0.5 SY - Application Method 17.8 0.76 0.0 32 4.1 0.0

[0115] like Figure 7 The diagram shows the coupling between the three-dimensional morphology of the dynamic stress locking band and the spatial distribution of the equivalent dynamic elastic modulus of the soil. The X and Y axes represent the planar coordinates of the test area, and the Z axis represents the burial depth. The locking band is displayed as a semi-transparent yellow solid model, with the transparency varying with the burial depth gradient. The three-dimensional dimensions (length, width, and thickness), focal spatial coordinates, and segmental boundaries of each sub-compensation unit are also labeled. The three-dimensional model is constructed at a 1:1 scale, and its morphology perfectly matches the field measurement data. Figure 8 The image shows the spatial distribution cloud map of the equivalent dynamic elastic modulus of the soil at a corresponding burial depth (8.5m), using the same... Figure 7 The X and Y plane coordinates are completely consistent, with color scales ranging from light yellow to dark red to indicate an elastic modulus that gradually changes from 0 to 25 MPa, with an accuracy of 0.1 MPa. The cloud map marks the average elastic modulus of each sub-compensation unit (white numbers), the maximum elastic modulus at the main settlement axis, and the distribution range of the initial elastic modulus of the soil (blue dashed boxes). From Figure 7 , Figure 8The linkage analysis clearly shows that the dynamic stress locking band is precisely distributed in three dimensions along the main settlement axis. The thickness of the locking band adaptively adjusts within the range of 0.6-1.1m according to the stiffness requirements of the sub-compensation unit, and the focal position completely coincides with the center of the sub-compensation unit. The equivalent dynamic elastic modulus of the soil at the corresponding burial depth is highly coupled with the shape of the locking band. The elastic modulus of the area covered by the locking band is 3-4 times higher than the initial value, reaching 21.5MPa at the main settlement axis. The elastic modulus of the non-locking band area does not change significantly. This directly proves that the method of this application can achieve precise, directional, and dynamic improvement of the soil stiffness in the target area by constructing a dynamic stress locking band, forming a stable virtual rigid support.

[0116] like Figure 9 The spatiotemporal evolution coupling analysis of the multi-core indicators shown in the figure can intuitively demonstrate the dynamic compensation and closed-loop control capabilities of the proposed method. The figure contains five core curves: micro-strain field residual norm (blue solid line), pore water pressure (red dashed line), energy coupling efficiency coefficient (green dotted line), equivalent dynamic elastic modulus (purple dotted line), and settlement rate (orange double dotted line). The data of each curve have been standardized and mapped to the corresponding dimensions. The safety threshold line (black horizontal dashed line, including pore water pressure threshold and energy coupling efficiency coefficient threshold), micro-strain field stability node (vertical red dashed line), and safety mechanism trigger warning interval (gray semi-transparent mask) are also marked. As can be seen from the attached figure, in the initial stage of the experiment, the residual norm of the micro-strain field, pore water pressure, and settling rate were all at a high level, the energy coupling efficiency coefficient rapidly increased to above 0.75 and remained stable, and the equivalent dynamic elastic modulus increased rapidly in tandem. As dynamic compensation control was implemented, the residual norm of the micro-strain field and the settling rate decreased exponentially, and the pore water pressure rose slightly and then stabilized below the threshold, with no cases exceeding the warning range. After the micro-strain field stabilized, all core indicators tended to stabilize, the energy coupling efficiency coefficient remained between 0.7 and 0.78, and the equivalent dynamic elastic modulus stabilized at 1819 MPa. This fully demonstrates that the dual-loop adaptive feedback control mechanism and improved optimization algorithm of the method in this application can achieve coordinated regulation of multiple core indicators, with significant closed-loop correction effect, no safety risks throughout the process, and excellent technical stability and reliability.

[0117] Compared with existing mainstream technologies, the method proposed in this application achieves a qualitative improvement in terms of settlement control accuracy, dynamic regulation capability, energy utilization efficiency, and safety protection performance. It has broad engineering application prospects in various soft strata foundation engineering projects, such as coastal silty clay, inland soft soil, and lacustrine sedimentary soft strata. It can significantly improve the long-term stability of soft strata foundations, reduce the later maintenance costs of engineering projects, and has significant engineering practical value and technology promotion value.

[0118] The above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. For those skilled in the art, the present invention can have various modifications and variations. Any changes, modifications, substitutions, integrations, and parameter changes made to these embodiments within the spirit and principles of the present invention, without departing from the principles and spirit of the present invention, through conventional substitutions or to achieve the same function, fall within the scope of protection of the present invention.

Claims

1. An adaptive compensation control method for uneven settlement in soft strata, characterized in that, include: Data on distributed micro-strain field and pore water pressure gradient at the bottom of the foundation of the weak stratum are obtained, and the data are input into a stratum rheological evolution prediction model constructed based on a physical information neural network. Based on the spatiotemporal attention encoding module in the formation rheological evolution prediction model, multi-scale spatiotemporal feature extraction is performed on the micro-strain field data to generate a formation creep trend vector; By using the virtual stiffness reconstruction algorithm embedded in the formation rheological evolution prediction model, the equivalent modulus correction coefficient of the target compensation area is calculated based on the formation creep trend vector, and a dynamic stress wave excitation parameter set is generated. Using a micro-array actuator embedded in the foundation bearing layer interface, a coherent stress beam with a specific frequency, phase and amplitude is emitted according to the dynamic stress wave excitation parameter set. The dynamic stress locking zone is instantaneously constructed in the soft soil layer by utilizing the constructive interference principle of waves, thereby increasing the equivalent dynamic elastic modulus of the soil in the target area to form a virtual rigid support effect. A dual-loop adaptive feedback control mechanism is adopted to monitor the residual of the micro-strain field under the action of the dynamic stress locking zone in real time. The dynamic stress wave excitation parameter set is corrected online through a heuristic optimization algorithm until the micro-strain field distribution is restored to a uniform state, thereby realizing the active suppression and adaptive compensation of uneven settlement of soft strata.

2. The adaptive compensation control method for uneven settlement in soft strata according to claim 1, characterized in that, The formation rheological evolution prediction model is constructed using a physical information neural network, specifically as follows: The Navier-Stokes equations and soil constitutive relations are embedded as physical constraints into the loss function of a neural network to construct a physical information loss term. The physical information loss item The calculation formula is: ,in, Indicates the first Micro-strain at each sampling point Indicates time, Indicates dependence on porosity The diffusion coefficient tensor, Indicates pore water pressure, Represents the soil viscoplastic rheological operator; by minimizing the data-driven loss term and the physical information loss term. The weighted sum of these factors is used to train the formation rheological evolution prediction model.

3. The adaptive compensation control method for uneven settlement in soft strata according to claim 2, characterized in that, The spatiotemporal attention encoding module is specifically used for: The input micro-strain field data is decomposed in the spatial domain and sliced ​​in the time domain using a sliding window to generate spatial feature maps and time series segments. Calculate the association weight matrix between grid nodes in the spatial feature map using a multi-head self-attention mechanism. And use causal convolutional networks to compute the temporal dependency weight vectors of time series segments. ; The correlation weight matrix With time-dependent weight vector Kronecker product operation is performed to generate spatiotemporal joint attention mask, and the original micro-strain field data is weighted and fused to output a formation creep trend vector containing high-order spatiotemporal coupling features.

4. The adaptive compensation control method for uneven settlement in soft strata according to claim 1, characterized in that, The specific steps for calculating the equivalent modulus correction coefficient using the virtual stiffness reconstruction algorithm include: Based on the formation creep trend vector, the main subsidence axis with the largest subsidence rate gradient is identified; several sub-compensation units are divided along the main subsidence axis, and the target equivalent dynamic elastic modulus is calculated for each sub-compensation unit. The target equivalent dynamic elastic modulus The calculation formula is: ,in, The initial static elastic modulus distribution, The sedimentation rate gradient vector. This is the stiffness enhancement gain coefficient. Gradient-sensitive threshold, The rheological compensation factor decays over time; based on the target equivalent dynamic elastic modulus. The difference between the measured equivalent modulus and the actual equivalent modulus is used to invert the equivalent modulus correction factor to obtain the required equivalent modulus correction factor.

5. The adaptive compensation control method for uneven settlement in soft strata according to claim 4, characterized in that, The dynamic stress wave excitation parameter set includes the center frequency. Initial phase and amplitude envelope ; The center frequency The determination method is as follows: analyze the particle size distribution curve and pore structure characteristics of the weak strata, and calculate the characteristic resonance frequency of the soil skeleton. The center frequency Set to slightly below the characteristic resonant frequency The sub-resonance frequency band, namely ,in This is the sub-resonance frequency adjustment coefficient, used to prevent soil liquefaction and maximize the interparticle frictional interlocking effect.

6. The adaptive compensation control method for uneven settlement in soft strata according to claim 5, characterized in that, The process by which the microarray actuator emits a coherent stress beam includes: The microarray actuator is divided into several phased array sub-apertures, each sub-aperture containing... A separate piezoelectric ceramic actuator unit; Based on the dynamic stress wave excitation parameter set, a delay control is applied to each actuator unit so that the spherical waves emitted by all sub-apertures undergo constructive interference at the focal point of the target compensation region. The synthetic wave field intensity at the focal point The formula is: ,in, For the first The amplitude of each unit, For the first Phase offset of each unit For wavenumber vectors, For the first The distance from each unit to the focal point; through dynamic adjustment The dynamic stress-locking band is controlled to move in shape and position in three-dimensional space.

7. The adaptive compensation control method for uneven settlement in soft strata according to claim 6, characterized in that, The formation mechanism of the dynamic stress-locking band is as follows: Under the action of the coherent stress beam, the soft soil particles generate high-frequency micro-amplitude vibrations, resulting in periodic fluctuations in the normal stress at the particle contact surface. By utilizing the non-monotonic friction law in particle mechanics, the effective normal stress between particles is increased during the compression half-cycle of the vibration cycle, and the contact is maintained without separation through the inertial effect during the tension half-cycle. Asymmetric contact mechanics behavior leads to a significant increase in the tangent modulus of the soil macroscopically, thereby inducing a macroscopically ordered force chain network reinforcement zone in the microscopically disordered soil structure, forming a dynamic stress-locked zone.

8. The adaptive compensation control method for uneven settlement in soft strata according to claim 1, characterized in that, The dual-loop adaptive feedback control mechanism includes an inner loop velocity control loop and an outer loop displacement shape control loop. The inner loop speed control loop uses micro-strain rate As a feedback quantity, a sliding mode variable structure controller is used to quickly suppress the sudden change in strain rate caused by high-frequency disturbances; The outer ring displacement mode control loop uses the spatial uniformity index of the micro-strain field distribution. As the feedback quantity, a model predictive controller is used to continuously optimize the dynamic stress wave excitation parameter set in the future time domain; the spatial uniformity index Defined as the Gini coefficient of the micro-strain field gradient, it is used to quantify the degree of non-uniformity in settlement.

9. The adaptive compensation control method for uneven settlement in soft strata according to claim 8, characterized in that, The heuristic optimization algorithm is an improved behavioral particle swarm optimization algorithm; in the behavioral particle swarm optimization algorithm, the particle position update equation introduces a potential well shrinkage factor. Similar to the random observation state collapse mechanism; the particle position update formula is: ,in, For the first During the nth iteration The particle in the first Dimensional parameter values, It is a random weighted average of the individual's historical best position and the global best position. The location of the population centroid. for Uniformly distributed random numbers within the interval; the frequency bandwidth and phase delay matrix of the dynamic stress wave excitation parameter set are dynamically adjusted through the position update formula to minimize the norm of the micro-strain field residual.

10. The adaptive compensation control method for uneven settlement in soft strata according to claim 1, characterized in that, The method also includes an energy dissipation monitoring step: The ratio of the acoustic power density radiated by the microarray actuator to the soil to the mechanical power density absorbed by the soil is calculated in real time, and the energy coupling efficiency coefficient is defined. When the rate of increase in pore water pressure in a local area exceeds a preset threshold, and the energy coupling efficiency coefficient... When an abnormally steep drop occurs, it is determined that the soil is about to enter the critical state of liquefaction; The safety mechanism is immediately triggered, forcibly switching the frequency of the coherent stress beam to the non-resonant dissipation band and gradually reducing the amplitude envelope. The beam is reduced to zero, and the reverse phase cancellation beam is activated to quickly dissipate the formed dynamic stress locking zone and prevent the foundation from becoming unstable.

Images