A site ground motion amplification coefficient self-adaptive reduction method based on intelligent response sandwich

Abstract

The application discloses a site seismic vibration amplification coefficient self-adaptive reduction method based on an intelligent response interlayer, and belongs to the technical field of rock-soil engineering earthquake resistance, shock absorption and disaster prevention. The method sets an interlayer composed of intelligent response materials in the site covering layer or near the interface between the covering layer and the bedrock. Under the conditions of small earthquakes and normal use, the interlayer maintains the initial stiffness and low damping state, and maintains the original dynamic characteristics of the site. When a strong earthquake occurs, the interlayer independently realizes modulus reduction and damping increase under the triggering of large earthquake strain or artificial intelligence early warning, so that the site predominant frequency moves to low frequency to realize frequency detuning, and the seismic wave energy is dissipated through high damping. The artificial intelligence model continuously predicts the response at the epicenter and optimizes the interlayer parameters. After the earthquake, the mechanical properties of the intelligent material automatically recover. The application realizes intelligent adaptive control of the site amplification coefficient, and has the advantages of recoverability and construction flexibility.

Description

Technical Field

[0001] This invention belongs to the field of earthquake resistance, vibration reduction and disaster prevention technology in geotechnical engineering, and specifically relates to an adaptive reduction method for the amplification factor of site ground motion based on intelligent response interlayer. Background Technology

[0002] The ground motion amplification effect is a significant factor contributing to variations in earthquake damage. During the propagation of seismic waves from bedrock to the surface, the weak near-surface overburden layer significantly amplifies these waves. The degree of amplification depends on dynamic parameters such as the soil layer's shear modulus, damping ratio, and thickness. When the dominant frequency of the overburden layer is close to the dominant frequency of the input ground motion, a resonance effect occurs, causing the intensity of the ground motion to increase exponentially, thus exacerbating damage to the superstructure. Therefore, for important engineering sites, effectively controlling the ground motion amplification effect is one of the key issues in seismic design.

[0003] In existing technologies, common measures to control the amplification effect of seismic ground motions at a site mainly fall into two categories: foundation treatment measures and base isolation measures. Foundation treatment measures include replacement, densification, and grouting, which aim to change the dominant frequency of the site by increasing soil stiffness, striving to avoid the frequency band where seismic energy is concentrated. Base isolation measures, on the other hand, reduce the acceleration input to the superstructure by installing isolation layers at the base of the building, extending the natural period of the structure and increasing damping. These techniques have been widely used in practical engineering and have played a positive role in mitigating earthquake disasters.

[0004] However, the existing measures mentioned above have fundamental shortcomings in addressing the randomness and intensity variations of earthquakes. First, once foundation treatment is completed, the site's dynamic parameters are essentially fixed. This "one-size-fits-all" strategy is insufficient to cope with earthquakes of different focal mechanisms, propagation paths, and intensities. For some earthquakes, the site may even fall into the resonance range due to changes in stiffness, resulting in an unfavorable amplification effect. Furthermore, the soil stiffness itself decreases due to nonlinearity under strong earthquakes, and fixed treatment measures cannot further increase energy consumption according to real-time needs. Second, while seismic isolation layers can provide protection during strong earthquakes, they only affect a single building and cannot improve the amplification of the entire site. Additionally, adding seismic isolation layers to existing buildings is extremely difficult and costly. Therefore, there is an urgent need for a site amplification factor control method that can adaptively adjust the site's dynamic characteristics based on earthquake intensity, balancing daily use with strong earthquake protection requirements.

[0005] In recent years, artificial intelligence (AI) technology has made groundbreaking progress in the field of earthquake engineering. Earthquake early warning models based on deep learning can achieve millisecond-level earthquake response prediction, and rapid prediction methods for site ground motion effects that integrate numerical simulation and machine learning can efficiently assess the ground motion amplification characteristics of complex sites. However, there is currently a lack of systematic solutions that combine AI technology with smart response materials to achieve adaptive control of the site amplification factor throughout the entire process. Summary of the Invention

[0006] This invention aims to address the shortcomings of existing site amplification factor control technologies, such as the inability to adaptively adjust according to seismic intensity, the lack of nonlinear energy dissipation enhancement mechanisms, and the difficulty in balancing daily use and strong earthquake protection. It provides an adaptive reduction method that can autonomously adjust the site dynamic characteristics according to seismic response and introduce artificial intelligence to achieve intelligent early warning, real-time prediction, and parameter optimization.

[0007] To solve the above-mentioned technical problems, the technical solution proposed in this application is as follows:

[0008] This invention provides a method for adaptively reducing the amplification factor of ground motion in a site based on a smart response interlayer, comprising the following steps: S1: An interlayer made of smart responsive material is set inside the overburden layer of the target site or near the interface between the overburden layer and the bedrock. The smart responsive material can undergo reversible modulus and damping changes under external excitation. S2: Under conditions of minor earthquakes, environmental vibrations, or normal use, the intelligent response interlayer maintains its initial stiffness and low damping, keeping the overall equivalent shear wave velocity and dominant frequency of the site constant. S3: When a strong earthquake occurs, the modulus of the intelligent response interlayer is reduced and the damping is increased by utilizing the large strain amplitude of the soil caused by the earthquake motion or the physical field triggered in advance by the earthquake early warning system. S4: By reducing the modulus of the interlayer, the overall equivalent shear stiffness of the site decreases, shifting the dominant frequency of the overburden layer to a lower frequency direction, causing it to deviate from the frequency range where seismic energy is concentrated, thus achieving resonant frequency detuning; at the same time, the increased damping of the interlayer dissipates the energy of the incoming seismic waves. S5: Under the combined effect of frequency detuning and energy dissipation, the peak value of the surface acceleration response spectrum is reduced, thereby achieving adaptive suppression of the ground motion amplification coefficient. S6: After the earthquake, the mechanical properties of the intelligent response material are restored to their initial state, thus restoring the dynamic characteristics of the site.

[0009] Further, in step S1, the intelligent response interlayer is a mixture of shape memory alloy and soil, wherein the shape memory alloy is in the form of granules, short filaments or mesh, and is uniformly mixed into sand or silt to form a composite interlayer; in step S3, under the large strain generated by a strong earthquake, the shape memory alloy undergoes stress-induced martensitic phase transformation, thereby increasing the equivalent damping and reducing the tangent modulus; in step S6, after the earthquake, the shape memory alloy returns to the austenitic state.

[0010] Further, in step S1, the intelligent response interlayer adopts a honeycomb grid structure filled with magnetorheological elastomer or electrorheological elastomer; in step S3, the physical field is a magnetic field or an electric field, and energized coils or electrode plates are arranged in the field. After the earthquake early warning system receives a strong earthquake signal, it energizes the energized coils or applies voltage to the electrode plates before the destructive seismic waves arrive, thereby changing the modulus of the intelligent response interlayer.

[0011] Furthermore, in step S1, the interlayer is configured as multiple layers with different depths, and each layer contains different types of smart response materials or trigger thresholds; shallower interlayers are set with lower strain trigger thresholds, which begin to soften and consume energy during moderate earthquakes, while deeper interlayers are set with higher trigger thresholds, which are activated during rare earthquakes.

[0012] Furthermore, in step S1, the smart response interlayer is set within the depth range where the maximum shear strain occurs in the cover layer.

[0013] Further, in step S1, the intelligent response interlayer is set inside the overburden layer of the target site or near the interface between the overburden layer and the bedrock by means of replacement, mixing or injection; the injection method is the drilling grouting method, in which boreholes are arranged in a grid in the site, and a mixture of shape memory alloy particles and cement bentonite is injected into the boreholes to form a horizontal or inclined hardened composite interlayer at a predetermined depth.

[0014] Furthermore, before step S1, step S0 is included: establishing an artificial intelligence-based site seismic response prediction model and intelligent early warning triggering system. The AI ​​model takes site geological parameters and real-time monitoring data as input, outputs real-time prediction results of site seismic response, and automatically outputs a trigger signal when the predicted site amplification factor exceeds a preset threshold. In step S3, the physical field is driven by the trigger signal output in advance by the AI ​​intelligent early warning triggering system.

[0015] Furthermore, the AI ​​model adopts a Transformer architecture, a convolutional neural network, or a physical information neural network. The training dataset is constructed by numerical simulation of no less than 500 sets of different seismic inputs on the target site. The AI ​​model has a magnification factor prediction accuracy better than 5% and a prediction latency of less than 10 milliseconds.

[0016] Furthermore, the preset threshold mentioned in step S3 is dynamically determined by the AI ​​model based on the spectral characteristics of the input ground motion; a lower trigger threshold is used for ground motions dominated by high-frequency components, and a higher trigger threshold is used for ground motions dominated by low-frequency components.

[0017] Furthermore, after step S3 and before step S4, step S3.1 is included: During the strong earthquake, the AI ​​model accesses the site monitoring data in real time, continuously predicts the real-time seismic response of the site, and dynamically optimizes the target mechanical parameters of the interlayer based on the prediction results. The physical field intensity is adjusted in real time through a feedback control loop, so that the mechanical properties of the interlayer continuously and adaptively adjust with the change of the earthquake motion time history. In step S5, the AI ​​model evaluates the reduction effect of the amplification factor in real time, and issues a parameter adjustment command when the reduction effect does not meet expectations, so as to ensure that the optimal suppression effect is maintained throughout the entire earthquake time history.

[0018] Furthermore, the feedback control loop employs a control strategy network trained using a deep reinforcement learning algorithm. The AI ​​model takes real-time monitoring data within a sliding time window as input and outputs a predicted field response value within the next 0.5 to 2 seconds to achieve rolling prediction and rolling optimization. The closed-loop delay is controlled within 50 milliseconds.

[0019] Furthermore, the AI ​​model continuously inverts the equivalent shear wave velocity and dominant frequency of the site in real time through environmental vibration data during the site's service life, and automatically issues a maintenance warning signal when abnormal degradation of the interlayer mechanical properties is detected.

[0020] Compared with the prior art, the present invention achieves the following beneficial technical effects: This invention, by setting up an intelligent response interlayer, enables the site to maintain its original dynamic characteristics under weak earthquakes and autonomously reduce stiffness and increase damping under strong earthquakes. It combines frequency detuning and energy dissipation mechanisms to synergistically suppress the amplification effect of seismic motion. Furthermore, it introduces artificial intelligence to achieve intelligent early warning triggering, real-time response prediction, and adaptive parameter optimization, effectively eliminating response delay and ensuring optimal suppression effect throughout the entire earthquake time series. At the same time, it also has post-earthquake recoverability, flexible applicability to existing sites, and intelligent health management capabilities throughout the entire life cycle, significantly improving the seismic resilience of the site. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention 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 the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 The flowchart illustrates a method for adaptively reducing the amplification factor of site ground motion based on an intelligent response interlayer, as provided in an embodiment of the present invention. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] like Figure 1 As shown in one embodiment of this application, an adaptive reduction method for the seismic amplification coefficient of a site based on a smart-response interlayer is provided. This method involves setting an interlayer composed of a smart-response material within the overburden layer of the target site or near the interface between the overburden layer and the bedrock. Utilizing the autonomous changes in the mechanical properties of this interlayer under strong earthquake excitation, combined with the early triggering of an earthquake early warning system and real-time prediction and parameter optimization by an AI model, adaptive suppression of the seismic amplification coefficient of the site is achieved. This method includes multiple steps executed sequentially, and the specific implementation of each step is described in detail below.

[0025] In step S1, one or more interlayers composed of smart responsive materials are first installed within the overburden layer of the target site or near the interface between the overburden layer and the bedrock. This smart responsive material exhibits reversible modulus or damping changes under external excitation, including but not limited to stress, strain, temperature, magnetic fields, and electric fields. The interlayers can be installed using replacement, hybrid, or injection methods. For newly constructed sites, the interlayers can be formed at a predetermined depth through excavation and backfilling during site construction; for existing sites, the interlayers can be formed without damaging the superstructure using methods such as drilling and grouting. The thickness, number of layers, and planar distribution of the interlayers need to be determined through numerical analysis based on the specific geological conditions and seismic requirements of the target site. After the interlayers are installed, they become part of the site overburden layer, contributing to the overall dynamic characteristics of the site.

[0026] In step S2, under conditions of minor earthquakes, environmental vibrations, or normal use, the smart response interlayer is maintained in its initial stiffness and low-damping state. At this point, the smart response material is not subjected to excitation sufficient to trigger changes in its mechanical properties, therefore the interlayer's mechanical parameters remain at their initial design values. Since the presence of the interlayer does not alter the overall equivalent shear wave velocity and dominant frequency of the site, the site's dynamic characteristics remain essentially unchanged, and existing buildings, precision equipment, and daily production activities on the site are unaffected. This step ensures the "invisibility" of the method under normal use conditions, avoiding resource waste and inconvenience caused by over-treatment in traditional foundation reinforcement methods.

[0027] In step S3, when a strong earthquake occurs, the material of the smart response interlayer is brought into a working state with reduced modulus and increased damping by utilizing the large strain amplitude of the soil caused by the earthquake motion or the physical field triggered in advance by the earthquake early warning system. Specifically, when the seismic waves generated by the strong earthquake propagate to the interlayer location, they will generate a large dynamic shear strain in the interlayer. When the amplitude of this shear strain exceeds the preset trigger threshold of the smart response material, microstructural changes occur inside the material—for example, stress-induced martensitic transformation in shape memory alloys, or rearrangement of granular chain structure in magnetorheological elastomers under the action of a magnetic field—resulting in changes in the macroscopic mechanical properties of the material, manifested as a decrease in equivalent shear modulus and an increase in equivalent damping ratio. On the other hand, for implementation schemes using field-controlled smart materials (such as magnetorheological elastomers or electrorheological elastomers), a trigger signal can be output by the earthquake early warning system before the arrival of destructive seismic waves (usually a few seconds to tens of seconds). This can control the energized coils deployed in the site to generate a magnetic field or control the electrode plates to generate an electric field, so that the smart response interlayer is already in a pre-response state with reduced modulus and increased damping when the strong seismic wave arrives, thereby eliminating the response delay.

[0028] In step S4, the overall equivalent shear stiffness of the site decreases by reducing the modulus of the interlayer, thereby shifting the dominant frequency of the overburden towards lower frequencies and deviating it from the frequency range where seismic energy is concentrated, thus achieving resonant frequency detuning. Simultaneously, the increased damping of the interlayer dissipates the energy of incoming seismic waves. Specifically, the dominant frequency of the overburden is proportional to the equivalent shear wave velocity of the overburden, which in turn is positively correlated with the equivalent shear modulus of the soil layer. When the modulus of the interlayer decreases, the equivalent shear wave velocity of the entire overburden decreases accordingly, and the dominant frequency shifts to the lower frequency end. Since the seismic energy of most strong earthquakes is concentrated within a specific frequency range, shifting the dominant frequency of the site out of this range effectively avoids the resonance amplification effect. At the same time, the increased damping of the interlayer means increased energy consumption per vibration cycle; the energy of incoming seismic waves is more converted into heat and dissipated as it passes through the interlayer, further reducing the energy input that can induce resonance.

[0029] In step S5, under the combined effect of frequency detuning and energy dissipation, the peak value of the surface acceleration response spectrum is significantly reduced, and the seismic amplification factor of the site is adaptively suppressed. Frequency detuning allows the dominant frequency of the site to avoid the frequency band where seismic energy is concentrated, thus preventing the generation of resonance amplification effect at the source; energy dissipation attenuates the energy of the seismic waves actually transmitted to the site, reducing the amplitude of surface ground motion. The two mechanisms work simultaneously, complementing and reinforcing each other, so that the surface acceleration response spectrum is effectively reduced across the entire frequency range, and the seismic amplification factor of the site is reduced to a safe level.

[0030] In step S6, after the earthquake, the mechanical properties of the smart response material are restored to their initial state, thus restoring the site's dynamic characteristics. Because the changes in the smart response material used in this invention are reversible—shape memory alloys, after completing a martensitic transformation, will revert to an austenitic state as stress is relieved and temperature rises, and magnetorheological or electrorheological elastomers will return to a disordered particle distribution state after the removal of the external field—therefore, the modulus and damping of the interlayer will automatically recover to their initial values ​​after the earthquake. The overall equivalent shear wave velocity and dominant frequency of the site subsequently recover to pre-earthquake levels, allowing the site to continue to withstand subsequent earthquakes or be used normally. This recoverable characteristic makes this invention particularly suitable for areas with frequent aftershocks, allowing it to cope with the next earthquake without any repair work after a strong earthquake.

[0031] In one embodiment of this application, the smart response interlayer in step S1 employs a mixture of shape memory alloy and soil. The shape memory alloy can be a nickel-titanium alloy (NiTi alloy), and its form can be granular, short filament, or mesh, uniformly mixed into sand or silt to form a composite interlayer. The diameter of the alloy wire can be selected from 0.3 mm to 1.0 mm, the length from 5 mm to 20 mm, and the mass content can be selected between 3% and 10% depending on the target modulus reduction. Under the large strain generated by a strong earthquake, the shape memory alloy undergoes a stress-induced martensitic phase transformation, that is, a transformation from austenite to martensite. During this phase transformation, the stress-strain curve of the alloy exhibits a full hysteresis loop, consuming a large amount of energy in each loading-unloading cycle, and the equivalent damping ratio can increase to over 15%; simultaneously, because the elastic modulus of the martensitic phase is lower than that of the austenitic phase, the tangential modulus of the alloy decreases by approximately 30% to 40%. After the earthquake, as stress is relieved and the temperature recovers above the martensitic inverse transformation temperature, the shape memory alloy reverts to its austenitic state, and its modulus and damping return to their initial values. The advantage of this implementation scheme is that it requires no external energy supply, relying entirely on seismic strain energy for autonomous triggering, making it suitable for remote sites lacking external power sources.

[0032] In one embodiment of this application, the smart responsive interlayer in step S1 employs a honeycomb grid structure filled with magnetorheological elastomers or electrorheological elastomers. The magnetorheological elastomer is made by dispersing micron-sized soft magnetic particles (such as carbonyl iron powder) in a polymer matrix (such as natural rubber or silicone rubber), and its shear modulus can change significantly under the influence of an external magnetic field. The electrorheological elastomer is made by dispersing dielectric particles in a polymer matrix, and its shear modulus can change significantly under the influence of an external electric field. Without an external field, the particles are randomly distributed, and the material as a whole exhibits a low shear modulus. When an external magnetic or electric field is applied, the particles align into a chain-like structure along the field direction, resulting in a significant increase in the overall shear modulus of the material. This change is continuous, rapid (millisecond-level), and completely reversible. The honeycomb grid structure is made of polymer material forming the grid skeleton, with the honeycomb pores of the grid filled with magnetorheological elastomers or electrorheological elastomers. This structure ensures both the overall stability of the interlayer and provides sufficient field-controlled material volume fraction to ensure modulus adjustment. Electrified coils (to generate a magnetic field) or electrode plates (to generate an electric field) are deployed within the site. Normally, there is no field effect, and the field-controlled material remains in a low-stiffness state. When the earthquake early warning system receives a strong earthquake signal, it energizes the coils or applies voltage to the electrode plates before the destructive seismic waves arrive, altering the shear modulus of the field-controlled material. Depending on the earthquake resistance strategy, the current or voltage can be increased during a strong earthquake to improve the interlayer stiffness and achieve "hardening deharmonicization," or the external field can be removed to reduce stiffness and enter a high-energy-consuming state. By controlling the intensity and timing of the external field, the site's dynamic characteristics can be flexibly adjusted to adapt to different types of seismic inputs.

[0033] In one embodiment of this application, the interlayer in step S1 is configured as multiple layers with different depths and different types or trigger thresholds of smart response materials contained in each layer. For example, two or more smart response interlayers can be set at different depths within the overburden layer. Shallower interlayers are buried at shallower depths and experience less confining pressure, so a lower strain trigger threshold is set, allowing them to soften and dissipate energy even during moderate earthquakes. Deeper interlayers are buried at greater depths and experience greater confining pressure, so a higher trigger threshold is set, activating only during rare earthquakes. Different smart response materials can be used for each layer—for example, shallow layers can use a mixture of shape memory alloy and soil to utilize its strain triggering characteristics, while deeper layers can use magnetorheological elastomer-filled grids to utilize their field-controlled rapid response characteristics—thereby achieving graded adaptive control under different seismic intensities. When setting up multiple interlayers, the spacing between layers, the thickness of each layer, and the planar distribution range of each layer need to be optimized through site seismic response analysis to ensure that the best amplification factor reduction effect can be obtained under different seismic inputs.

[0034] Specifically, shallower interlayers are configured with lower strain triggering thresholds, initiating softening and energy dissipation even during moderate earthquakes; deeper interlayers are configured with higher triggering thresholds, activating only during rare earthquakes. This tiered triggering design ensures appropriate site protection under earthquakes of varying intensities: during moderate earthquakes, the shallower interlayers respond first, providing moderate frequency detuning and energy dissipation to prevent excessive amplification of the site's seismic amplification factor; during rare earthquakes, both the shallow and deep interlayers work together to provide maximum frequency detuning and energy dissipation, ensuring site safety during earthquakes. The tiered triggering thresholds can be adjusted by modifying the content and type of smart response materials and the depth and confining pressure of the interlayers—for example, reducing the mass content of shape memory alloys can increase their triggering strain threshold, and increasing the interlayer depth can utilize confining pressure to suppress martensitic phase transformation, thereby increasing the triggering threshold.

[0035] In one embodiment of this application, step S1 involves setting up a smart response interlayer within the depth range where the maximum shear strain occurs in the overburden layer. Site seismic response analysis can determine the distribution of shear strain along depth in the overburden layer under different seismic inputs, thereby identifying the depth range where the maximum shear strain occurs. Setting up the smart response interlayer within this depth range ensures that the interlayer can generate a sufficiently large strain amplitude during an earthquake to trigger changes in the mechanical properties of the smart response material, thus fully utilizing the modulus reduction and damping increase effects of the interlayer. The depth range of maximum shear strain is typically located several meters to over twenty meters below the surface, with the specific location depending on the thickness of the overburden layer, the soil stiffness profile, and the characteristics of the input seismic motion. By setting the interlayer within this depth range, the maximum control effect can be achieved with minimal material usage.

[0036] In one embodiment of this application, the smart response interlayer in step S1 is disposed within the overburden layer of the target site or near the interface between the overburden layer and the bedrock through a replacement method, a mixing method, or an injection method. The replacement method is suitable for newly constructed sites or existing sites with excavation conditions, where the interlayer is formed by excavating to a predetermined depth, backfilling with a mixture of smart response material and soil, and compacting. The mixing method is suitable for site construction, where smart response material is mixed with in-situ soil within a predetermined depth range to form a composite interlayer. The injection method is suitable for existing sites without large-area excavation conditions, where smart response material is injected in slurry form within a predetermined depth range through drilling, and the interlayer is formed after the slurry solidifies. Each of the three methods has its applicable scenarios and can be flexibly selected according to the specific conditions of the target site and construction feasibility.

[0037] Specifically, the injection method employs borehole grouting. Boreholes are arranged in a grid pattern within the site, with the spacing and depth determined through numerical analysis based on the site's geological conditions and interlayer design requirements. A mixture of shape memory alloy particles and cement-bentonite is injected into the boreholes. Under pressure, the grout penetrates radially along the boreholes, filling soil pores and forming a horizontal or inclined hardened composite interlayer at a predetermined depth. After borehole grouting, the hardened bodies formed at each borehole are distributed in a grid pattern on the plane. Although the hardened bodies are not physically continuous, when the borehole spacing is sufficiently small and the mechanical properties of the hardened bodies differ significantly from the surrounding soil, the dynamic characteristics of the entire grid area will exhibit an equivalent interlayer effect. The borehole spacing and injection volume need to be optimized through numerical analysis to ensure that the resulting grid-like hardened bodies can generate sufficient site dynamic characteristic regulation effects as a whole. The outstanding advantage of this borehole grouting method is its minimal ground disturbance, allowing for construction during normal building use, and making it particularly suitable for in-situ seismic retrofitting of existing urban building complexes.

[0038] In one embodiment of this application, the method further includes step S0, prior to step S1, a step of establishing an AI-based site seismic response prediction model and an intelligent early warning triggering system. The AI ​​intelligent early warning triggering system employs a deep learning model based on the Transformer architecture. The input features for model training include the thickness of the site overburden, shear wave velocity and density profiles of each soil layer, peak ground acceleration and dominant frequency of the bedrock input ground motion, and monitoring data from a real-time acceleration sensor array within the site. The output target is the predicted values ​​of the site surface acceleration response spectrum and amplification factor. The training dataset is constructed as follows: a one-dimensional or two-dimensional site seismic response analysis model is established for the target site (e.g., using the DEEPSOIL program), and bedrock ground motion time histories of different intensities and spectral characteristics (no fewer than 500 sets) are input. A site response database covering different ground motion input conditions is generated through numerical simulation. This database is used to train the AI ​​model, enabling the model to accurately predict the site amplification effect caused by subsequent ground motions within milliseconds after receiving an earthquake early warning signal, based on limited first-arrival information of the seismic waves. The AI-powered intelligent early warning triggering system accesses real-time data from earthquake monitoring networks or accelerometer arrays deployed on-site. Using the AI ​​model, it performs millisecond-level predictions of the site response after the arrival of seismic waves. When the predicted peak ground acceleration or amplification factor exceeds a preset threshold, it automatically outputs a trigger signal to the physical field generator. In step S3, the physical field is driven by the trigger signal output in advance by the AI-powered intelligent early warning triggering system.

[0039] In one embodiment of this application, the trigger threshold in step S3 is dynamically determined by the AI ​​model based on the spectral characteristics of the input ground motion: a lower trigger threshold is used for ground motions dominated by high-frequency components, and a higher trigger threshold is used for ground motions dominated by low-frequency components, thereby achieving differentiated triggering that adapts to the spectral characteristics of the ground motion. Compared with a fixed threshold, the AI ​​dynamic threshold determination mechanism can adapt to the differences in the spectral characteristics of different ground motions, thus achieving more accurate "on-demand triggering".

[0040] In one embodiment of this application, step S3.1 is included after step S3 and before step S4: During the sustained strong earthquake, the AI ​​model accesses site monitoring data in real time, continuously predicts the real-time seismic response of the site, and dynamically optimizes the target mechanical parameters of the interlayer based on the prediction results. The physical field intensity is adjusted in real time through a feedback control loop, so that the mechanical properties of the interlayer continuously and adaptively adjust with the changes in the seismic motion time history. Specifically, an accelerometer array (no less than 3 measuring points, located at the ground surface, the interlayer location, and the bedrock surface, respectively) is deployed within the site. Sensor data is collected in real time at a sampling rate of no less than 200 Hz and transmitted to an AI computing platform on-site or in the cloud. The AI ​​model takes a sliding time window (e.g., monitoring data from the most recent 2 seconds) as input and outputs a predicted site response value for the next 0.5 to 2 seconds, achieving rolling prediction. Based on the prediction results, the AI ​​model outputs the optimal physical field control command (e.g., target magnetic field strength or target electric field strength) through a pre-trained control strategy network (which may employ deep reinforcement learning algorithms), driving the physical field generator via a digital-to-analog conversion module. The latency of the entire "perception-prediction-decision-execution" closed loop is controlled within 50 milliseconds, ensuring that the mechanical properties of the interlayer can be adjusted in real time in response to the non-stationary time history characteristics of the ground motion. For passively triggered interlayers using shape memory alloys, the AI ​​model does not directly control the mechanical properties of the interlayer, but instead outputs seismic damage assessment and aftershock risk warning information to site managers through real-time prediction results.

[0041] In one embodiment of this application, the burial depth, thickness, number of layers, planar distribution range of the interlayer, and the type and proportion of intelligent response materials are determined using an AI-assisted intelligent optimization design method. The specific steps are as follows: A parametric numerical model is established for the target site, with the interlayer location, thickness, and material parameters as variable design parameters; at least 2000 sets of design parameter combinations are generated using methods such as Latin hypercube sampling; numerical simulations of the site seismic response are performed on each parameter combination to calculate the surface amplification factor under different seismic inputs; a surrogate model (such as Gaussian process regression or deep neural network) is trained using the above simulation data to establish a mapping relationship between the design parameters and the amplification factor reduction effect; with the goal of maximizing the amplification factor reduction, intelligent optimization algorithms such as Bayesian optimization or genetic algorithms are used to perform global optimization on the surrogate model to obtain the optimal interlayer design scheme. Compared with the traditional trial-and-error method, this AI-assisted design method can improve design efficiency by tens of times and can discover optimal parameter combinations that are difficult to cover with human experience.

[0042] In one embodiment of this application, the AI ​​model continuously inverts the equivalent shear wave velocity and dominant frequency of the site in real time through environmental vibration data during the site's service life. When abnormal degradation of the interlayer mechanical properties is detected, it automatically issues a maintenance warning signal, prompting the interlayer performance to be tested or repaired, thereby achieving intelligent health management throughout the entire life cycle.

[0043] Example 1 A key industrial plant is located on a deep, soft soil surface with a 50-meter-thick overburden layer. The shear wave velocity is 160 m / s, the dominant frequency is approximately 0.9 Hz, and the amplification factor reaches 5.0. The plant contains continuously operating precision production lines, and significant changes to the daily vibration environment are not permitted, nor can protection be interrupted under strong earthquakes.

[0044] Engineering surveys determined that the maximum shear strain within the overburden typically occurs within a range of several meters to over twenty meters below the surface. A smart response interlayer with a thickness of approximately 4 meters was selected at a burial depth of 8 to 12 meters. Short nickel-titanium shape memory alloy filaments were mixed with medium-coarse sand; the alloy filaments had a diameter of 0.5 mm, a length of 5–10 mm, and a mass content of approximately 5%. After compaction, a composite interlayer was formed.

[0045] During normal production, the environmental vibration intensity is low, the interlayer strain amplitude is extremely small, the alloy wire is in an elastic martensitic state, exhibiting linear high stiffness, the site's dominant frequency remains unchanged, and the operation of precision equipment is unaffected. Under strong earthquakes, surface movement causes large shear strain in the interlayer (exceeding approximately 0.1%), and the alloy wire enters the stress-induced martensitic positive phase transformation stage. The stress-strain curve exhibits a full hysteresis loop, and the equivalent damping ratio increases sharply to over 15%; simultaneously, the tangent modulus decreases by approximately 30%–40%. The softening of the interlayer reduces the overall equivalent shear wave velocity of the site, shifting the dominant frequency from 0.9 Hz to around 0.6 Hz. Since most of the dominant energy from strong earthquakes is concentrated in the 0.8–1.2 Hz range, frequency detuning is achieved. High damping dissipates a significant amount of input energy. Calculations and simulations show that under the design earthquake, the peak value of the surface acceleration response spectrum decreases by approximately 35%, and the amplification factor decreases from 5.0 to around 3.3. After the earthquake, the alloy wires recovered without residual deformation, and the mechanical properties of the interlayer were restored, making it capable of withstanding aftershocks.

[0046] Example 2 The site of a nuclear power plant has a sandy overburden layer of approximately 20 meters above the bedrock, with a dominant frequency of about 5 Hz. Nuclear power plants have extremely high seismic redundancy requirements and must consider extreme earthquakes.

[0047] Two intelligent interlayers were installed at the bottom of the overburden layer near the bedrock surface. The first layer was a magnetorheological elastomer-filled grid containing ferromagnetic particles, with a thickness of 0.5 meters; the second layer was a shape memory alloy mesh-reinforced sand layer, with a thickness of 1 meter. The site was equipped with an earthquake early warning system, an AI intelligent decision-making platform, and a controllable electromagnet array.

[0048] The AI ​​model was pre-trained based on the site's geological parameters and extensive numerical simulation data to establish a site seismic response prediction model. Normally, the electromagnet is off, and the magnetorheological elastomer (MEE) is in a low-modulus state; the interlayer is inactive during minor earthquakes. When the AI ​​intelligent early warning system receives a signal from the seismic network and predicts within milliseconds that the site's peak ground acceleration will exceed a set threshold and the amplification factor will exceed a safety limit, the system energizes the electromagnet a few seconds before the arrival of the S-wave. This rapidly increases the modulus of the MEE interlayer, temporarily increasing the overall site stiffness and raising the dominant frequency to approximately 7 Hz, far exceeding the dominant frequency of the input seismic motion, achieving "hardening deharmonicization." During strong earthquakes, the AI ​​model continuously accesses monitoring data from the site's acceleration sensor array, continuously predicting the site response and dynamically optimizing the electromagnetic current intensity, allowing the interlayer's mechanical properties to adjust in real time according to the seismic motion's time history. Simultaneously, the second layer of shape memory alloy mesh begins to consume high energy under large strain. The two interlayers work synergistically, using a combination of frequency shifting strategies to balance the spectral differences between P-waves and S-waves, effectively reducing the amplification factor by more than 50%. After the power outage following the earthquake, the system was restored, and the AI ​​model recorded the entire process data for subsequent model iteration and optimization.

[0049] Example 3 For existing urban building complexes, excavating large areas of interlayers is prohibitively expensive. A drilled grouting intelligent interlayer method is employed. Holes are drilled in a grid pattern within the site, and a mixture of shape memory alloy particles and cement-bentonite grout is injected to form a horizontal or inclined hardened composite interlayer at a predetermined depth. After the grout solidifies, it forms multiple discontinuous but overall reinforcing layers with an interlayer effect. The drill spacing and injection volume are optimized through numerical analysis to ensure the formation of an overall effect. This method causes minimal ground disturbance and can be implemented during the normal use of the building.

[0050] The method provided by this invention uses smart response materials that are already in industrial production, and the construction technology can draw on foundation treatment and grouting processes. It can be applied in the seismic design of various important projects and has significant practical value in improving the seismic toughness of sites.

[0051] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for adaptively reducing the amplification factor of ground motion in a site based on an intelligent response interlayer, characterized in that, Includes the following steps: S1: An interlayer made of smart responsive material is set inside the overburden layer of the target site or near the interface between the overburden layer and the bedrock. The smart responsive material can undergo reversible modulus and damping changes under external excitation. S2: Under conditions of minor earthquakes, environmental vibrations, or normal use, the intelligent response interlayer maintains its initial stiffness and low damping, keeping the overall equivalent shear wave velocity and dominant frequency of the site constant. S3: When a strong earthquake occurs, the modulus of the intelligent response interlayer is reduced and the damping is increased by utilizing the large strain amplitude of the soil caused by the earthquake motion or the physical field triggered in advance by the earthquake early warning system. S4: By reducing the modulus of the interlayer, the overall equivalent shear stiffness of the site decreases, shifting the dominant frequency of the overburden layer to a lower frequency direction, causing it to deviate from the frequency range where seismic energy is concentrated, thus achieving resonant frequency detuning; at the same time, the increased damping of the interlayer dissipates the energy of the incoming seismic waves. S5: Under the combined effect of frequency detuning and energy dissipation, the peak value of the surface acceleration response spectrum is reduced, thereby achieving adaptive suppression of the ground motion amplification coefficient. S6: After the earthquake, the mechanical properties of the intelligent response material are restored to their initial state, thus restoring the dynamic characteristics of the site.

2. The method according to claim 1, characterized in that, In step S1, the intelligent response interlayer is a mixture of shape memory alloy and soil. The shape memory alloy is in the form of granules, short filaments, or mesh, and is uniformly mixed into sand or silt to form a composite interlayer. In step S3, under the large strain generated by a strong earthquake, the shape memory alloy undergoes a stress-induced martensitic transformation, thereby increasing the equivalent damping and reducing the tangent modulus. In step S6, after the earthquake, the shape memory alloy returns to the austenitic state.

3. The method according to claim 1, characterized in that, In step S1, the intelligent response interlayer adopts a honeycomb grid structure filled with magnetorheological elastomer or electrorheological elastomer; in step S3, the physical field is a magnetic field or an electric field, and energized coils or electrode plates are arranged in the field. After the earthquake early warning system receives a strong earthquake signal, it energizes the energized coils or applies voltage to the electrode plates before the destructive seismic waves arrive, thereby changing the modulus of the intelligent response interlayer.

4. The method according to claim 1, characterized in that, In step S1, the interlayer is configured as multiple layers with different depths, and each layer contains different types of smart response materials or trigger thresholds; shallower interlayers are set with lower strain trigger thresholds, which begin to soften and consume energy during moderate earthquakes, while deeper interlayers are set with higher trigger thresholds, which are activated during rare earthquakes.

5. The method according to claim 1, characterized in that, In step S1, the smart response interlayer is set within the depth range where the maximum shear strain occurs in the cover layer.

6. The method according to claim 1, characterized in that, In step S1, the intelligent response interlayer is set inside the overburden layer of the target site or near the interface between the overburden layer and the bedrock by means of replacement, mixing or injection. The injection method is a drilling grouting method, in which boreholes are arranged in a grid in the site, and a mixture of shape memory alloy particles and cement bentonite is injected into the boreholes to form a horizontal or inclined hardened composite interlayer at a predetermined depth.

7. The method according to any one of claims 1 to 6, characterized in that, Before step S1, there is also step S0: establishing an artificial intelligence-based site seismic response prediction model and intelligent early warning triggering system. The AI ​​model takes site geological parameters and real-time monitoring data as input, outputs real-time prediction results of site seismic response, and automatically outputs a trigger signal when the predicted site amplification factor exceeds a preset threshold. In step S3, the physical field is driven by the trigger signal output in advance by the AI ​​intelligent early warning triggering system.

8. The method according to claim 7, characterized in that, The AI ​​model adopts a Transformer architecture, a convolutional neural network, or a physical information neural network. The training dataset is constructed by numerical simulation of no less than 500 sets of different seismic inputs on the target site. The AI ​​model has a magnification factor prediction accuracy of better than 5% and a prediction delay of less than 10 milliseconds.

9. The method according to claim 7, characterized in that, The preset threshold mentioned in step S3 is dynamically determined by the AI ​​model based on the spectral characteristics of the input ground motion; a lower trigger threshold is used for ground motions dominated by high-frequency components, and a higher trigger threshold is used for ground motions dominated by low-frequency components.

10. The method according to claim 7, characterized in that, The process includes step S3.1 after step S3 and before step S4: During the strong earthquake, the AI ​​model accesses the site monitoring data in real time, continuously predicts the real-time seismic response of the site, and dynamically optimizes the target mechanical parameters of the interlayer based on the prediction results. The physical field intensity is adjusted in real time through a feedback control loop, so that the mechanical properties of the interlayer continuously and adaptively adjust with the change of the earthquake time history. In step S5, the AI ​​model evaluates the reduction effect of the amplification factor in real time, and issues parameter adjustment instructions when the reduction effect does not meet expectations, so as to ensure that the optimal suppression effect is maintained throughout the entire earthquake time history.

11. The method according to claim 10, characterized in that, The feedback control loop employs a control strategy network trained using a deep reinforcement learning algorithm. The AI ​​model takes real-time monitoring data within a sliding time window as input and outputs a predicted field response value within 0.5 to 2 seconds in the future, thereby achieving rolling prediction and rolling optimization. The closed-loop delay is controlled within 50 milliseconds.

12. The method according to claim 7, characterized in that, During the service life of the site, the AI ​​model continuously inverts the equivalent shear wave velocity and dominant frequency of the site in real time using environmental vibration data, and automatically issues a maintenance warning signal when abnormal degradation of the interlayer mechanical properties is detected.

Images