Micro-displacement reinforcement method for equipment foundation under high-frequency and high-strength vibration

Abstract

The application provides a device foundation micro-displacement reinforcement method under high-frequency high-strength vibration, and belongs to the technical field of wind tunnels. The distributed optical fiber strain sensor network and the acoustic emission sensor array are pre-buried at the bottom of the foundation of the wind tunnel driving system to collect interface strain and debonding acoustic signals in real time. The monitoring data is input into an interface gap identification model constructed by a deep neural network to output the gap position, width and development rate. These parameters are input into a double-layer game optimization model to calculate the optimized grouting pressure and the optimized grouting speed. The high-flow micro-expansion slurry is injected into the gap area according to the optimized parameters to complete the reinforcement and verify the effect, and the technical problem that the interface gap at the bottom of the foundation of the high-frequency high-strength vibration device cannot be accurately identified and dynamically optimized and reinforced is solved.

Description

Technical Field

[0001] This invention belongs to the field of wind tunnel technology, and more specifically, relates to a method for reinforcing the foundation of equipment under high-frequency and high-intensity vibration with micro-displacement. Background Technology

[0002] In the operation and maintenance of high-frequency, high-intensity vibration equipment such as wind tunnel drive systems, the interface between the equipment foundation and the subgrade soil is prone to develop micro-cracks under long-term vibration, leading to foundation instability. Traditional monitoring techniques mainly rely on periodic manual inspections and experience-based judgment to identify the interface debonding state. This involves deploying a small number of discrete sensors for periodic measurements, typically once a day or once a week, and using fixed grouting pressure and speed for reinforcement. However, in current practices for reinforcing vibration equipment foundations, the insufficient spatiotemporal resolution of monitoring methods makes it impossible to capture the dynamic development process of the cracks. The identification error for crack width and location generally exceeds 20%. The selection of grouting parameters relies heavily on engineering experience, lacking an adaptive adjustment mechanism for different crack states, resulting in unsatisfactory reinforcement effects or excessive disturbance to the foundation during the grouting process. In other words, existing technologies suffer from the technical problem of inaccurate identification and reinforcement of interface cracks at the bottom of high-frequency, high-intensity vibration equipment foundations. Summary of the Invention

[0003] In view of this, the present invention provides a method for reinforcing the micro-displacement of equipment foundations under high-frequency and high-intensity vibration, which can solve the technical problem in the prior art that the gaps at the bottom interface of equipment foundations under high-frequency and high-intensity vibration cannot be accurately identified and the foundations can be reinforced.

[0004] This invention is implemented as follows: This invention provides a method for micro-displacement reinforcement of equipment foundations under high-frequency, high-intensity vibration, comprising the following steps: pre-embedding a distributed fiber optic strain sensor network and an acoustic emission sensor array at the bottom of the foundation of a wind tunnel drive system; establishing a dynamic constitutive model considering the soil strain rate effect, and performing finite element simulation of the foundation system using explicit dynamic time history analysis; during the operation of the wind tunnel drive system, the distributed fiber optic strain sensor network collects real-time interface strain data at the bottom of the foundation, and the acoustic emission sensors simultaneously record debonding acoustic signals, obtaining interface strain time history data and acoustic emission signal characteristic parameters; and then... Variable time-history data and acoustic emission signal characteristic parameters are input into the interface gap identification model, which outputs the gap location coordinates, gap width, and gap development rate at the bottom of the foundation. Grouting holes are drilled in the corresponding area at the bottom of the foundation based on the gap location coordinates. The gap width, gap development rate, and grouting hole location parameters are input into a grouting parameter optimization game model, which outputs optimized grouting pressure and optimized grouting speed. High-fluidity micro-expansion grout is injected into the gap area through the grouting holes, and grouting is performed according to the optimized grouting pressure and optimized grouting speed. After grouting is completed, foundation displacement is monitored to verify the reinforcement effect.

[0005] Specifically, the distributed fiber optic strain sensor network is deployed by arranging the fiber optic sensors in a grid pattern before the concrete is poured at the bottom of the foundation. The fiber optic sensors achieve distributed strain monitoring through optical time-domain reflectometry.

[0006] Specifically, the acoustic emission sensor array is arranged such that acoustic emission sensors are evenly distributed around the four sides of the foundation, and the location of the sound source is determined by a time difference positioning algorithm between the sensors.

[0007] Specifically, the dynamic constitutive model considering the soil strain rate effect adopts the modified Cambridge model as the basic framework and introduces strain rate parameters and cyclic softening parameters to improve the model.

[0008] The specific steps of the explicit dynamic time history analysis method are as follows: establish a three-dimensional finite element model of the foundation, use the measured vibration acceleration time history as the load input, and use an explicit integration algorithm to calculate the dynamic time history.

[0009] The process includes establishing a dynamic constitutive model that takes into account the effect of soil strain rate, followed by a step of inverting and verifying the model parameters, using a genetic algorithm to invert and optimize the soil constitutive model parameters.

[0010] The interface gap refers to the debonding and separation area between the bottom of the foundation and the contact surface of the foundation soil caused by vibration.

[0011] Specifically, the structure of the interface gap identification model is as follows: the input layer receives the interface strain time history data of the fiber optic strain sensing point and the acoustic emission signal characteristic parameters of the acoustic emission sensor; the three hidden layers use rectified linear units as activation functions; the third hidden layer introduces an attention mechanism; and the output layer outputs the gap probability, gap width, and gap development rate of the potential gap region at the bottom of the foundation.

[0012] Specifically, the steps for establishing the training dataset for the interface gap recognition model involve setting up a scaled-down foundation vibration test bench in the laboratory, artificially prefabricating gaps of different widths at the bottom of the foundation, applying vibration loads of different frequencies and amplitudes to the test bench, simultaneously collecting fiber optic strain data and acoustic emission signals, and measuring the actual location and width of the gaps using a high-precision three-dimensional laser scanner as label data.

[0013] Specifically, the interface gap recognition model training step uses mini-batch gradient descent for training, and the loss function is a weighted combination of mean squared error and cross-entropy. A dropout method is introduced to prevent overfitting.

[0014] The grouting parameter optimization game model includes an upper-level model aimed at minimizing foundation disturbance and a lower-level model aimed at maximizing the density of gap filling. The objective function of the upper-level model is used to minimize the additional displacement and stress disturbance generated in the foundation during grouting. Inputs include optimized grouting pressure, optimized grouting speed, and gap width; the output is the foundation disturbance index. The objective function of the lower-level model is used to maximize the density of gap filling by the grout during grouting. Inputs include optimized grouting pressure, optimized grouting speed, and gap development rate; the output is the density of gap filling.

[0015] Specifically, the solution steps of the grouting parameter optimization game model are as follows: set initial values ​​for optimized grouting pressure and optimized grouting speed based on the gap width and gap development rate; fix the optimized grouting pressure and solve for the optimized grouting speed that maximizes the gap filling density in the lower-level model; then fix the optimized grouting speed and solve for the optimized grouting pressure that minimizes the foundation disturbance index in the upper-level model; and iterate repeatedly until the changes in optimized grouting pressure and optimized grouting speed are both less than the set threshold.

[0016] The high-fluidity micro-expansion slurry is prepared by mixing cement, fly ash, bentonite, water-reducing agent and water in a specific mass ratio. The water-reducing agent used is a polycarboxylate-based high-efficiency water-reducing agent.

[0017] The real-time monitoring and control method for the grouting process specifically involves using low-pressure, slow-speed grouting in the initial stage to allow the grout to fully penetrate into the gaps. When the outlet pressure of the grouting pipe rises to a set ratio of the set pressure, grouting is performed according to the calculated values ​​of the optimized grouting pressure and optimized grouting speed. Grouting is stopped when the outlet pressure of the grouting pipe reaches the set ratio of the optimized grouting pressure and remains stable for a set time. The displacement of the foundation surface is monitored in real time during the grouting process.

[0018] This invention establishes a collaborative monitoring system combining a distributed fiber optic strain sensor network and an acoustic emission sensor array. It acquires interface strain time-history data and debonding acoustic signals at a high frequency of 10000 Hz, inputting multi-source monitoring information into a deep neural network model containing three hidden layers and an attention mechanism. This enables real-time and accurate identification of the gap's location coordinates, width, and development rate, overcoming the shortcomings of traditional discrete monitoring methods that suffer from low spatiotemporal resolution and large errors in gap feature parameter identification. Furthermore, this invention constructs a two-layer game-theoretic optimization model with minimizing foundation disturbance as the upper-level objective and maximizing gap filling density as the lower-level objective. Based on the identified gap width and development rate, iterative solutions are used to obtain optimized grouting pressure and speed, achieving adaptive dynamic adjustment of grouting parameters according to the gap state. This overcomes the limitations of fixed-parameter grouting methods, which cannot simultaneously address filling effect and foundation disturbance control. In summary, this invention solves the technical problem mentioned in the background art of accurately identifying and reinforcing interface gaps at the bottom of high-frequency, high-intensity vibration equipment foundations. Attached Figure Description

[0019] Figure 1 This is a time-history curve of the gap development rate in the embodiment.

[0020] Figure 2 This is a graph showing the foundation displacement monitoring during the grouting process in the embodiment.

[0021] Figure 3 The graph shows the pressure variation curves of the grouting in the three gaps in the embodiment. Detailed Implementation

[0022] 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.

[0023] This invention provides a method for reinforcing equipment foundations under high-frequency, high-intensity vibration with micro-displacement, comprising the following steps:

[0024] S01. A distributed fiber optic strain sensor network is pre-embedded at the bottom of the foundation of the wind tunnel drive system, with a sensor spacing of 200mm. At the same time, an acoustic emission sensor array is arranged around the foundation, with no less than 16 sensors.

[0025] S02. Establish a dynamic constitutive model that considers the soil strain rate effect, and use the explicit dynamic time history analysis method to perform finite element simulation of the foundation-subsoil system. The mesh size is controlled within 50 mm, and the time step is set to 0.0001 s.

[0026] S03. During the operation of the wind tunnel drive system, the distributed fiber optic strain sensor network collects the strain data of the interface at the bottom of the foundation in real time. The acquisition frequency is set to 10000Hz. The acoustic emission sensor records the debonding acoustic signal simultaneously to obtain the interface strain time history data and acoustic emission signal characteristic parameters.

[0027] S04. Input the interface strain time history data and acoustic emission signal characteristic parameters into the interface gap identification model. The interface gap identification model outputs the coordinates of the gap location at the bottom of the foundation, the gap width, and the gap development rate.

[0028] S05. When five consecutive test results show that the gap width exceeds 0.05mm, it is determined to be a continuous abnormal state and the emergency grouting process is initiated; when only a single test result exceeds 0.05mm, the test frequency is increased to 20000Hz for secondary confirmation.

[0029] S06. Based on the coordinates of the gap location, drill grouting holes in the corresponding area at the bottom of the foundation. The hole diameter is 30mm, and the hole depth penetrates the foundation thickness and enters the foundation soil 50mm deep to obtain the grouting hole location parameters.

[0030] S07. Input the gap width, gap development rate and grouting hole location parameters into the grouting parameter optimization game model. The grouting parameter optimization game model outputs the optimized grouting pressure and optimized grouting speed.

[0031] S08. Inject high-flowability micro-expansion grout into the crack area through the grouting hole, and perform grouting according to the optimized grouting pressure and optimized grouting speed. The optimized grouting pressure is controlled between 0.3MPa and 0.8MPa, and the optimized grouting speed is between 5L / min and 15L / min.

[0032] S09. 72 hours after grouting is completed, foundation displacement monitoring is carried out. High-precision displacement sensors are used to measure horizontal displacement and vertical settlement. The measurement accuracy reaches 0.01mm to verify whether the reinforcement effect meets the control requirements of horizontal displacement less than 0.1mm and vertical settlement less than 4mm.

[0033] The distributed fiber optic strain sensor network is deployed as follows: before the concrete is poured at the bottom of the foundation, the fiber optic sensors are arranged in a grid pattern with a longitudinal and transverse spacing of 200 mm. The fiber optic sensors are protected with polyimide coating, with a temperature resistance range of -40℃ to 180℃ and a strain measurement range of -10000με to 10000με. The sensors achieve distributed strain monitoring through optical time-domain reflectometry.

[0034] The acoustic emission sensor array is arranged as follows: 16 to 24 acoustic emission sensors are evenly arranged on the four sides of the foundation. The operating frequency range of the sensors is 20kHz to 400kHz, and the sensitivity is not less than -65dB. The sensors are fixed to the foundation surface by magnetic attraction or bolts. The location of the sound source is determined by the time difference positioning algorithm between the sensors.

[0035] The dynamic constitutive model considering the soil strain rate effect is as follows: the modified Cambridge model is used as the basic framework, and strain rate parameters and cyclic softening parameters are introduced to improve the model. The strain rate parameters are obtained through dynamic triaxial tests of soil, with the test loading frequency covering the range of 10Hz to 200Hz. The cyclic softening parameters are determined through cyclic loading and unloading tests, with no less than 1000 test cycles. The model parameters include soil elastic modulus, Poisson's ratio, internal friction angle, cohesion, strain rate sensitivity coefficient, and cyclic softening coefficient.

[0036] The implementation steps of the explicit dynamic time history analysis method are as follows: First, a three-dimensional finite element model of the foundation-subsoil is established. The foundation is simulated using solid elements, and the subsoil area is taken as 3 times the size of the foundation. The boundary is simulated using a viscoelastic artificial boundary to simulate the infinite domain effect. Then, the vibration acceleration time history measured in the field is used as the load input, and the time history length is not less than 60s. The dynamic time history is calculated using an explicit integration algorithm. The calculation time step is determined according to the minimum size of the element and the material wave velocity. The stress-strain time history and displacement response at the bottom of the foundation are output.

[0037] The steps for model parameter inversion verification are as follows: During the trial operation of the wind tunnel drive system, accelerometers and displacement sensors are arranged on the foundation surface and in the foundation soil to measure the actual vibration response and displacement data; the measured data are compared with the finite element simulation results to calculate the response error; the genetic algorithm is used to invert and optimize the soil constitutive model parameters, and the optimization objective function is to minimize the root mean square error between the measured response and the simulated response; the process is iterated repeatedly until the error converges to within 10%, thus completing the model parameter verification.

[0038] Among them, the interface gap refers to the debonding and separation area between the bottom of the foundation and the contact surface of the foundation soil caused by vibration, and the gap width ranges from 0.01mm to 2.00mm.

[0039] The interface gap recognition model has the following structure: the input layer contains 96 neurons, which receive interface strain time history data from 12 fiber optic strain sensing points and acoustic emission signal characteristic parameters from 16 acoustic emission sensors; the first hidden layer contains 256 neurons, which use rectified linear units as activation functions; the second hidden layer contains 128 neurons, which also use rectified linear units for activation; the third hidden layer contains 64 neurons, which introduce an attention mechanism to enhance the ability to extract key features; the output layer contains 32 neurons, which output the gap probability, gap width, and gap development rate of 32 potential gap regions at the bottom of the base; the model is equipped with a dynamic monitoring mechanism. When the output gap development rate exceeds the preset threshold of 0.15 mm / h, the first-level early warning adjustment mechanism is immediately activated and the monitoring operation mode is switched to the protection state. When the gap development rate continues to rise and exceeds the critical value of 0.30 mm / h, the second-level emergency compensation algorithm is simultaneously activated to prevent system overload.

[0040] The steps for establishing the training dataset for the interface gap recognition model are as follows: First, a foundation-subsoil vibration test bench with a scale of 1:5 is built in the laboratory. Gaps of different widths are artificially prefabricated at the bottom of the foundation, ranging from 0.01mm to 2.00mm, and the gap locations cover different areas at the bottom of the foundation. Then, vibration loads of different frequencies and amplitudes are applied to the test bench, with frequencies ranging from 20Hz to 150Hz and acceleration amplitudes ranging from 0.5g to 5.0g. Fiber optic strain data and acoustic emission signals are collected simultaneously, with each test set lasting 300s. The actual location and width of the gaps are measured using a high-precision 3D laser scanner as label data. A total of 500 sets of tests under different working conditions are conducted, resulting in 150,000 labeled samples, which are divided into training, validation, and test sets in a ratio of 8:1:1.

[0041] The training steps for the interface gap recognition model are as follows: Mini-batch gradient descent with a batch size of 64 is used for training. The initial learning rate is set to 0.001, and it decays to 90% of the original value every 30 training epochs. The loss function uses a weighted combination of mean squared error and cross-entropy, with weight coefficients of 0.6 and 0.4, respectively. A dropout method is introduced to prevent overfitting, with a dropout rate of 0.3. The total number of training epochs is 200. After each epoch, the model performance is evaluated on the validation set, and the model parameters are recorded when the validation loss is minimized. Training is stopped early when the validation loss does not decrease for 15 consecutive epochs. After training, the model is evaluated on the test set, requiring a gap location recognition accuracy of over 92% and a gap width prediction relative error of less than 8%.

[0042] The grouting parameter optimization game model includes an upper-level model that aims to minimize basic disturbance and a lower-level model that aims to maximize the density of gap filling.

[0043] The objective function of the upper-level model is used to minimize the additional displacement and stress disturbance generated in the foundation during the grouting process. The inputs include optimized grouting pressure, optimized grouting speed, and gap width, and the output is the foundation disturbance index. The upper-level objective function is expressed as follows: the foundation disturbance index divided by the standard disturbance value equals the optimized grouting pressure divided by the square of the baseline grouting pressure multiplied by the optimized grouting speed divided by the baseline grouting speed plus the gap width divided by the maximum gap width multiplied by the grouting pressure-speed coupling coefficient. The unit of the foundation disturbance index is mm, the standard disturbance value is 0.05 mm, the unit of the optimized grouting pressure is MPa, the baseline grouting pressure is 0.5 MPa, the unit of the optimized grouting speed is L / min, the baseline grouting speed is 10 L / min, the unit of the gap width is mm, the maximum gap width is 2.00 mm, and the grouting pressure-speed coupling coefficient is a dimensionless coefficient with a value of 0.3.

[0044] The constraints of the upper-level model include: the optimized grouting pressure is not less than 0.3 MPa and not more than 0.8 MPa; the optimized grouting speed is not less than 5 L / min and not more than 15 L / min; and the basic disturbance index is not more than 0.05 mm.

[0045] The objective function of the lower-level model is used to maximize the filling density of the cracks by the grout during the grouting process. The inputs include optimized grouting pressure, optimized grouting speed, and crack development rate, and the output is the crack filling density. The lower-level objective function is expressed as follows: the crack filling density is equal to the optimized grouting pressure divided by the baseline grouting pressure multiplied by the optimized grouting speed divided by the square root of the baseline grouting speed minus the crack development rate divided by the maximum development rate multiplied by the grouting aging decay coefficient. The crack filling density is a dimensionless index with a value range of 0 to 1. The unit of optimized grouting pressure is MPa, and the baseline grouting pressure is 0.5 MPa. The unit of optimized grouting speed is L / min, and the baseline grouting speed is 10 L / min. The unit of crack development rate is mm / h, and the maximum development rate is 0.30 mm / h. The grouting aging decay coefficient is a dimensionless coefficient with a value of 0.25.

[0046] The constraints of the lower-level model include: the optimized grouting pressure is not less than 0.3 MPa and not more than 0.8 MPa; the optimized grouting speed is not less than 5 L / min and not more than 15 L / min; and the gap filling density is not less than 0.85.

[0047] The grouting pressure-velocity coupling coefficient is the coupling term coefficient between the upper-level objective function and the lower-level objective function. It is used to characterize the degree of synergistic influence of grouting pressure and grouting velocity on foundation disturbance. When the optimized grouting pressure is high and the optimized grouting velocity is fast, the foundation disturbance increases, and the density of the gap filling increases. The two objectives can be balanced and optimized by adjusting the grouting pressure-velocity coupling coefficient.

[0048] The solution steps of the game model for optimizing grouting parameters are as follows: First, set the initial values ​​of the optimized grouting pressure and optimized grouting speed according to the gap width and gap development rate; then, fix the optimized grouting pressure and solve for the optimized grouting speed that maximizes the gap filling density in the lower-level model; then, fix the optimized grouting speed and solve for the optimized grouting pressure that minimizes the foundation disturbance index in the upper-level model; iterate the above process repeatedly until the changes in the optimized grouting pressure and optimized grouting speed are both less than 1%, and output the final optimized grouting pressure and optimized grouting speed.

[0049] The high-fluidity micro-expansion slurry is prepared by mixing cement, fly ash, bentonite, water-reducing agent, and water according to the following mass ratio: 425 ordinary Portland cement is used; the fly ash fineness is no greater than 45μm; the bentonite sodiumization degree is no less than 85%; the water-reducing agent is a polycarboxylate-based high-efficiency water-reducing agent; and the water-cement ratio is controlled between 0.45 and 0.60. The initial setting time of the slurry is controlled between 4h and 8h, the final setting time is controlled between 12h and 24h, the 28-day compressive strength is no less than 15MPa, the slurry expansion rate is 0.1% to 0.5%, and the fluidity is no less than 280mm.

[0050] The positioning accuracy requirements for drilling grouting holes are as follows: based on the coordinates of the gap location, a total station is used to lay out the grouting hole positions, with the layout error controlled within 10mm; diamond drill bits are used for drilling, and the drilling speed is controlled between 0.5m / min and 1.5m / min to avoid disturbing the foundation and subgrade during the drilling process; after drilling is completed, an endoscope is used to check the hole wall quality to ensure that the hole wall is flat and free of cracks, and to complete the acquisition of the grouting hole position parameters.

[0051] The real-time monitoring and control method for the grouting process is as follows: In the initial stage of grouting, low-pressure, slow-speed grouting is adopted, with the grouting pressure controlled at 0.3 MPa and the grouting speed controlled at 5 L / min, lasting for 10 to 20 minutes, to allow the grout to fully penetrate into the gaps; when the outlet pressure of the grouting pipe rises to 70% of the set pressure, grouting is carried out according to the calculated values ​​of the optimized grouting pressure and optimized grouting speed; when the outlet pressure of the grouting pipe reaches 95% of the optimized grouting pressure and remains stable for 3 minutes, grouting is stopped; during the grouting process, the displacement of the foundation surface is monitored in real time, and when the sudden change in displacement exceeds 0.05 mm, the grouting pressure is immediately reduced or grouting is suspended.

[0052] The steps for verifying the reinforcement effect are as follows: 72 hours after grouting, 8 to 12 high-precision displacement sensors are arranged on the top surface of the foundation, with the measuring points covering the edge and center area of ​​the foundation; the horizontal displacement and vertical settlement of the foundation are continuously monitored under the operating conditions of the wind tunnel drive system, with a monitoring time of not less than 168 hours; the maximum and average values ​​of the displacement during the monitoring period are calculated to determine whether the control requirements of horizontal displacement being less than 0.1 mm and vertical settlement being less than 4 mm are met; at the same time, a distributed fiber optic strain sensor network is used to monitor the strain distribution at the bottom of the foundation to verify whether the gaps are effectively filled and compacted.

[0053] The specific implementation methods of the above steps are described in detail below.

[0054] The specific implementation of step S01 involves pre-embedding a distributed fiber optic strain sensor network and arranging an acoustic emission sensor array at the bottom of the wind tunnel drive system foundation to achieve comprehensive monitoring of the foundation bottom strain and debonding acoustic signals. Firstly, before pouring the concrete at the foundation bottom, fiber optic sensors protected by a polyimide coating are arranged in a grid pattern with a longitudinal and transverse spacing of 200mm, forming a distributed sensing network. The fiber optic sensors have a temperature resistance range of -40℃ to 180℃ and a strain measurement range of -10000με to 10000με. The sensors achieve distributed strain monitoring through optical time-domain reflectometry (OTDR), which utilizes the time delay and intensity variation of the Rayleigh scattering signal in the optical fiber to determine the strain distribution, enabling strain measurement with continuous spatial resolution. Then, 16 to 24 acoustic emission sensors are evenly arranged around the four sides of the foundation. The sensors operate in the frequency range of 20kHz to 400kHz, with a sensitivity of no less than -65dB. The sensors are fixed to the foundation surface using magnetic or bolt methods. The location of the sound source is determined by a time-difference positioning algorithm. This algorithm is based on the propagation speed of sound in a solid medium and the time difference between the signals received by each sensor, and calculates the coordinates of the debonding position using the principle of triangulation. The fiber optic strain sensor network can capture minute strain changes at the bottom interface of the foundation, while the acoustic emission sensor array can identify the elastic wave signals generated during interface debonding. The collaborative work of the two sensing systems provides multi-source data support for subsequent gap identification.

[0055] The specific implementation of step S02 involves establishing a dynamic constitutive model considering the soil strain rate effect and performing finite element simulation to predict the dynamic response of the foundation and the stress-strain distribution at the interface under high-frequency vibration. First, a modified Cambridge model is used as the foundation framework. This model can describe the elasto-plastic deformation behavior of soil under cyclic loading. Strain rate parameters and cyclic softening parameters are introduced to improve the model. The strain rate parameters are obtained through dynamic triaxial tests of soil, with test loading frequencies covering the range of 10Hz to 200Hz. The cyclic softening parameters are determined through cyclic loading and unloading tests, with at least 1000 test cycles. Model parameters include the soil elastic modulus, Poisson's ratio, internal friction angle, cohesion, strain rate sensitivity coefficient, and cyclic softening coefficient. Then, a three-dimensional finite element model of the foundation-subsoil is established. The foundation is simulated using solid elements, and the subsoil area is taken as three times the foundation size to ensure that boundary effects do not affect the calculation results. Viscoelastic artificial boundaries are used to simulate infinite domain effects; this boundary condition can absorb outwardly propagating wave energy and avoid false reflections. The mesh size is controlled within 50mm to meet the requirements of high-frequency vibration wavelength. The measured vibration acceleration time history is used as the load input, with a time history length of no less than 60 seconds. An explicit integration algorithm is used for dynamic time history calculation, with a time step set to 0.0001 seconds to ensure calculation stability. The time step is determined based on the minimum element size and material wave velocity to satisfy the Coulomb stability condition. The stress-strain time history and displacement response at the bottom of the foundation are output. During the trial operation of the wind tunnel drive system, acceleration and displacement sensors are deployed on the foundation surface and in the foundation soil to measure the actual vibration response and displacement data. The measured data are compared with the finite element simulation results to calculate the response error. A genetic algorithm is used to inversely optimize the soil constitutive model parameters. The optimization objective function is to minimize the root mean square error between the measured and simulated responses. The genetic algorithm searches for the optimal parameter combination through simulated natural selection and genetic mechanisms, iterating repeatedly until the error converges to within 10% to complete the model parameter verification, ensuring that the numerical model can accurately predict the actual engineering response.

[0056] The specific implementation of step S03 involves real-time acquisition of interface strain data and acoustic emission signals at the bottom of the foundation during the operation of the wind tunnel drive system, providing raw monitoring data for gap identification. A distributed fiber optic strain sensor network acquires interface strain data at the bottom of the foundation in real time at a sampling frequency of 10000Hz. This frequency can capture strain fluctuations caused by high-frequency vibrations. Acoustic emission sensors simultaneously record debonding acoustic signals. When interface debonding occurs, the elastic strain energy stored within the material is rapidly released, generating transient stress waves. The acoustic emission sensors can capture these high-frequency stress wave signals, obtaining interface strain time history data and acoustic emission signal characteristic parameters. These acoustic emission signal characteristic parameters include amplitude, rise time, duration, energy, and frequency components. These parameters reflect the degree and development state of interface debonding. The high sampling frequency ensures accurate capture of strain changes and debonding events caused by vibration, providing a high-quality data source for subsequent intelligent identification models.

[0057] The specific implementation of step S04 involves inputting the interface strain time history data and acoustic emission signal feature parameters into the interface gap recognition model, and using deep learning methods to identify the location, width, and development rate of gaps at the bottom of the foundation. The interface gap recognition model employs a multi-layer neural network structure. The input layer contains 96 neurons, receiving interface strain time history data from 12 fiber optic strain sensing points and acoustic emission signal feature parameters from 16 acoustic emission sensors, achieving multi-source information fusion processing. The first hidden layer contains 256 neurons, using rectified linear units as the activation function. This function introduces nonlinear mapping capabilities while maintaining gradient propagation efficiency. The second hidden layer contains 128 neurons, also activated by rectified linear units. The third hidden layer contains 64 neurons and introduces an attention mechanism to enhance key feature extraction capabilities. The attention mechanism adaptively assigns different weights to different sensor data to highlight important information. The output layer contains 32 neurons, outputting the gap probability, gap width, and gap development rate of 32 potential gap regions at the bottom of the foundation. The model training dataset was established using a 1:5 scale foundation-ground vibration test bench in the laboratory. Artificial prefabricated gaps with widths ranging from 0.01 mm to 2.00 mm were placed at the bottom of the foundation. Vibration loads with frequencies ranging from 20 Hz to 150 Hz and acceleration amplitudes ranging from 0.5 g to 5.0 g were applied. Fiber optic strain data and acoustic emission signals were collected simultaneously. The actual location and width of the gaps were measured using a high-precision 3D laser scanner as label data, resulting in a total of 150,000 labeled samples. The model training employed mini-batch gradient descent with a batch size of 64. The initial learning rate was set to 0.001 and decayed to 90% of its original value every 30 training epochs. The loss function used a weighted combination of mean squared error and cross-entropy with weight coefficients of 0.6 and 0.4. A dropout mechanism was introduced to prevent overfitting, with a dropout rate of 0.3. The total number of training epochs was 200, and an early stopping strategy was implemented on the validation set. Training was stopped early when the validation loss did not decrease for 15 consecutive epochs. Ultimately, the gap location identification accuracy on the test set reached over 92%, and the relative error of gap width prediction was less than 8%. A dynamic monitoring mechanism was incorporated into the model. When the output gap development rate exceeded the preset threshold of 0.15 mm / h, the first-level early warning adjustment mechanism was immediately activated. When the gap development rate continued to rise and exceeded the critical value of 0.30 mm / h, the second-level emergency compensation algorithm was simultaneously activated to prevent system overload.

[0058] The specific implementation of step S05 involves determining the abnormal state based on the gap width output by the interface gap identification model. This aims to distinguish between sporadic measurement fluctuations and continuous structural deterioration. When five consecutive test results show a gap width exceeding 0.05 mm, this threshold, determined based on the allowable debonding width at the foundation-subsoil interface, is classified as a continuous abnormal state, and an emergency grouting process is initiated. This continuity determination criterion avoids false alarms caused by single measurement errors. When only a single test result exceeds 0.05 mm, the detection frequency is increased to 20000 Hz for secondary confirmation. By increasing the sampling frequency, more detailed strain fluctuation information is obtained, verifying the authenticity of the abnormal state. This strategy avoids unnecessary reinforcement work while ensuring safety.

[0059] The specific implementation of step S06 involves drilling grouting holes in the corresponding area at the bottom of the foundation according to the coordinates of the gap location, providing a channel for subsequent grout injection. First, a total station is used to lay out the grouting hole positions according to the gap location coordinates, with the layout error controlled within 10mm to ensure the grouting holes accurately correspond to the gap locations. Then, a diamond drill bit is used to drill the grouting holes with a diameter of 30mm to meet the requirements for smooth grout injection. The drilling speed is controlled between 0.5m / min and 1.5m / min to avoid disturbing the foundation and subgrade during drilling. The hole depth penetrates the foundation thickness and enters the subgrade soil 50mm deep, allowing the grout to fully penetrate into the gap area. After drilling, an endoscope is used to inspect the hole wall quality to ensure the hole wall is smooth and free of cracks, obtaining the grouting hole location parameters to provide input conditions for grouting parameter optimization.

[0060] The specific implementation of step S07 involves inputting the gap width, gap development rate, and grouting hole location parameters into a grouting parameter optimization game model, and solving for the optimal grouting pressure and grouting speed through a two-layer optimization. The grouting parameter optimization game model includes an upper-layer model aimed at minimizing foundation disturbance and a lower-layer model aimed at maximizing gap filling density. The upper-layer model aims to minimize the additional displacement and stress disturbance generated in the foundation during grouting. Inputs include optimized grouting pressure, optimized grouting speed, and gap width; the output is a foundation disturbance index, which is related to the square of the optimized grouting pressure, optimized grouting speed, and gap width. The grouting pressure-speed coupling coefficient is set to 0.3 to characterize the degree of synergistic influence of grouting pressure and grouting speed on foundation disturbance. The goal of the lower-level model is to maximize the filling density of the cracks during grouting. Inputs include optimized grouting pressure, optimized grouting speed, and crack development rate. Output is the crack filling density. The crack filling density is positively correlated with the square root of the optimized grouting pressure and optimized grouting speed, and negatively correlated with the crack development rate. The grouting aging attenuation coefficient is set to 0.25, reflecting the weakening effect of continuous crack development on the filling effect. The model solution employs an iterative optimization strategy. First, initial values ​​for optimized grouting pressure and optimized grouting speed are set based on the crack width and crack development rate. Then, with the optimized grouting pressure fixed, the lower-level model solves for the optimized grouting speed that maximizes the crack filling density. Next, with the optimized grouting speed fixed, the upper-level model solves for the optimized grouting pressure that minimizes the foundation disturbance index. This process is iterated until the changes in optimized grouting pressure and optimized grouting speed are both less than 1%, at which point the final optimized grouting pressure and optimized grouting speed are output. The constraints include an optimized grouting pressure of not less than 0.3 MPa and not more than 0.8 MPa, an optimized grouting speed of not less than 5 L / min and not more than 15 L / min, a foundation disturbance index of not more than 0.05 mm, and a gap filling density of not less than 0.85. These constraints ensure that the grouting parameters are within the feasible range of the project and meet the requirements of the reinforcement effect.

[0061] The specific implementation of step S08 involves injecting a highly fluid micro-expansion grout into the crack area through grouting holes, and performing grouting according to optimized grouting pressure and speed to effectively fill the crack. The highly fluid micro-expansion grout is prepared by mass ratio of 425 ordinary Portland cement, fly ash with a fineness of no more than 45μm, bentonite with a sodium degree of no less than 85%, polycarboxylate-based high-efficiency water-reducing agent, and water. The water-cement ratio is controlled between 0.45 and 0.60, the initial setting time is controlled between 4h and 8h, the final setting time is controlled between 12h and 24h, the 28-day compressive strength is no less than 15MPa, the grout expansion rate is 0.1% to 0.5%, which can compensate for shrinkage during the hardening process and generate pre-stress on the crack, and the fluidity is no less than 280mm to ensure that the grout can fully penetrate into the narrow crack. In the initial stage of grouting, low-pressure, slow-speed grouting is employed, with the grouting pressure controlled at 0.3 MPa and the grouting speed at 5 L / min for 10 to 20 minutes. This allows the grout to fully penetrate the gaps, avoiding foundation disturbance caused by high-pressure, rapid grouting. When the outlet pressure of the grouting pipe rises to 70% of the set pressure, grouting is carried out according to the calculated values ​​of the optimized grouting pressure and speed. Grouting is stopped when the outlet pressure of the grouting pipe reaches 95% of the optimized grouting pressure and remains stable for 3 minutes, indicating that the grout has filled the gap area. During the grouting process, the foundation surface displacement is monitored in real time. If the sudden change in displacement exceeds 0.05 mm, the grouting pressure is immediately reduced or grouting is paused. This control strategy, based on displacement feedback, achieves dynamic regulation of the grouting process to ensure foundation safety.

[0062] The specific implementation of step S09 involves monitoring the foundation displacement 72 hours after grouting to verify the reinforcement effect. The purpose is to assess whether the grouting reinforcement meets the displacement control requirements. Eight to twelve high-precision displacement sensors are arranged on the top surface of the foundation. The measuring points cover the foundation's edge and center areas to comprehensively reflect the foundation's deformation state. High-precision displacement sensors with a measurement accuracy of 0.01 mm are used to measure horizontal displacement and vertical settlement. The foundation's displacement response is continuously monitored under wind tunnel drive system operating conditions for at least 168 hours to cover multiple operating cycles. The maximum and average displacement values ​​during the monitoring period are calculated to determine whether the control requirements of horizontal displacement less than 0.1 mm and vertical settlement less than 4 mm are met. This control standard is determined based on the accuracy requirements of the wind tunnel drive system for foundation displacement. Simultaneously, a distributed fiber optic strain sensor network is used to monitor the strain distribution at the bottom of the foundation. The continuity and uniformity of the strain field verify whether the gaps are effectively filled and compacted. If the strain distribution is uniform and there is no local stress concentration, it indicates that the gaps have been filled with grout, eliminating the stress fracture surface.

[0063] The key technical ideas of this invention include multi-source information fusion monitoring based on a distributed fiber optic strain sensor network and an acoustic emission sensor array, intelligent identification of interface gaps based on a deep learning attention mechanism, and collaborative optimization of grouting parameters based on a two-layer game model. The multi-source information fusion monitoring technology achieves continuous spatial resolution measurement of interface strain through the principle of optical time-domain reflectometry. Combined with an acoustic emission time-difference positioning algorithm, it captures transient elastic wave signals of debonding events. Compared with traditional single-sensor monitoring, it can comprehensively characterize the interface state from both strain field and sound field dimensions, significantly improving the early identification capability and positioning accuracy of minute gaps. The deep learning attention mechanism identification technology extracts deep features from strain time histories and acoustic emission signals through multi-layer neural networks. The attention mechanism adaptively assigns weights to different sensor data to highlight key information. Compared with traditional threshold discrimination methods, it can accurately identify gap features under complex vibration backgrounds, achieving quantitative prediction of gap location, width, and development rate. The two-layer game theory model optimization technique achieves coordinated optimization of grouting pressure and grouting speed through iterative solutions of minimizing foundation disturbance in the upper-layer model and maximizing filling density in the lower-layer model. Compared with traditional empirical parameter selection methods, it can minimize the disturbance impact of the grouting process on the foundation of operating equipment while ensuring reinforcement quality. The synergistic effect of the three technical approaches is that multi-source fusion monitoring provides high-quality data support for intelligent identification, the gap parameters output by intelligent identification provide accurate input conditions for grouting optimization, and the optimized grouting parameters ensure the safety and controllability of the reinforcement process. This forms a closed-loop reinforcement system from monitoring and identification to parameter optimization and effect verification, which significantly improves the intelligence level and engineering reliability of micro-displacement reinforcement of equipment foundations under high-frequency and high-intensity vibration environments compared with traditional methods.

[0064] It should be noted that this invention also solves the following technical problem: the difficulty in predicting the dynamic evolution of interface cracks at the bottom of the foundation under high-frequency and high-intensity vibration. Traditional finite element analysis methods typically employ static or quasi-static assumptions, failing to fully consider the strain rate effect and cyclic softening characteristics of the soil under cyclic loading, resulting in insufficient accuracy in predicting crack development trends. This invention establishes a dynamic constitutive model considering the soil strain rate effect, introduces strain rate parameters and cyclic softening parameters to improve the modified Cambridge model, and uses an explicit dynamic time history analysis method to perform refined simulation of the foundation system, controlling the mesh size within 50 mm and setting the time step to 0.0001 seconds. Simultaneously, a genetic algorithm is used to inversely verify the model parameters using field measured data, ensuring that the finite element calculation results accurately reflect the stress-strain state and crack development law at the bottom of the foundation. This provides reliable theoretical support for generating training samples for the crack identification model and optimizing grouting parameters, thereby achieving accurate prediction and early warning of the dynamic evolution process of cracks.

[0065] Specifically, the principle of this invention is as follows: The invention solves this technical problem by constructing a complete technical chain from high-density real-time monitoring to intelligent identification and adaptive optimization. A high-density monitoring grid formed by distributed fiber optic strain sensors with a 200mm spacing can fully cover the bottom interface of the foundation. A 10,000Hz acquisition frequency ensures the capture of transient changes in the strain field during vibration. An acoustic emission sensor array uses a time-difference positioning algorithm to pinpoint the location of the debonding sound source to the centimeter level. The strain time-history data and acoustic emission characteristic parameters acquired by the two types of sensors are used as complementary information and input into a deep neural network model. A three-layer hidden layer structure composed of 256, 128, and 64 neurons is used to progressively extract the spatial distribution characteristics and temporal evolution of the gap. An attention mechanism enhances the model's ability to focus on key features, thereby increasing the gap location identification accuracy to over 92% and controlling the width prediction relative error to within 8%. The two-layer game optimization model sets minimizing foundation disturbance and maximizing gap filling density as mutually constraining optimization objectives. By fixing one variable and solving the other through iteration, the optimal solution is searched in the parameter space of grouting pressure and grouting speed. This ensures that the grouting process can both guarantee that the grout fully penetrates into the gap to achieve dense filling and control the additional foundation displacement caused by grouting within the allowable range, thus realizing dynamic optimization reinforcement based on accurate gap identification.

[0066] The following provides a specific embodiment 1 of the present invention, and the specific implementation of each step in this embodiment 1 is described in detail below.

[0067] The specific implementation methods of steps S01, S03, S05, S06, S08 and S09 are the same as those described above, and will not be repeated in detail here.

[0068] The specific implementation of step S02 involves establishing a dynamic constitutive model considering the soil strain rate effect and performing finite element simulation to predict the dynamic response of the foundation and the interface stress-strain distribution under high-frequency vibration. First, a modified Cambridge model is used as the basic framework, and strain rate parameters and cyclic softening parameters are introduced to improve the model. The soil stress-strain relationship is expressed as follows:

[0069] ;

[0070] In the formula, The total stress tensor of the soil, in units of ; These are static stress components, in units of... It was obtained through static triaxial testing; This represents the dynamic stress increment, in units of... ; The reference stress value is set to 1. .

[0071] The dynamic stress increment takes into account the strain rate effect, as specifically expressed below:

[0072] ;

[0073] In the formula, This represents the dynamic stress increment, in units of... ; The reference stress is set to 1. ; This refers to the elastic modulus of soil, in units of... The results were obtained through dynamic triaxial tests on soil, with test loading frequencies covering 10... Up to 200 scope; The reference modulus is set to 100. ; This is the actual strain rate, in units of... ; The reference strain rate is set to a value of ; The strain rate sensitivity coefficient is a dimensionless parameter obtained through fitting dynamic triaxial tests, with an empirical value of 0.15 to 0.35. The cyclic softening coefficient is a dimensionless parameter determined through cyclic loading and unloading tests, with at least 1000 test cycles. The empirical value is 0.08 to 0.20. The loop count is a dimensionless parameter. This is the cyclic decay exponent, a dimensionless parameter, with a default value of 0.4.

[0074] Then, a three-dimensional finite element model of the foundation-subsoil system is established, with the mesh size controlled at 50. Within this range, the time step is determined based on the Courant stability condition, as specifically expressed below:

[0075] ;

[0076] In the formula, To calculate the time step, the unit is... The value is 0.0001. ; For reference time, the value is 1. ; This is the Courant number, a dimensionless parameter with a default value of 0.8. This is the minimum grid size, in units of The value is 0.05. ; The unit is the soil compression wave velocity. The value was obtained through wave velocity testing experiments, with an empirical value ranging from 150 to 400. .

[0077] During the trial operation of the wind tunnel drive system, a genetic algorithm was used to invert and optimize the parameters of the soil constitutive model. The optimization objective function is expressed as follows:

[0078] ;

[0079] In the formula, To optimize the objective function value, dimensionless parameters are used, and the convergence requirement is less than 0.1; The number of measurement points is a dimensionless parameter. For the first Simulated acceleration response at each measuring point, in units of ; For the first The measured acceleration response at each measuring point, in units of ; This represents the maximum measured acceleration response, in units of... .

[0080] The specific implementation of step S04 involves inputting the interface strain time history data and acoustic emission signal characteristic parameters into the interface gap identification model, and using deep learning methods to identify the location, width, and development rate of the gap at the bottom of the foundation. The interface gap identification model employs a multi-layer neural network structure, with the input layer containing 96 neurons, receiving interface strain time history data from 12 fiber optic strain sensing points and acoustic emission signal characteristic parameters from 16 acoustic emission sensors. The model training loss function is expressed as follows:

[0081] ;

[0082] In the formula, The total loss function value is a dimensionless parameter. is the mean square error weighting coefficient, a dimensionless parameter with a value of 0.6; is the cross-entropy weight coefficient, a dimensionless parameter with a value of 0.4; The mean squared error loss is a dimensionless parameter, specifically expressed as: ; Cross-entropy loss is a dimensionless parameter, specifically expressed as: .in, The sample size is a dimensionless parameter. For the first The predicted gap width for each sample, in units of ; For the first The actual gap width of each sample, in units of ; The maximum gap width is set to 2.00. ; For the first The true gap probability of a sample, a dimensionless parameter, takes the value of 0 or 1; For the first The predicted gap probability for each sample, a dimensionless parameter with a value ranging from 0 to 1.

[0083] The specific implementation of step S07 involves inputting the gap width, gap development rate, and grouting hole location parameters into a grouting parameter optimization game model, and solving for the optimal grouting pressure and grouting speed through a two-layer optimization. The objective function of the upper-layer model is expressed as follows:

[0084] ;

[0085] In the formula, Basic disturbance index, unit: ; The standard disturbance value is 0.05. ; To optimize grouting pressure, the unit is... The value range is 0.3. Up to 0.8 ; The reference grouting pressure is set at 0.5. ; To optimize grouting speed, the unit is... The value range is 5 Up to 15 ; The baseline grouting speed is set to 10. ; The gap width is expressed in units of 1000 mm. ; The maximum gap width is set to 2.00. ; Let be the grouting pressure-velocity coupling coefficient, a dimensionless parameter with a value of 0.3. This objective function considers the coupling effect between grouting pressure and velocity. The first term... The second term characterizes the foundation disturbance caused by grouting dynamics. The effect of the gap width on the disturbance is characterized, and the sum of the two factors is normalized to obtain the basic disturbance index.

[0086] The objective function of the lower-level model is expressed as follows:

[0087] ;

[0088] In the formula, The density of the gap filling is a dimensionless parameter, ranging from 0 to 1, and is required to be no less than 0.85. To optimize grouting pressure, the unit is... ; The reference grouting pressure is set at 0.5. ; To optimize grouting speed, the unit is... ; The baseline grouting speed is set to 10. ; The rate of crack development is expressed in units of... ; The maximum growth rate is set to 0.30. ; The grouting aging attenuation coefficient is a dimensionless parameter with a value of 0.25. This objective function considers the dynamic balance between the grouting filling effect and crack development. The first term... The second term characterizes the positive contributions of grouting pressure and velocity to filling density. This characterizes the negative impact of continuous crack development on filling effectiveness.

[0089] The model solution employs an iterative optimization strategy, at the... In this iteration, the updated equation for optimizing the grouting pressure is expressed as follows:

[0090] ;

[0091] In the formula, For the first The optimized grouting pressure for the next iteration is expressed in units of... ; For the first The optimized grouting pressure for the next iteration is expressed in units of... ; This is a pressure reference value, with a value of 1. ; The pressure update step size, in units of The empirical value is 0.01; The partial derivative of the basic disturbance index with respect to the optimized grouting pressure, in units of... Specifically, it is expressed as .

[0092] The updated equation for optimizing the grouting rate is expressed as follows:

[0093] ;

[0094] In the formula, For the first The optimized grouting speed for the next iteration is expressed in units of... ; For the first The optimized grouting speed for the next iteration is expressed in units of... ; This is a speed reference value, with a value of 1. ; To update the step size quickly, the unit is... The empirical value is 0.5; The partial derivative of the gap filling density with respect to the optimized grouting speed, in units of... Specifically, it is expressed as .

[0095] The iterative convergence condition is stated as follows:

[0096] and ;

[0097] In the formula, The grouting pressure convergence threshold is a dimensionless parameter with a value of 0.01. The grouting speed convergence threshold is a dimensionless parameter with a value of 0.01. It is the absolute value symbol.

[0098] It should be noted that the variables involved in this invention are explained in detail in Table 1.

[0099] Table 1. Variable Explanation Table

[0100]

[0101] To better understand and implement this invention, a specific application scenario, Example 2, is provided below: A technical team is responsible for vibration control and reinforcement of the foundation of the wind tunnel drive system in a large-scale wind tunnel test facility construction project. The equipment foundation of the wind tunnel drive system adopts... Concrete pouring, with silty clay soil beneath the foundation, the characteristic value of which is 280. The technical team adopted the method for reinforcing equipment foundations under high-frequency and high-intensity vibration as described in this invention to systematically address the micro-displacement problem of foundations under high-frequency vibration.

[0102] The technical team first pre-embedded a distributed fiber optic strain sensor network at the bottom of the foundation. The fiber optic sensors, protected by a polyimide coating, were arranged in a grid pattern with a longitudinal and transverse spacing of 200mm, totaling 576 fiber optic sensing points. The strain measurement range was -10000. Up to 10000 The temperature resistance range is -40℃ to 180℃. Twenty acoustic emission sensors are evenly distributed around the four sides of the foundation, with an operating frequency range of 20... Up to 400 Sensitivity is -62 The sensor is bolted to the foundation surface. The technical team established a dynamic constitutive model considering the soil strain rate effect, using a modified Cambridge model as the basic framework. Strain rate parameters were obtained through field dynamic triaxial tests of the soil, with test loading frequencies covering 10Hz to 200Hz. Cyclic softening parameters were determined through cyclic loading and unloading tests, with 1500 cycles. Model parameters include the soil elastic modulus. =42 Poisson's ratio =0.32, internal friction angle =28°, cohesion =35 Strain rate sensitivity coefficient =0.18, cyclic softening coefficient =0.24.

[0103] The technical team established a three-dimensional finite element model of the foundation-subsoil using explicit dynamic time history analysis. The foundation was simulated using solid elements with a mesh size controlled at 48mm. The subsoil area was three times the size of the foundation, and viscoelastic artificial boundaries were used to simulate infinite domain effects. The team used the field-measured vibration acceleration time history as the load input, with a time history length of 120s. An explicit integration algorithm was used to calculate the dynamic time history with a time step of 0.0001s, outputting the stress-strain time history and displacement response at the bottom of the foundation. During the trial operation of the wind tunnel drive system, the team deployed 15 acceleration sensors and 12 displacement sensors on the foundation surface and in the subsoil to measure the actual vibration response and displacement data. A genetic algorithm was used to invert and optimize the soil constitutive model parameters. After eight iterations, the error converged to 7.8%, completing the model parameter verification.

[0104] After the wind tunnel drive system was officially put into operation, the distributed fiber optic strain sensor network collected real-time strain data of the interface at the bottom of the foundation at a sampling frequency of 10,000 Hz, while the acoustic emission sensor simultaneously recorded the debonding acoustic signal. After 72 hours of operation, the technical team discovered an abnormal strain signal in the southeast region of the foundation bottom, and continuous monitoring showed that the strain amplitude in this area continued to increase. The technical team input the interface strain time history data and acoustic emission signal characteristic parameters into the interface gap recognition model. This model is a deep neural network structure with an input layer containing 96 neurons, a first hidden layer containing 256 neurons, a second hidden layer containing 128 neurons, a third hidden layer containing 64 neurons and incorporating an attention mechanism, and an output layer containing 32 neurons. The model output shows three cracks in the southeastern region at the bottom of the foundation, with coordinates of (18.6m, 4.2m), (19.2m, 4.8m), and (19.8m, 5.1m), crack widths of 0.12mm, 0.18mm, and 0.15mm, and crack development rates of 0.22mm / h, 0.28mm / h, and 0.25mm / h, respectively. Figure 1 As shown, the development rate of the cracks accelerates over time, with the development rate of the cracks located at (19.2m, 4.8m) approaching the critical value of 0.30mm / h.

[0105] Based on five consecutive tests showing a gap width exceeding 0.05mm, the technical team determined it to be a persistent abnormal state and initiated an emergency grouting procedure. Using the gap location coordinates, the team laid out the grouting hole positions in the corresponding area at the bottom of the foundation using a total station, controlling the layout error within 8mm. Three grouting holes were drilled, each with a diameter of 30mm, penetrating 3.5m of foundation thickness and entering 50mm into the foundation soil, at a drilling speed of 1.2m / min. The team input the gap width, gap development rate, and grouting hole location parameters into a grouting parameter optimization game model. This model includes an upper-level model aiming to minimize foundation disturbance and a lower-level model aiming to maximize gap filling density. The objective function of the upper-level model is the ratio of the foundation disturbance index to the standard disturbance value of 0.05mm, while the objective function of the lower-level model is the gap filling density. Through iterative calculations, the model output optimized grouting parameters for the three gaps, as shown in Table 2.

[0106] Table 2 Optimization results of gap grouting parameters

[0107]

[0108] The technical team formulated a high-fluidity, micro-expansion slurry, using 425 ordinary Portland cement and fly ash (fineness 42) in the specified mass ratio. The grout was prepared by mixing bentonite (88% sodium content), polycarboxylate superplasticizer, and water, with a water-cement ratio controlled at 0.52. The initial setting time of the grout was 6 hours, the final setting time was 18 hours, and the 28-day compressive strength was 18.5. The grout expansion rate was 0.3%, and the fluidity was 295 mm. During the grouting process, the technical team first adopted low-pressure, slow-speed grouting, controlling the grouting pressure at 0.3... Grouting speed controlled at 5 The grouting process lasted for 15 minutes to allow the grout to fully penetrate the crack. When the outlet pressure of the grouting pipe reached 70% of the set pressure, grouting was performed according to the optimized grouting pressure and speed in Table 1. For the crack located at (19.2m, 4.8m), the grouting pressure was 0.68... The grouting speed is 11.2. When the outlet pressure of the grouting pipe reaches 0.65 Stop grouting after the grouting has remained stable for 3 minutes. Figure 2 As shown, during the grouting process, the displacement monitoring of the foundation surface indicated that the displacement change was controlled within 0.03 mm, and no abrupt changes occurred. Figure 3 As shown, the grouting pressure curves of the three gaps show a trend of first low, then high and then stable, reflecting the transition process of grout from permeation filling to compaction reinforcement.

[0109] Seventy-two hours after grouting, the technical team deployed 10 high-precision displacement sensors on the top surface of the foundation, achieving a measurement accuracy of 0.01 mm. The measuring points covered the foundation's edges and center. Continuous monitoring of the foundation's horizontal displacement and vertical settlement under wind tunnel drive system operation conditions was conducted for 168 hours. Monitoring results showed a maximum horizontal displacement of 0.08 mm and an average of 0.05 mm, and a maximum vertical settlement of 3.2 mm and an average of 2.6 mm, both meeting the control requirements of less than 0.1 mm for horizontal displacement and less than 4 mm for vertical settlement. The technical team also used a distributed fiber optic strain sensor network to monitor the strain distribution at the bottom of the foundation. Verification results showed a 62% decrease in strain amplitude in the original gap area, with the strain distribution becoming more uniform, indicating that the gaps were effectively filled and compacted.

[0110] The advancements of this invention over traditional grouting reinforcement methods primarily lie in the accurate identification and adaptive control of foundation gaps under vibration conditions. Traditional methods rely mainly on empirical judgment and fixed-parameter grouting, making it difficult to accurately capture the dynamic evolution of gaps at the foundation bottom under high-frequency vibration. This results in a lack of scientific basis for grouting parameter selection, easily leading to insufficient or excessive grouting. This invention achieves full-domain perception of the foundation bottom interface state through the collaborative monitoring of a distributed fiber optic strain sensor network and an acoustic emission sensor array. Combined with a deep neural network recognition model, it can accurately output the location, width, and development rate of gaps, providing precise target parameters for subsequent grouting. This invention employs a two-layer game optimization model to determine grouting parameters. The upper-layer model aims to minimize foundation disturbance, avoiding additional damage to the foundation during grouting. The lower-layer model aims to maximize gap filling density, ensuring reinforcement effectiveness. The two-layer model achieves a dynamic balance between grouting pressure and grouting speed through iterative solutions, ensuring reinforcement effectiveness while minimizing foundation disturbance. This adaptive optimization mechanism is unattainable with traditional fixed-parameter grouting methods. The dynamic constitutive model established in this invention, which considers the effect of soil strain rate, can more realistically reflect the mechanical behavior of foundation soil under high-frequency vibration, providing theoretical support for the analysis of crack formation mechanism and the prediction of grouting effect, and overcoming the limitations of traditional static analysis methods in dynamic problems.

[0111] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for reinforcing equipment foundations under high-frequency, high-intensity vibration with micro-displacement, characterized in that, Includes the following steps: A distributed fiber optic strain sensor network and an acoustic emission sensor array are pre-embedded at the bottom of the foundation of the wind tunnel drive system; a dynamic constitutive model considering the soil strain rate effect is established, and the foundation system is simulated by finite element method using explicit dynamic time history analysis; during the operation of the wind tunnel drive system, the distributed fiber optic strain sensor network collects the interface strain data at the bottom of the foundation in real time, and the acoustic emission sensor records the debonding acoustic signal simultaneously, thereby obtaining the interface strain time history data and acoustic emission signal characteristic parameters; The interface strain time history data and acoustic emission signal characteristic parameters are input into the interface gap identification model. The interface gap identification model outputs the gap location coordinates, gap width, and gap development rate at the bottom of the foundation. Grouting holes are drilled in the corresponding area at the bottom of the foundation according to the gap location coordinates. The gap width, gap development rate, and grouting hole location parameters are input into the grouting parameter optimization game model. The grouting parameter optimization game model outputs the optimized grouting pressure and optimized grouting speed. High-fluidity micro-expansion grout is injected into the gap area through the grouting holes, and grouting is performed according to the optimized grouting pressure and optimized grouting speed. After grouting is completed, foundation displacement is monitored to verify the reinforcement effect.

2. The method according to claim 1, characterized in that, The deployment method of the distributed fiber optic strain sensor network is as follows: before the concrete is poured at the bottom of the foundation, the fiber optic sensors are arranged in a grid pattern, and the fiber optic sensors realize distributed strain monitoring through optical time domain reflection technology.

3. The method according to claim 2, characterized in that, The acoustic emission sensor array is arranged such that acoustic emission sensors are evenly distributed around the four sides of the foundation, and the location of the sound source is determined by a time difference positioning algorithm between the sensors.

4. The method according to claim 3, characterized in that, The dynamic constitutive model that considers the soil strain rate effect specifically adopts the modified Cambridge model as the basic framework and introduces strain rate parameters and cyclic softening parameters to improve the model.

5. The method according to claim 4, characterized in that, The implementation steps of the explicit dynamic time history analysis method are as follows: establish a three-dimensional finite element model of the foundation, use the measured vibration acceleration time history as the load input, and use an explicit integration algorithm to calculate the dynamic time history.

6. The method according to claim 5, characterized in that, After establishing a dynamic constitutive model that considers the effect of soil strain rate, the next step is to verify the model parameters by inverting them and using a genetic algorithm to optimize the soil constitutive model parameters.

7. The method according to claim 6, characterized in that, The interface gap refers to the debonding and separation area between the bottom of the foundation and the contact surface of the foundation soil caused by vibration.

8. The method according to claim 7, characterized in that, The structure of the interface gap identification model is as follows: the input layer receives the interface strain time history data of the fiber optic strain sensing point and the acoustic emission signal characteristic parameters of the acoustic emission sensor; the three hidden layers use rectified linear units as activation functions; the third hidden layer introduces an attention mechanism; and the output layer outputs the gap probability, gap width, and gap development rate of the potential gap region at the bottom of the foundation.

9. The method according to claim 8, characterized in that, The steps for establishing the training dataset for the interface gap recognition model are as follows: a scaled-down foundation vibration test bench is built in the laboratory, gaps of different widths are artificially prefabricated at the bottom of the foundation, vibration loads of different frequencies and amplitudes are applied to the test bench, fiber optic strain data and acoustic emission signals are collected simultaneously, and the real location and width of the gaps are measured using a high-precision three-dimensional laser scanner as label data.

10. The method according to claim 9, characterized in that, The interface gap recognition model training steps specifically employ mini-batch gradient descent for training, and the loss function uses a weighted combination of mean squared error and cross-entropy, while a dropout method is introduced to prevent overfitting.

Images