A method for detecting the ultimate bearing capacity of a pile foundation

Abstract

The present application relates to the pile foundation engineering detection technical field, disclose a kind of pile foundation ultimate bearing capacity detection, this method is by arranging multiple sensor array at pile top, vibration, displacement, soil pressure multi-source signal is synchronously collected, 2Dstack data fusion strategy is used to carry out pretreatment, in combination with the optimized and corrected Heili piling formula and 2Dstack-CNN deep learning model, realize the accurate calculation of pile foundation ultimate bearing capacity.The present application has high detection precision, small bearing capacity detection error, defect positioning error ≤±0.15m;High detection efficiency, single pile detection time 2-3 hours, more than 80% than static load test;Nondestructive and strong adaptability, suitable for in-service pile foundation, complex geology and large-diameter pile foundation;High degree of intelligence, without manual waveform interpretation;Detection cost is reduced by more than 50% compared with traditional method, with very high engineering application value.

Description

Technical Field

[0001] This invention relates to the field of pile foundation engineering testing technology, and more specifically, to a method for testing the ultimate bearing capacity of pile foundations. Background Technology

[0002] As the core load-bearing foundation of an engineering structure, the ultimate bearing capacity of pile foundations directly determines the safety and stability of the superstructure. Therefore, pile foundation ultimate bearing capacity testing is a crucial step in pile foundation engineering quality control. Currently, commonly used methods for testing the ultimate bearing capacity of pile foundations in engineering mainly include static load testing, high-strain method, low-strain method, and core drilling method. However, all of these methods have significant limitations and cannot meet the requirements for accurate, efficient, and non-destructive testing in complex engineering scenarios.

[0003] The existing testing methods mentioned above mostly rely on single-point data acquisition, which is susceptible to environmental interference leading to data distortion. Furthermore, data analysis depends on manual interpretation, resulting in low efficiency, large errors, and difficulty in achieving precise quantitative detection of ultimate bearing capacity. Simultaneously, existing methods have poor adaptability to piles with variable cross-sections, ultra-long piles, and pile foundations under complex geological conditions, significantly reducing detection accuracy. Therefore, developing a precise, efficient, non-destructive, and highly adaptable method for detecting the ultimate bearing capacity of pile foundations has become an urgent technical problem to be solved in the field of pile foundation testing. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides a method for detecting the ultimate bearing capacity of pile foundations. Through multi-source data fusion and intelligent analysis, it achieves accurate, efficient, and non-destructive testing of the ultimate bearing capacity of pile foundations, improves testing efficiency and accuracy, expands the scope of application of testing, and solves testing challenges in complex scenarios.

[0005] To achieve the above objectives, the technical solution of the present invention is as follows:

[0006] A method for testing the ultimate bearing capacity of pile foundations includes:

[0007] At least three acceleration sensors are arranged at equal intervals on the exposed section at the top of the pile, a displacement sensor is arranged at the center of the pile top, and soil pressure sensors are arranged in the soil around the pile foundation. All sensors are connected to the data acquisition terminal to complete sensor calibration and data acquisition terminal parameter settings.

[0008] A graded excitation method was adopted, in which excitation was applied directly above the pile top measuring point in the same direction as the receiving direction of the accelerometer. The excitation intensity increased in stages from low to high, and remained stable for 3-5 minutes after each excitation. The pile vibration signal, pile top settlement and displacement signal, and soil stress signal of the surrounding strata were collected simultaneously by the accelerometer, displacement sensor, and soil pressure sensor. At least 4096 time series data points were collected for each excitation, and the data acquisition terminal stored the collected multi-source signal data in real time.

[0009] The vibration signals collected by the accelerometers are preprocessed to remove environmental interference noise, signal drift, and abnormal data. A 2D stack data fusion strategy is adopted to stack the vibration signals collected by multiple accelerometers into a two-dimensional tensor. The displacement signal and earth pressure signal are smoothed and normalized. Wavelet transform is used to perform time-frequency analysis on the vibration signal to extract key feature parameters from the signal.

[0010] The preprocessed multi-source data is input into a preset deep learning model, and combined with the optimized and corrected Haili pile driving formula, the predicted value of the ultimate bearing capacity of the pile foundation is calculated. The predicted value is corrected by combining the variation law of earth pressure signal and displacement signal to obtain the final detection value of the ultimate bearing capacity of the pile foundation.

[0011] The ultimate bearing capacity test value of the pile foundation is compared with the ultimate bearing capacity design value of the pile foundation to determine whether the pile foundation meets the design requirements; combined with the characteristics of the pile vibration signal, it is determined whether there are defects in the pile body. If defects are found, the location and severity of the defects are output simultaneously; finally, a visual test report is generated.

[0012] As a preferred embodiment of the present invention, before arranging the acceleration sensor and the displacement sensor, the top of the pile foundation to be tested is cleaned to remove the laitance and debris on the pile top and ensure that the pile top is flat.

[0013] As a preferred embodiment of the present invention, the arrangement of the acceleration sensors is adjusted according to the pile diameter: when the pile diameter is ≤1.2m, three sensors are arranged in an equilateral triangle distribution; when the pile diameter is >1.2m, four acceleration sensors are arranged in a square distribution.

[0014] The earth pressure sensor is arranged around the pile foundation within a range of 1-2 times the pile diameter, at a depth consistent with the bearing layer depth of the pile foundation.

[0015] As a preferred embodiment of the present invention, the graded vibration adopts an electro-hydraulic servo vibration method with an excitation frequency range of 10-100Hz. The increment of the excitation load for each stage is 10%-15% of the design ultimate bearing capacity, until the settlement at the top of the pile reaches the ultimate settlement standard specified in the specification or obvious pile foundation failure characteristics appear, at which point the vibration is stopped. During the vibration process, sensor data is monitored in real time. If an abnormal signal occurs, the vibration is stopped immediately, the fault is investigated, and the data is collected again.

[0016] As a preferred embodiment of the present invention, wavelet threshold denoising method is used to remove environmental interference noise in vibration signal, linear interpolation method is used to correct signal drift, and 3σ criterion is used to remove abnormal data; normalization processing adopts min-max normalization method to map data to the [0,1] interval; the key feature parameters include wave velocity, amplitude, frequency, and attenuation coefficient.

[0017] As a preferred embodiment of the present invention, the deep learning model is a 2D stack-CNN model, which is generated by training large-scale pile foundation detection data. The training process includes: collecting pile foundation detection data with different geological conditions, different pile types, and different defect types to construct a training database; after preprocessing the data in the database, dividing it into training set, validation set, and test set, using gradient descent method to train the model, adjusting the model parameters until the prediction accuracy of the model is ≥99.8% and the validation set R² is ≥0.998.

[0018] As a preferred embodiment of the present invention, the test results are verified by sampling verification using the static load test method, with a sampling ratio of not less than 5% of the total number of tests. If the error between the test value of this method and the test value of the static load test is ≤3%, the test result is deemed valid; if the error is >3%, the sensor arrangement, data acquisition and analysis process are re-examined, corrected, and the test is repeated.

[0019] As a preferred embodiment of the present invention, the defects in the pile body include segregation, necking, fracture, and mud inclusion, and the defect positioning error is ≤ ±0.15m.

[0020] As a preferred embodiment of the present invention, the optimized and modified Haili piling formula is as follows: ;

[0021] in The weight of the hammer core is directly input from the technical parameters of the electro-hydraulic servo excitation device. For hammer-assisted high jump, the technical parameters of the excitation device are directly input; The energy transfer reduction factor for pile hammers is determined by matching from an empirical database or by on-site calibration based on pile type, pile length, and geological conditions. The hammering efficiency is determined by the performance parameters of the excitation device and the on-site working conditions; The final hammer penetration depth is the cumulative permanent settlement at the pile top monitored by displacement sensors. Total number of hammer blows Calculated, i.e. ; The total elastic deformation of the pile-soil system during hammering is directly determined by the rebound value at the pile top measured by the on-site displacement sensor.

[0022] A pile foundation ultimate bearing capacity testing system includes:

[0023] The sensor array module includes no less than three accelerometers equally spaced on the exposed section of the pile top, a displacement sensor arranged at the center of the pile top, and an earth pressure sensor arranged in the soil strata around the pile foundation. All sensors are connected to the data acquisition terminal for sensor calibration and data acquisition terminal parameter settings.

[0024] The graded excitation module is used to apply excitation directly above the pile top measuring point in the same direction as the receiving direction of the acceleration sensor. The excitation intensity increases in stages from low to high, and the pile-soil system remains stable for 3-5 minutes after each excitation.

[0025] The synchronous data acquisition terminal is electrically connected to the sensor array module and the graded excitation module, respectively, and is used to control the acceleration sensor, displacement sensor and earth pressure sensor to synchronously acquire pile vibration signal, pile top settlement displacement signal and stress signal of the surrounding strata of the pile foundation. Each level of excitation acquires no less than 4096 time series data points and stores the acquired multi-source signal data in real time.

[0026] The data preprocessing unit, electrically connected to the synchronous data acquisition terminal, is used to preprocess the vibration signals acquired by the accelerometer to remove environmental interference noise, signal drift, and abnormal data; it employs a 2Dstack data fusion strategy to stack the vibration signals acquired by multiple accelerometers into a two-dimensional tensor; it performs smoothing and normalization processing on the displacement signal and earth pressure signal; and it performs time-frequency analysis on the vibration signal through wavelet transform to extract key feature parameters from the signal.

[0027] The analysis and calculation unit is electrically connected to the data preprocessing unit. It has a built-in preset deep learning model and an optimized and corrected Hailey pile driving formula. It is used to input the preprocessed multi-source data into the deep learning model, calculate the predicted value of the ultimate bearing capacity of the pile foundation by combining the optimized and corrected Hailey pile driving formula, and then correct the predicted value by combining the variation law of earth pressure signal and displacement signal to obtain the final detection value of the ultimate bearing capacity of the pile foundation.

[0028] The result output unit is electrically connected to the analysis and calculation unit. It is used to compare the detected value of the ultimate bearing capacity of the pile foundation with the design value of the ultimate bearing capacity of the pile foundation to determine whether the pile foundation meets the design requirements; it combines the characteristics of the pile vibration signal to determine whether there are defects in the pile body. If there are defects, it outputs the location and severity of the defects simultaneously; and finally generates a visual inspection report.

[0029] The beneficial technical effects of this invention are:

[0030] A multi-sensor array is used to synchronously collect multi-source signals of vibration, displacement, and soil pressure. The complete data information is preserved through a 2Dstack data fusion strategy. Combined with the optimized Haili pile driving formula and deep learning model, the ultimate bearing capacity is quantitatively and accurately detected with an error of ≤3% and a defect location error of ≤±0.15m, which far exceeds the accuracy level of traditional detection methods. This solves the problems of insufficient detection accuracy and reliance on manual interpretation of traditional methods.

[0031] By adopting a graded excitation and automated data acquisition and analysis mode, there is no need for complex reaction devices. The testing time for a single pile foundation is shortened to 2-3 hours. Compared with the static load test that requires 1-3 days for a single pile, the testing efficiency is improved by more than 80%. It can realize rapid sampling inspection of large-scale engineering piles and significantly reduce the testing cycle and labor costs.

[0032] The testing process generates only slight vibrations through excitation, without damaging the pile foundation structure. It is applicable to various pile types, including in-service pile foundations, precast piles, cast-in-place piles, and rock-socketed piles. It is also suitable for testing complex geological conditions, variable cross-section piles, and ultra-long piles. It solves the shortcomings of traditional methods, such as poor adaptability and easy damage to pile foundations, and expands the scope of application of the testing.

[0033] By introducing deep learning models, we can achieve automatic analysis, feature extraction, and ultimate bearing capacity prediction of multi-source data, eliminating the need for manual waveform interpretation and reducing human error.

[0034] No large testing equipment or complex on-site setup is required. The sensors are reusable and no additional consumables are needed during the testing process. Compared with traditional methods such as static load testing and core drilling, the testing cost is reduced by more than 50%. It balances testing accuracy and economy and has extremely high engineering application value. Attached Figure Description

[0035] Figure 1 This is a schematic diagram of the overall process of the present invention.

[0036] Figure 2 This is a schematic diagram of the ultimate bearing capacity test for pile foundations.

[0037] Figure 3 This is a schematic diagram of the deep learning model of the present invention. Detailed Implementation

[0038] In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, the specific embodiments of the present invention will be further described in detail below with reference to the accompanying drawings and examples. The following examples are used to illustrate the present invention, but are not intended to limit the scope of the present invention.

[0039] Combination Figures 1-3 The present invention provides the following embodiments:

[0040] Example 1:

[0041] A method for testing the ultimate bearing capacity of pile foundations includes:

[0042] At least three acceleration sensors are arranged at equal intervals on the exposed section at the top of the pile, a displacement sensor is arranged at the center of the pile top, and soil pressure sensors are arranged in the soil around the pile foundation. All sensors are connected to the data acquisition terminal to complete sensor calibration and data acquisition terminal parameter settings.

[0043] A graded excitation method was adopted, in which excitation was applied directly above the pile top measuring point in the same direction as the receiving direction of the accelerometer. The excitation intensity increased in stages from low to high, and remained stable for 3-5 minutes after each excitation. The pile vibration signal, pile top settlement and displacement signal, and soil stress signal of the surrounding strata were collected simultaneously by the accelerometer, displacement sensor, and soil pressure sensor. At least 4096 time series data points were collected for each excitation, and the data acquisition terminal stored the collected multi-source signal data in real time.

[0044] The vibration signals collected by the accelerometers are preprocessed to remove environmental interference noise, signal drift, and abnormal data. A 2D stack data fusion strategy is adopted to stack the vibration signals collected by multiple accelerometers into a two-dimensional tensor. The displacement signal and earth pressure signal are smoothed and normalized. Wavelet transform is used to perform time-frequency analysis on the vibration signal to extract key feature parameters from the signal.

[0045] The preprocessed multi-source data is input into a preset deep learning model, and combined with the optimized and corrected Haili pile driving formula, the predicted value of the ultimate bearing capacity of the pile foundation is calculated. The predicted value is corrected by combining the variation law of earth pressure signal and displacement signal to obtain the final detection value of the ultimate bearing capacity of the pile foundation.

[0046] The ultimate bearing capacity test value of the pile foundation is compared with the ultimate bearing capacity design value of the pile foundation to determine whether the pile foundation meets the design requirements; combined with the characteristics of the pile vibration signal, it is determined whether there are defects in the pile body. If defects are found, the location and severity of the defects are output simultaneously; finally, a visual test report is generated.

[0047] Furthermore, before arranging the acceleration sensor and the displacement sensor, the top of the pile foundation to be tested is cleaned to remove laitance and debris, ensuring that the top of the pile is flat.

[0048] Latex and debris on the pile top can cause severe attenuation of excitation energy and poor coupling between the sensor and the pile top, introducing high-frequency noise. After cleaning, the pile top should be flat and hard so that the excitation energy can be effectively transmitted along the pile body, ensuring that the vibration signal truly reflects the dynamic response of the pile-soil system.

[0049] Furthermore, the arrangement of the acceleration sensors is adjusted according to the pile diameter: when the pile diameter is ≤1.2m, three sensors are arranged in an equilateral triangle distribution; when the pile diameter is >1.2m, four acceleration sensors are arranged in a square distribution.

[0050] The earth pressure sensor is arranged around the pile foundation within a range of 1-2 times the pile diameter, at a depth consistent with the bearing layer depth of the pile foundation.

[0051] Adjusting the number and geometric distribution of sensors according to the pile diameter can completely cover the dynamic response area at the pile top and accurately capture the vibration mode of the pile body; the earth pressure sensor is buried at the bearing layer depth to directly measure the comprehensive stress response of the pile end resistance and deep side friction, providing real pile-soil interaction boundary conditions for bearing capacity correction.

[0052] Furthermore, the graded vibration adopts an electro-hydraulic servo vibration method with an excitation frequency range of 10-100Hz. The increase in excitation load for each stage is 10%-15% of the design ultimate bearing capacity, until the settlement at the pile top reaches the ultimate settlement standard specified in the code or obvious pile foundation failure characteristics appear, at which point the vibration stops. Sensor data is monitored in real time during the vibration process. If an abnormal signal is detected, the vibration is stopped immediately, the fault is investigated, and the data is collected again.

[0053] Electro-hydraulic servo excitation enables precise control of load and frequency, and incrementally simulates the loading process of static load tests to avoid instantaneous plastic failure of the soil around the pile caused by impact loads. Each stage is stabilized for 3-5 minutes to allow the pore water pressure around the pile to dissipate fully, and the quasi-static deformation characteristics of the pile-soil system are obtained. Real-time monitoring can protect the pile foundation and sensors in time when equipment or signals are abnormal.

[0054] Furthermore, wavelet threshold denoising is used to remove environmental interference noise from the vibration signal, linear interpolation is used to correct signal drift, and the 3σ criterion is used to remove abnormal data; the normalization process uses the min-max normalization method to map the data to the [0,1] interval; the key feature parameters include wave velocity, amplitude, frequency, and attenuation coefficient.

[0055] Mechanical vibration and electromagnetic interference exist at the engineering site. Wavelet threshold denoising can separate effective signals from broadband noise in the time and frequency domains. Linear interpolation and the 3σ criterion eliminate sensor temperature drift and occasional strong interference. Min-max normalization eliminates dimensional differences in multi-source heterogeneous data. Wave velocity and attenuation coefficient are directly related to the quality of pile concrete and the damping of the surrounding soil. Amplitude and frequency reflect the stiffness of the pile-soil system, providing interpretable physical characteristics for bearing capacity assessment.

[0056] Furthermore, the deep learning model is a 2D stack-CNN model, which is generated by training large-scale pile foundation detection data. The training process includes: collecting pile foundation detection data with different geological conditions, different pile types, and different defect types to build a training database; after preprocessing the data in the database, it is divided into training set, validation set and test set; the gradient descent method is used to train the model and adjust the model parameters until the prediction accuracy of the model is ≥99.8% and the validation set R² is ≥0.998.

[0057] Traditional manual interpretation relies on personal experience, which is highly subjective and inefficient. 2Dstack-CNN automatically extracts the spatiotemporal coupling features of multi-channel vibration signals through convolutional kernels, and establishes a highly nonlinear mapping between pile vibration response and ultimate bearing capacity. Large-scale training data across geological and pile types endow the model with strong generalization ability, and high accuracy and coefficient of determination ensure the reliability of prediction results in engineering applications.

[0058] Furthermore, the test results are verified by sampling using the static load test method, with a sampling ratio of no less than 5% of the total number of tests. If the error between the test value and the static load test value is ≤3%, the test result is deemed valid; if the error is >3%, the sensor arrangement, data acquisition and analysis process are re-examined, corrected, and the test is repeated.

[0059] Static load testing is an industry-recognized benchmark for load-bearing capacity testing. Using it as a sampling reference, the intelligent prediction model of this invention can be calibrated in a closed loop. The 5% sampling ratio balances the representativeness of the verification with the economy of the test. The 3% error threshold matches the high precision of the model. Once the limit is exceeded, it will indicate the existence of systematic error. By tracing back the sensor layout and data quality, the test process can achieve self-correction.

[0060] Furthermore, pile defects include segregation, necking, fracture, and mud inclusion, with defect location error ≤ ±0.15m.

[0061] Furthermore, the optimized and corrected Haili piling formula is as follows: ;

[0062] in The weight of the hammer core is directly input from the technical parameters of the electro-hydraulic servo excitation device. For hammer-assisted high jump, the technical parameters of the excitation device are directly input; The energy transfer reduction factor for pile hammers is determined by matching from an empirical database or by on-site calibration based on pile type, pile length, and geological conditions. The hammering efficiency is determined by the performance parameters of the excitation device and the on-site working conditions; The final hammer penetration depth is the cumulative permanent settlement at the pile top monitored by displacement sensors. Total number of hammer blows Calculated, i.e. ; The total elastic deformation of the pile-soil system during hammering is directly determined by the rebound value at the pile top measured by the on-site displacement sensor.

[0063] The Haley formula establishes a balance between the hammer impact kinetic energy and the deformation work of the pile-soil system based on the law of conservation of energy. It replaces the traditional empirical value of elastic deformation C with the rebound value of the pile top measured by displacement sensors, and the final hammer penetration e is also calculated from the measured settlement, so that the theoretical formula is transformed from empirical estimation to being driven by measured data. The physical meaning of each parameter is clear and can be traced on site, which significantly improves the calculation accuracy and adaptability of the formula under complex geological conditions.

[0064] Example 2:

[0065] A pile foundation ultimate bearing capacity testing system includes:

[0066] The sensor array module includes no less than three accelerometers equally spaced on the exposed section of the pile top, a displacement sensor arranged at the center of the pile top, and an earth pressure sensor arranged in the soil strata around the pile foundation. All sensors are connected to the data acquisition terminal for sensor calibration and data acquisition terminal parameter settings.

[0067] The graded excitation module is used to apply excitation directly above the pile top measuring point in the same direction as the receiving direction of the acceleration sensor. The excitation intensity increases in stages from low to high, and the pile-soil system remains stable for 3-5 minutes after each excitation.

[0068] The synchronous data acquisition terminal is electrically connected to the sensor array module and the graded excitation module, respectively, and is used to control the acceleration sensor, displacement sensor and earth pressure sensor to synchronously acquire pile vibration signal, pile top settlement displacement signal and stress signal of the surrounding strata of the pile foundation. Each level of excitation acquires no less than 4096 time series data points and stores the acquired multi-source signal data in real time.

[0069] The data preprocessing unit, electrically connected to the synchronous data acquisition terminal, is used to preprocess the vibration signals acquired by the accelerometer to remove environmental interference noise, signal drift, and abnormal data; it employs a 2Dstack data fusion strategy to stack the vibration signals acquired by multiple accelerometers into a two-dimensional tensor; it performs smoothing and normalization processing on the displacement signal and earth pressure signal; and it performs time-frequency analysis on the vibration signal through wavelet transform to extract key feature parameters from the signal.

[0070] The analysis and calculation unit is electrically connected to the data preprocessing unit. It has a built-in preset deep learning model and an optimized and corrected Hailey pile driving formula. It is used to input the preprocessed multi-source data into the deep learning model, calculate the predicted value of the ultimate bearing capacity of the pile foundation by combining the optimized and corrected Hailey pile driving formula, and then correct the predicted value by combining the variation law of earth pressure signal and displacement signal to obtain the final detection value of the ultimate bearing capacity of the pile foundation.

[0071] The result output unit is electrically connected to the analysis and calculation unit. It is used to compare the detected value of the ultimate bearing capacity of the pile foundation with the design value of the ultimate bearing capacity of the pile foundation to determine whether the pile foundation meets the design requirements; it combines the characteristics of the pile vibration signal to determine whether there are defects in the pile body. If there are defects, it outputs the location and severity of the defects simultaneously; and finally generates a visual inspection report.

[0072] Example 3: Ultimate Bearing Capacity Testing of Cast-in-Place Piles in a Residential Building Project

[0073] In this embodiment, the pile foundation to be tested is a cast-in-place concrete pile with a diameter of 1.0m and a length of 25m. The bearing layer is a silty clay layer, and the design ultimate bearing capacity is 1800kN. The method of this invention is used for testing, and the specific steps are as follows:

[0074] 1. Pre-test preparation: Clean the pile top of laitance and debris, and grind the pile top until smooth; for pile diameters 1.0m ≤ 1.2m, arrange three accelerometers in an equilateral triangle on the exposed section of the pile top, with a sensor spacing of 0.4m, and place one high-precision laser displacement sensor at the center of the pile top, with a measurement accuracy of 0.01mm. Within a 1.5 times the pile diameter range around the pile foundation, arrange two earth pressure sensors at a depth consistent with the bearing layer depth of 25m; connect all sensors to the data acquisition terminal, complete sensor calibration, and set the data acquisition frequency to 100Hz to ensure data acquisition accuracy.

[0075] 2. Multi-source signal acquisition: An electro-hydraulic servo excitation device is used to apply excitation directly above the pile top measuring point. The excitation frequency range is 10-80Hz. The increment of the excitation load for each level is 10% of the design ultimate bearing capacity, i.e., 180kN, which gradually increases from 180kN to 1980kN. After each level of excitation, the system is kept stable for 4 minutes. Vibration signals are collected synchronously through three accelerometers, pile top settlement displacement signals are collected by a displacement sensor, and ground stress signals are collected by an earth pressure sensor. 4096 time series data points are collected for each level of excitation. All data are stored in real time at the data acquisition terminal.

[0076] 3. Data Preprocessing: Wavelet threshold denoising was used to remove environmental noise from the vibration signals, linear interpolation was used to correct signal drift, and the 3σ criterion was used to remove one outlier data point. A 2D stack data fusion strategy was used to stack the vibration signals from the three accelerometers into a two-dimensional tensor with 3 rows and 4096 columns. The displacement and earth pressure signals were smoothed, and the min-max normalization method was used to map all data to the [0,1] interval. Wavelet transform was used to extract the wave velocity, amplitude, frequency and other characteristic parameters of the vibration signals.

[0077] 4. Intelligent Analysis and Ultimate Bearing Capacity Calculation: Preprocessed multi-source data is input into a trained 2D stack-CNN model. The 2D stack-CNN model achieves a prediction accuracy of 99.85% and a validation set R² = 0.9989. The model automatically extracts signal features and, combined with the optimized Haili pile driving formula, inputs parameters such as hammer core weight, hammer impact height, final hammer penetration, and pile top rebound value to calculate the predicted ultimate bearing capacity of the pile foundation as 1825 kN. The predicted value is then corrected based on the variation patterns of earth pressure and displacement signals, ultimately yielding a detected ultimate bearing capacity value of 1818 kN.

[0078] 5. Verification and Output of Test Results: The test value of 1818kN was compared with the design ultimate bearing capacity of 1800kN, and the error was 1%, which meets the specification requirements. Through vibration signal characteristic analysis, it was determined that there were no obvious defects in the pile body. Three pile foundations were selected for verification using the static load test method. The static load test values ​​were 1822kN, 1815kN, and 1820kN, respectively, and the average error with the test value of this method was 0.3%, indicating that the test results were valid. A visual test report was generated, clearly stating that the ultimate bearing capacity of this batch of pile foundations meets the design requirements.

[0079] Example 4: Ultimate bearing capacity test of in-service pile foundations for a bridge project

[0080] In this embodiment, the pile foundation to be tested is an in-service precast reinforced concrete pile with a diameter of 1.5m and a length of 35m. The bearing stratum is moderately weathered rock, and the design ultimate bearing capacity is 3200kN. Since the superstructure has already been put into use, traditional static load tests cannot be carried out. Therefore, the method of this invention is used for testing. The specific steps are as follows:

[0081] 1. Pre-test preparation: Clear debris from the pile top to ensure it is flat; based on the pile diameter of 1.5m > 1.2m, arrange four accelerometers in a square pattern on the exposed section of the pile top, with a sensor spacing of 0.5m; arrange one laser displacement sensor at the center of the pile top; and arrange three soil pressure sensors within a range of twice the pile diameter around the pile foundation, at a depth consistent with the bearing layer depth of 35m; connect the sensors to the data acquisition terminal, complete the calibration, and set the acquisition frequency to 120Hz to avoid interference from the vibration of the superstructure on the test data.

[0082] 2. Multi-source signal acquisition: An electro-hydraulic servo excitation device is used to apply graded excitation with an excitation frequency range of 20-100Hz. The excitation load increment of each stage is 12% of the design ultimate bearing capacity, i.e., 384kN, gradually increasing from 384kN to 3584kN. Each stage of excitation is stabilized for 3 minutes, and multi-source signals of vibration, displacement, and earth pressure are acquired simultaneously. 5000 time series data points are collected for each stage of excitation to ensure that the data can truly reflect the interaction between the pile foundation and the stratum.

[0083] 3. Data preprocessing: Wavelet threshold denoising method is used to remove interference noise generated by the vibration of the superstructure, correct signal drift, and eliminate abnormal data; 2Dstack data fusion strategy is used to process vibration signals, and displacement and earth pressure signals are smoothed and normalized; characteristic parameters of vibration signals are extracted to provide support for subsequent analysis.

[0084] 4. Intelligent analysis and ultimate bearing capacity calculation: The preprocessed multi-source data is input into the 2D stack-CNN model and combined with the optimized Haili pile driving formula to calculate the predicted value of the ultimate bearing capacity as 3230kN; combined with the change of earth pressure signal, the final detected value is 3225kN after correction.

[0085] 5. Verification and output of test results: The test value of 3225kN was compared with the design ultimate bearing capacity of 3200kN, and the error was 0.78%, which meets the specification requirements; through signal analysis, it was determined that there were no defects in the pile body and the vibration of the superstructure did not affect the test results; a test report was generated, which clearly stated that the ultimate bearing capacity of the in-service pile foundation meets the usage requirements and no reinforcement treatment is required.

[0086] Application example:

[0087] The method for testing the ultimate bearing capacity of pile foundations includes the following steps:

[0088] Step 1: Pre-test preparation

[0089] 1.1 Pile Top Treatment

[0090] Clean the top of the pile foundation to be tested, remove the laitance, loose concrete and debris from the top of the pile, and grind it until it is flat and hard to ensure that the excitation energy is effectively transmitted.

[0091] 1.2 Sensor Array Arrangement

[0092] According to pile diameter Determine the sensor placement scheme. (Regarding the pile diameter...) At that time, the pile top exposed section is arranged at equal intervals. Several accelerometers are arranged in an equilateral triangle; when the pile diameter... At that time, arrangements There are 10 accelerometers, arranged in a square. The spacing between all the accelerometers is... All meet A high-precision laser displacement sensor is placed at the absolute geometric center of the pile top, with a measurement accuracy of no more than [value missing]. Around the pile foundation Arrangement within the range of multiple pile diameters A soil pressure sensor, typically taken as an example The burial depth of the sensor With the depth of the bearing layer of the pile foundation Consistency, that is This is to ensure that the collected ground stress signals can accurately reflect the interaction between the pile foundation and the bearing layer.

[0093] 1.3 Excitation point location

[0094] The excitation device uses an electro-hydraulic servo exciter, and the vibration point must be located directly above the centroid of the shape formed by the accelerometer sensors. For example... Figure 2 As shown, for an equilateral triangle arrangement, the centroid coordinates are... Planar coordinates of three sensors , , Calculate using the following formula:

[0095]

[0096] For a square arrangement, the centroid coordinates are the midpoint of the diagonal:

[0097]

[0098] The excitation direction should be consistent with the receiving direction of the accelerometer to avoid eccentric excitation.

[0099] 1.4 System Connection and Calibration

[0100] Connect all sensors to the data acquisition terminal to complete initial reading calibration and sensitivity calibration. Set the data acquisition frequency. Generally take If the pile foundation to be tested is an in-service pile foundation or there is vibration interference from the superstructure, then take... Set the number of sampling points for each excitation stage. Corresponding to single-point sampling time interval Confirm that the terminal has real-time data display, anomaly alarm, and storage functions.

[0101] Step 2: Multi-source signal acquisition

[0102] 2.1 Staged excitation control

[0103] Staged excitation is implemented using an electro-hydraulic servo excitation device. Excitation frequency range. Controlled The load increases in stages from low to high, with the increment of each single load stage being [not specified]. Take the design ultimate bearing capacity of ,Right now A constant load is maintained after each excitation stage. After the signal stabilizes, proceed to the next stage. The termination condition is: when the settlement at the top of the pile reaches the limit settlement standard specified in the code, or when obvious pile foundation failure characteristics appear (such as a sudden increase in settlement, cracking of the pile concrete, sudden change in earth pressure, etc.), the excitation should be stopped immediately.

[0104] 2.2 Multi-source synchronous acquisition

[0105] In the Under staged excitation, three types of raw timing signals are simultaneously acquired through a data acquisition terminal to ensure that the timestamp deviation of each channel is less than [value missing]. Specifically, this includes: by Vibration signals of the pile body collected by an accelerometer ,in , The settlement displacement signal at the top of the pile was acquired by a laser displacement sensor. , ;Depend on Earth pressure sensor collects stress signals from the soil strata around the pile. ,in , .

[0106] 2.3 Data Acquisition Quality Control

[0107] During the excitation process, sensor data is monitored in real time. If signal saturation, communication interruption, or abnormal fluctuation occurs, the excitation is stopped immediately, and the data is collected again after troubleshooting.

[0108] Step 3: Data Preprocessing and 2DStack Integration

[0109] 3.1 Wavelet Thresholding Denoising

[0110] For acceleration signals Perform a Discrete Wavelet Transform (DWT) to decompose it into approximate coefficients. With detail coefficient :

[0111]

[0112] Use a soft thresholding function to handle detail coefficients:

[0113]

[0114] Where the threshold Based on the number of sampling points Adaptive determination:

[0115]

[0116] In the formula To estimate the noise standard deviation, take the median absolute deviation of the detail coefficients at the finest scale and divide by... To obtain.

[0117] 3.2 Linear Interpolation Drift Correction

[0118] Estimated baseline drift And deduct it:

[0119]

[0120] The corrected signal is denoted as :

[0121]

[0122] 3.33σ Criterion for Anomaly Removal

[0123] Calculate the sample mean of the cleaned signal. with standard deviation If the signal satisfies at a certain moment If the value is not found, the point is considered an outlier and is replaced with a linear interpolation of the adjacent valid point.

[0124] 3.42Dstack Data Fusion

[0125] Will The cleaned acceleration channels are stacked according to sensor number to form a two-dimensional vibration tensor. Its dimensions are :

[0126]

[0127] The first dimension of this tensor corresponds to the spatial channel (sensor number), and the second dimension corresponds to the temporal sampling point, preserving temporal continuity and directly serving as the input feature map for a two-dimensional convolutional neural network. Simultaneous processing of displacement signals... Earth pressure signal Smoothing is performed using a moving average filter, and the window width is typically set to... .

[0128] 3.5 Normalization Processing

[0129] Using the min-max normalization method, all data are mapped to Interval:

[0130]

[0131] Normalized vibration tensors were obtained respectively. Normalized displacement sequence and normalized earth pressure sequence .

[0132] 3.6 Time-Frequency Feature Extraction

[0133] Key physical characteristic parameters were extracted using continuous wavelet transform (CWT). The stress wave velocity in the pile body was also analyzed. Utilizing the time difference of reflected waves from the pile bottom With pile length calculate:

[0134]

[0135] Simultaneously extract the main frequency (Peak frequency of power spectrum) and attenuation coefficient (Logarithmic decay method):

[0136]

[0137] In the formula and The first Next and first Second peak amplitude.

[0138] Step 4: Intelligent Analysis and Ultimate Bearing Capacity Calculation

[0139] This step employs a three-level fusion strategy of dual-branch prediction and multi-source response correction.

[0140] 4.1 Branch A: 2D stack-CNN deep learning model

[0141] Constructing the input layer By normalized vibration tensor With auxiliary feature vectors It is constructed by splicing together statistical features of displacement and earth pressure signals.

[0142]

[0143] in This represents the mean. Indicates standard deviation, This represents the slope of the displacement curve.

[0144] Will Input the trained 2D stack-CNN model, and denote the model parameters as follows: Preliminary predicted value of ultimate bearing capacity after forward propagation:

[0145]

[0146] This model was trained using large-scale pile foundation testing data. The training process employed gradient descent to adjust parameters until the model's prediction accuracy was no less than [a certain percentage]. Coefficient of determination of the validation set The model can quickly adapt to different types of pile foundation testing needs based on actual testing scenarios through transfer learning.

[0147] 4.2 Branch Road B: Optimized Haili Pile Driving Formula

[0148] By incorporating on-site sensor measured parameters, the traditional Haili piling formula is optimized and corrected as follows:

[0149]

[0150] The definitions of each parameter in the formula and the data source are as follows: This represents the theoretical value of the ultimate bearing capacity of the pile. The weight of the hammer core is directly input from the technical parameters of the electro-hydraulic servo excitation device. For hammer-assisted high jump, the technical parameters of the excitation device are directly input; The energy transfer reduction factor for pile hammers is determined by matching from an empirical database or by on-site calibration based on pile type, pile length, and geological conditions. The hammering efficiency is determined by the performance parameters of the excitation device and the on-site working conditions; The final hammer penetration depth is the cumulative permanent settlement at the pile top monitored by displacement sensors. Total number of hammer blows Calculated, i.e. ; The total elastic deformation of the pile-soil system during hammering is directly determined by the rebound value at the pile top measured by the on-site displacement sensor.

[0151] 4.3 Primary Fusion

[0152] A weighted fusion strategy is adopted, using CNN predictions. Primarily based on the theoretical value of the Heyley formula Assuming physical constraints, fused prediction values ​​are obtained. :

[0153]

[0154] Fusion weights The validation set performance is used as the metric, typically taking... Prioritize trusting data-driven models, but retain physical boundary constraints.

[0155] 4.4 Multi-source response correction

[0156] Introducing displacement response factor Earth pressure response factor Establish a comprehensive correction coefficient .

[0157] Displacement response factor Defined as the slope of the load-settlement curve under the current graded excitation. slope of theoretical limit state The ratio:

[0158]

[0159] Earth pressure response factor Defined as the measured peak earth pressure Ultimate lateral resistance of soil layers at corresponding depths The ratio:

[0160]

[0161] Comprehensive correction coefficient Calculate using the following formula:

[0162]

[0163] In the formula and In response to the weighting coefficients, satisfying The value is determined by regression from the training samples; if there is no prior data, an empirical value can be used. , .

[0164] 4.5 Ultimate bearing capacity

[0165] Merge predicted values Multiply by the overall correction factor The final test value of the ultimate bearing capacity of the pile foundation is obtained. :

[0166]

[0167] Its physical meaning is: when the pile-soil response reaches the limit state ( and )hour, The correction is zero; when the displacement and earth pressure response are not fully utilized. The predicted value is reduced; when an abnormal hardening response occurs, The predicted value should be appropriately increased.

[0168] Step 5: Verification and Output of Test Results

[0169] 5.1 Design Compliance Judgment

[0170] Final detection value The design value of the ultimate bearing capacity of the pile foundation as required by the standard Compare them. If If the pile foundation meets the design requirements; The design requirements have been determined not to be met, and further evaluation or reinforcement is recommended.

[0171] 5.2 Synchronous Identification of Pile Defects

[0172] By combining the reflected wave characteristics of pile vibration signals, the integrity of the pile body is simultaneously identified, along with the location of defects. Calculated using the stress wave reflection method:

[0173]

[0174] In the formula The stress wave velocity of the pile body extracted in step 3.6. The time difference between the defect-reflected wave and the incident first arrival wave is extracted from the acceleration signal time history analysis. The defect location error should satisfy no greater than [value missing]. .

[0175] Defect severity Assessment based on the degree of attenuation of the reflected wave amplitude:

[0176]

[0177] in The amplitude of the reflected wave from the defect. The amplitude of the incident wave. The larger the defect, the more severe it is. Identifiable defect types include segregation, necking, fracture, and inclusion, which are determined based on the phase and spectral distortion characteristics of the reflected wave.

[0178] 5.3 Sampling Verification and Error Control

[0179] The static load test method is used to sample and verify the test results, and the sampling ratio is not less than the total number of tests. Calculate the test values ​​obtained by this method and the test values ​​obtained by static load test. relative error :

[0180]

[0181] like If the test result is deemed valid; If necessary, the sensor layout, data acquisition and analysis process should be re-examined, corrected, and then re-tested.

[0182] 5.4 Visual Inspection Report Generation

[0183] The final test report is generated, which includes the following data stream final state summary: test parameters (pile type, pile diameter) Pile length Bearing layer depth Number of sensors and Sampling frequency The rate of increase of excitation load Multi-source signal curves (vibration time history curve, settlement-time curve, earth pressure-time curve); ultimate bearing capacity calculation results (CNN predicted value) Reference values ​​for the Heyley formula Fusion correction value and design value (Comparison); Static load verification error (If applicable); Results of pile defect analysis (defect type, depth, and location) Severity ); and the final test results and engineering recommendations.

[0184] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for testing the ultimate bearing capacity of pile foundations, characterized in that, include: At least three acceleration sensors are arranged at equal intervals on the exposed section at the top of the pile, a displacement sensor is arranged at the center of the pile top, and soil pressure sensors are arranged in the soil around the pile foundation. All sensors are connected to the data acquisition terminal to complete sensor calibration and data acquisition terminal parameter settings. A graded excitation method was adopted, in which excitation was applied directly above the pile top measuring point in the same direction as the receiving direction of the accelerometer. The excitation intensity increased in stages from low to high, and remained stable for 3-5 minutes after each excitation. The pile vibration signal, pile top settlement and displacement signal, and soil stress signal of the surrounding strata were collected simultaneously by the accelerometer, displacement sensor, and soil pressure sensor. At least 4096 time series data points were collected for each excitation, and the data acquisition terminal stored the collected multi-source signal data in real time. The vibration signals collected by the accelerometers are preprocessed to remove environmental interference noise, signal drift, and abnormal data. A 2D stack data fusion strategy is adopted to stack the vibration signals collected by multiple accelerometers into a two-dimensional tensor. The displacement signal and earth pressure signal are smoothed and normalized. Wavelet transform is used to perform time-frequency analysis on the vibration signal to extract key feature parameters from the signal. The preprocessed multi-source data is input into a preset deep learning model, and combined with the optimized and corrected Haili pile driving formula, the predicted value of the ultimate bearing capacity of the pile foundation is calculated. The predicted value is corrected by combining the variation law of earth pressure signal and displacement signal to obtain the final detection value of the ultimate bearing capacity of the pile foundation. The ultimate bearing capacity test value of the pile foundation is compared with the ultimate bearing capacity design value of the pile foundation to determine whether the pile foundation meets the design requirements.

2. The method for detecting the ultimate bearing capacity of pile foundations according to claim 1, characterized in that, Before installing the acceleration sensor and the displacement sensor, clean the top of the pile foundation to be tested, remove the laitance and debris from the top of the pile, and ensure that the top of the pile is flat.

3. The method for detecting the ultimate bearing capacity of pile foundations according to claim 1, characterized in that, The arrangement of the acceleration sensors is adjusted according to the pile diameter: when the pile diameter is ≤1.2m, three sensors are arranged in an equilateral triangle; when the pile diameter is >1.2m, four acceleration sensors are arranged in a square. The earth pressure sensor is arranged around the pile foundation within a range of 1-2 times the pile diameter, at a depth consistent with the bearing layer depth of the pile foundation.

4. The method for detecting the ultimate bearing capacity of pile foundations according to claim 1, characterized in that, The graded vibration adopts an electro-hydraulic servo vibration method with an excitation frequency range of 10-100Hz. The increase in excitation load for each stage is 10%-15% of the design ultimate bearing capacity, until the settlement at the pile top reaches the ultimate settlement standard specified in the code or obvious pile foundation failure characteristics appear, at which point the vibration stops. Sensor data is monitored in real time during the vibration process. If an abnormal signal is detected, the vibration is stopped immediately, the fault is investigated, and the data is collected again.

5. The method for detecting the ultimate bearing capacity of pile foundations according to claim 1, characterized in that, Wavelet thresholding is used to remove environmental noise from vibration signals, linear interpolation is used to correct signal drift, and the 3σ criterion is used to remove outlier data.

6. The method for detecting the ultimate bearing capacity of pile foundations according to claim 1, characterized in that, The normalization process uses the min-max normalization method to map the data to the [0,1] interval; the key feature parameters include wave velocity, amplitude, frequency, and attenuation coefficient.

7. The method for testing the ultimate bearing capacity of pile foundations according to claim 1, characterized in that, The deep learning model is a 2D stack-CNN model, which is generated by training large-scale pile foundation detection data. The training process includes: collecting pile foundation detection data with different geological conditions, different pile types, and different defect types to build a training database; after preprocessing the data in the database, it is divided into training set, validation set and test set; the gradient descent method is used to train the model and adjust the model parameters until the prediction accuracy of the model is ≥99.8% and the validation set R² is ≥0.

998.

8. The method for detecting the ultimate bearing capacity of pile foundations according to claim 1, characterized in that, The test results are verified by sampling using the static load test method, with a sampling ratio of no less than 5% of the total number of tests. If the error between the test value and the static load test value is ≤3%, the test result is deemed valid. If the error is >3%, the sensor arrangement, data acquisition and analysis process are re-examined, corrected, and the test is repeated.

9. The method for detecting the ultimate bearing capacity of pile foundations according to claim 1, characterized in that, The optimized and corrected Haili piling formula is as follows: ; in The weight of the hammer core is directly input from the technical parameters of the electro-hydraulic servo excitation device. For hammer-assisted high jump, the technical parameters of the excitation device are directly input; The energy transfer reduction factor for pile hammers is determined by matching from an empirical database or by on-site calibration based on pile type, pile length, and geological conditions. The hammering efficiency is determined by the performance parameters of the excitation device and the on-site working conditions; The final hammer penetration depth is the cumulative permanent settlement at the pile top monitored by displacement sensors. Total number of hammer blows Calculated, i.e. ; The total elastic deformation of the pile-soil system during hammering is directly determined by the rebound value at the pile top measured by the on-site displacement sensor.

10. A pile foundation ultimate bearing capacity testing system, characterized in that, include: The sensor array module includes no less than three accelerometers equally spaced on the exposed section of the pile top, a displacement sensor arranged at the center of the pile top, and an earth pressure sensor arranged in the soil strata around the pile foundation. All sensors are connected to the data acquisition terminal for sensor calibration and data acquisition terminal parameter settings. The graded excitation module is used to apply excitation directly above the pile top measuring point in the same direction as the receiving direction of the acceleration sensor. The excitation intensity increases in stages from low to high, and the pile-soil system remains stable for 3-5 minutes after each excitation. The synchronous data acquisition terminal is electrically connected to the sensor array module and the graded excitation module, respectively, and is used to control the acceleration sensor, displacement sensor and earth pressure sensor to synchronously acquire pile vibration signal, pile top settlement displacement signal and stress signal of the surrounding strata of the pile foundation. Each level of excitation acquires no less than 4096 time series data points and stores the acquired multi-source signal data in real time. The data preprocessing unit, electrically connected to the synchronous data acquisition terminal, is used to preprocess the vibration signals acquired by the accelerometer to remove environmental interference noise, signal drift, and abnormal data; it employs a 2Dstack data fusion strategy to stack the vibration signals acquired by multiple accelerometers into a two-dimensional tensor; it performs smoothing and normalization processing on the displacement signal and earth pressure signal; and it performs time-frequency analysis on the vibration signal through wavelet transform to extract key feature parameters from the signal. The analysis and calculation unit is electrically connected to the data preprocessing unit. It has a built-in preset deep learning model and an optimized and corrected Hailey pile driving formula. It is used to input the preprocessed multi-source data into the deep learning model, calculate the predicted value of the ultimate bearing capacity of the pile foundation by combining the optimized and corrected Hailey pile driving formula, and then correct the predicted value by combining the variation law of earth pressure signal and displacement signal to obtain the final detection value of the ultimate bearing capacity of the pile foundation. The result output unit is electrically connected to the analysis and calculation unit and is used to compare the detected value of the ultimate bearing capacity of the pile foundation with the design value of the ultimate bearing capacity of the pile foundation to determine whether the pile foundation meets the design requirements.

Images