A pile foundation state real-time monitoring and diagnosis system based on digital twinning technology

Abstract

The application relates to the technical field of pile foundation monitoring, and discloses a pile foundation state real-time monitoring and diagnosis system based on digital twin technology. The system comprises a multi-source data acquisition module, a data confidence evaluation module and an acoustic emission monitoring decision module. The multi-source data acquisition module comprises a plurality of sensor groups arranged at different depths of the pile foundation, each group comprising strain, acceleration, acoustic emission and temperature sensors, and the data of the pile foundation can be collected in multiple dimensions; the data confidence evaluation module receives original data, generates a corrected data sequence through time sequence noise separation and reconstruction, and calculates the data confidence according to the distribution discrete degree of the corrected data; and the acoustic emission monitoring decision module judges whether to start acoustic emission monitoring according to the data confidence, and controls the acoustic emission sensor array at the top of the pile foundation to collect acoustic emission signals when the acoustic emission monitoring is started. The system can comprehensively acquire the data of the pile foundation, improve the data accuracy, realize early damage identification, and guarantee the safety of the pile foundation.

Description

Technical Field

[0001] This invention relates to the field of pile foundation monitoring technology, specifically to a real-time monitoring and diagnosis system for pile foundation status based on digital twin technology. Background Technology

[0002] In fields such as building construction, transportation engineering, and water conservancy engineering, pile foundations serve as crucial load-bearing structures, and their operational status directly impacts the stability and safety of the entire engineering structure. As the scale of construction projects continues to expand, the geological environments in which pile foundations are located become increasingly complex. Special geological conditions, such as soft soil strata and karst development areas, can easily lead to problems like stress concentration, crack initiation, and structural damage during pile foundation use. Furthermore, during long-term service, pile foundations are continuously affected by external load changes, temperature fluctuations, and groundwater erosion. If potential structural hazards are not detected and addressed in a timely manner, they may lead to serious safety accidents such as decreased pile foundation bearing capacity, structural deformation, or even collapse, causing not only huge economic losses but also threatening human lives.

[0003] Monitoring methods for pile foundation conditions mainly include traditional manual inspection and conventional automated monitoring. Manual inspection typically employs techniques such as ultrasonic testing and low-strain reflection wave methods, requiring on-site operation by personnel. This is not only time-consuming and labor-intensive, but the results are also easily affected by factors such as human skill level and the testing environment. It suffers from limitations such as long inspection cycles and the inability to achieve real-time monitoring, making it difficult to meet the need for continuous dynamic monitoring of pile foundation conditions. While conventional automated monitoring systems can achieve continuous data acquisition, most rely on only a single type of sensor to obtain monitoring data, such as using only strain sensors to monitor pile foundation stress changes or only using accelerometers to capture vibration signals. This fails to comprehensively reflect the multi-dimensional state information of the pile foundation. Furthermore, existing automated monitoring systems often do not fully consider the impact of sensor noise and external interference on the accuracy of monitoring data during data processing, resulting in significant errors in the collected raw data and making it difficult to accurately determine the actual working state of the pile foundation. When it is necessary to monitor minute damages such as microcracks that may occur inside the pile foundation, the existing system lacks an effective monitoring triggering mechanism and accurate signal acquisition methods. It cannot capture key signals reflecting damage characteristics in a timely manner, making it difficult to achieve early identification and diagnosis of pile foundation damage. As a result, safety hazards of the pile foundation structure cannot be investigated and dealt with in a timely manner, which seriously affects the overall safety and reliability of the engineering structure. Summary of the Invention

[0004] The purpose of this invention is to provide a real-time monitoring and diagnosis system for pile foundation status based on digital twin technology, so as to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides a real-time monitoring and diagnosis system for pile foundation status based on digital twin technology, the system comprising:

[0006] The multi-source data acquisition module includes multiple sensor groups deployed at different depths in the pile foundation. Each sensor group contains a strain sensor, an acceleration sensor, an acoustic emission sensor, and a temperature sensor.

[0007] The data confidence assessment module is used to receive the raw monitoring data of the multiple sensor groups, perform time-series noise separation and reconstruction on the raw monitoring data of each sensor, generate a corrected monitoring data sequence, and calculate the confidence of the pile foundation data at the current moment based on the distribution dispersion of the corrected monitoring data sequence of all sensors.

[0008] The acoustic emission monitoring decision module is used to determine whether to start acoustic emission monitoring based on the confidence level of the pile foundation data output by the data confidence assessment module. When it is determined that acoustic emission monitoring should be started, the acoustic emission sensor array deployed on the top of the pile foundation is controlled to collect the acoustic emission signal of the pile foundation.

[0009] Preferably, each sensor group in the multi-source data acquisition module further includes positioning markers set on the sidewall of the pile foundation;

[0010] The acoustic emission signal collected by the acoustic emission sensor array includes the reflected waveform of the positioning marker point.

[0011] Preferably, the data confidence assessment module performs time-series noise separation and reconstruction on the raw monitoring data of each sensor to generate a corrected monitoring data sequence, including:

[0012] For each sensor, extract continuous sampled values ​​of its raw monitoring data within a preset time window;

[0013] Adaptive mode decomposition is performed on the continuous sampled values ​​to separate high-frequency noise components and low-frequency feature components;

[0014] The high-frequency noise components are filtered based on a preset noise threshold, and the filtered high-frequency noise components and low-frequency feature components are reconstructed into a corrected monitoring data sequence.

[0015] Preferably, the data confidence assessment module calculates the confidence level of the pile foundation data at the current moment based on the distribution dispersion of the corrected monitoring data sequence from all sensors, including:

[0016] Acquire corrected monitoring data from all sensors at the same timestamp;

[0017] Calculate the standard deviation of all corrected monitoring data at the same timestamp;

[0018] Calculate the moving average of the standard deviation based on a series of standard deviations from multiple consecutive timestamps;

[0019] The standard deviation moving average is input into a preset confidence mapping function to output the confidence level of the pile foundation data.

[0020] Preferably, the acoustic emission monitoring decision module determines whether to initiate acoustic emission monitoring based on the confidence level of the pile foundation data output by the data confidence assessment module, including:

[0021] When the confidence level of the pile foundation data is lower than the preset confidence threshold and the duration exceeds the preset duration threshold, an acoustic emission monitoring start command is generated;

[0022] The acoustic emission sensor array performs signal acquisition in response to the acoustic emission monitoring start command.

[0023] Preferably, the system further includes:

[0024] The digital twin feature library module is used to store the pile foundation geometric model, material parameter model, and historical health status feature vectors.

[0025] The real-time diagnostic engine module is used to fuse the corrected monitoring data sequence from the multi-source data acquisition module with the acoustic emission signals acquired by the acoustic emission sensor array, and combine the pile foundation geometric model and material parameter model in the digital twin feature library module to generate a pile stress distribution cloud map and damage location coordinates.

[0026] Preferably, the real-time diagnostic engine module generates a pile stress distribution cloud map and damage location coordinates, including:

[0027] The sound wave propagation path deviation is calculated based on the time delay difference of the reflected waveforms of the positioning markers in the acoustic emission signal.

[0028] The coordinates of the positioning markers in the pile foundation geometric model are corrected based on the deviation of the sound wave propagation path.

[0029] The corrected monitoring data sequence of the multi-source data acquisition module is mapped onto the surface of the corrected pile foundation geometric model;

[0030] The stress gradient distribution of the pile body is calculated using a material parameter model, and a stress distribution cloud map of the pile body is generated.

[0031] Identify the coordinates of abnormal stress concentration areas in the pile stress distribution cloud map and use them as damage location coordinates.

[0032] Preferably, the system further includes:

[0033] The status trend prediction module is used to extract historical health status feature vectors from the digital twin feature library module;

[0034] Use the current damage location coordinates and stress gradient values ​​as new feature vectors;

[0035] The fused historical health status feature vector sequence is compressed in terms of feature dimension and aligned in time sequence.

[0036] State evolution rate analysis is performed based on the compressed and aligned feature vector sequence, and the damage development rate and remaining lifetime prediction are output.

[0037] Preferably, the system further includes:

[0038] The health status grading module is used to receive the damage location coordinates from the real-time diagnostic engine module and the damage development rate from the status trend prediction module.

[0039] The overall deterioration index of the pile foundation is calculated based on the weighted product of the depth distribution density of the damage location coordinates and the damage development rate.

[0040] The overall deterioration index of the pile foundation is matched with the preset grading threshold range, and the pile foundation health status level code is output.

[0041] Preferably, the system further includes:

[0042] The maintenance strategy generation module is used to receive the pile foundation health status level code from the health status grading module;

[0043] A pre-defined maintenance measures library is established based on the pile foundation health status level code index;

[0044] The remaining life prediction value of the state trend prediction module is combined to optimize the maintenance measure execution time parameters;

[0045] Generate a list of pile foundation maintenance strategies that includes maintenance measure types and execution time windows.

[0046] Compared with the prior art, the beneficial effects of the present invention are:

[0047] This real-time monitoring and diagnosis system for pile foundation status based on digital twin technology, by setting up a multi-source data acquisition module and deploying multiple sensor groups including strain sensors, acceleration sensors, acoustic emission sensors, and temperature sensors at different depths of the pile foundation, can collect monitoring data of the pile foundation from multiple dimensions such as stress and strain, vibration response, acoustic emission characteristics, and temperature changes. It breaks through the limitations of traditional monitoring methods that rely on only a single type of sensor to obtain information, and comprehensively captures multi-dimensional state parameters of the pile foundation at different depths and under different working conditions. This allows staff to have a more comprehensive understanding of the overall working condition of the pile foundation, avoids missing key state information due to a single monitoring dimension, and provides a rich data foundation for subsequent accurate judgment of the pile foundation status.

[0048] The data confidence assessment module performs time-series noise separation and reconstruction on the raw monitoring data of each sensor, generating a corrected monitoring data sequence. This effectively filters out the influence of sensor noise and external environmental interference on the monitoring data, reducing errors in the raw data and improving the accuracy and reliability of the monitoring data. Simultaneously, this module calculates the confidence level of the pile foundation data at the current moment based on the distribution dispersion of the corrected monitoring data sequences from all sensors. This directly reflects the reliability of the current monitoring data, providing a scientific basis for deciding whether to initiate more precise monitoring methods. It avoids misjudgments or omissions caused by relying on inaccurate data, ensuring that the assessment of the pile foundation's condition is based on reliable data.

[0049] The acoustic emission monitoring decision module determines whether to activate acoustic emission monitoring based on the confidence level of the pile foundation data. This enables on-demand activation of acoustic emission monitoring, avoiding unnecessary energy consumption and data redundancy caused by the long-term continuous operation of the acoustic emission sensor array, and reducing system operating costs and data processing pressure. When acoustic emission monitoring is activated, the acoustic emission sensor array deployed on top of the pile foundation collects acoustic emission signals from the pile foundation. It can accurately capture acoustic emission characteristic signals generated by minute damages such as microcracks that may occur inside the pile foundation. Especially when the data confidence level is high and there may be structural anomalies, timely activation of targeted acoustic emission monitoring helps to detect potential damage to the pile foundation earlier, enabling early identification and diagnosis of pile foundation damage. This buys time for staff to take timely repair and reinforcement measures, preventing further damage and ensuring the long-term stability and safety of the pile foundation structure, thereby maintaining the safe operation of the entire engineering structure.

[0050] The entire system organically combines multi-source data acquisition, data confidence assessment, and acoustic emission monitoring and decision-making to form a complete real-time monitoring and diagnosis process for pile foundation status. With the support of digital twin technology, it can achieve real-time monitoring of pile foundation status, accurate data processing, and scientific decision-making. This changes the problems of low efficiency and poor accuracy of traditional monitoring methods and also makes up for the shortcomings of conventional automated monitoring systems in data processing and monitoring triggering mechanisms. It provides a more efficient, accurate, and comprehensive monitoring solution for the field of pile foundation monitoring, and is suitable for pile foundation monitoring needs under different geological conditions and engineering scenarios, with broad application prospects. Attached Figure Description

[0051] Figure 1 This is a timing diagram of the real-time monitoring and diagnosis system for pile foundation status based on digital twin technology described in this invention.

[0052] Figure 2 This is a flowchart of multi-source data acquisition and acoustic emission signal acquisition including positioning markers;

[0053] Figure 3A flowchart for calculating the confidence level of pile foundation data;

[0054] Figure 4 This is a flowchart illustrating the collaboration between a digital twin feature library and a real-time diagnostic engine. Detailed Implementation

[0055] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0056] Please see Figure 1 The present invention provides a real-time monitoring and diagnosis system for pile foundation status based on digital twin technology. The system includes: a multi-source data acquisition module, a data confidence assessment module, and an acoustic emission monitoring and decision module.

[0057] The multi-source data acquisition module acquires data through multiple sensor groups deployed at different depths in the pile foundation. Each sensor group includes a strain sensor, an acceleration sensor, an acoustic emission sensor, and a temperature sensor. The data confidence assessment module receives the raw monitoring data from multiple sensor groups, performs time-series noise separation and reconstruction on the raw monitoring data from each sensor, generates a corrected monitoring data sequence, and calculates the pile foundation data confidence level at the current moment based on the distribution dispersion of all sensor corrected monitoring data sequences. The acoustic emission monitoring decision module determines whether to activate acoustic emission monitoring based on the pile foundation data confidence level output by the data confidence assessment module. When it determines to activate acoustic emission monitoring, it controls the acoustic emission sensor array deployed at the top of the pile foundation to collect the acoustic emission signals from the pile foundation.

[0058] Example 1: See Figure 2 Positioning markers, made of highly reflective material, are placed at specific locations on the sidewalls of the pile foundation to effectively reflect acoustic signals. During installation, the spatial coordinates of each sensor array are precisely measured and recorded using a total station. When acquiring signals, the acoustic emission sensor array records clear reflected waveforms from these positioning markers. These reflected waveforms exhibit a distinguishable time delay compared to the directly propagating acoustic emission signal; this delay is directly related to the propagation path length from the sound source to the marker and then to the sensor.

[0059] For the raw monitoring data collected by each sensor, the data processing procedure first defines a preset time window. The length of this window is determined based on the dynamic characteristics of the monitored object and the sampling frequency. The system extracts all sampled values ​​within this window from the continuous data stream, forming a discrete data sequence to be processed. Subsequently, adaptive mode decomposition is performed on this data sequence. This decomposition process adaptively decomposes the complex raw signal into a series of eigenmode function components with different frequency characteristics. These components are arranged in descending order of frequency. High-frequency components typically correspond to random noise, electronic thermal noise, or transient environmental interference in the measurement system, while low-frequency components more often reflect physical phenomena such as strain, vibration, or temperature changes generated by the pile foundation structure under load.

[0060] After completing mode decomposition, a preset noise threshold is used to distinguish and filter out high-frequency noise components. This threshold is not fixed but dynamically adjusted based on the sensor type, the statistical characteristics of its historical data (such as standard deviation and peak distribution), and current environmental conditions. For example, during thunderstorms or when large machinery is operating nearby, electromagnetic interference and mechanical vibrations are aggravated, and the system will correspondingly lower the noise threshold to more strictly filter out suspicious high-frequency components. The filtering process typically uses soft or hard threshold functions to attenuate or zero out the identified high-frequency noise components.

[0061] The filtered high-frequency components (which may still contain some useful high-frequency information) are reconstructed along with all low-frequency feature components. The reconstruction process essentially involves linearly superimposing all processed intrinsic mode function components to generate a purified, more accurate, and corrected monitoring data sequence that better reflects the physical state of the pile foundation. This corrected sequence significantly reduces random noise interference, preserves key signal features, and provides a higher-quality data foundation for subsequent data consistency assessment and fusion diagnosis.

[0062] The establishment of positioning markers and the acquisition of their reflected waveforms provide an indispensable spatial reference for subsequent precise positioning using acoustic emission signals. When acoustic emission signals propagate through the pile material, their propagation path and arrival time are affected by material heterogeneity, internal defects, and minor errors in sensor installation. By identifying the reflected waveforms of the positioning markers within the acoustic emission signals, the deviation between the actual propagation path of the sound wave and the theoretical path in an ideal homogeneous model can be calculated. This path deviation information can be used to fine-tune and correct the actual coordinates of the sensor nodes, thereby constructing a geometric model in the digital twin model that more closely resembles the actual physical space. This significantly improves the accuracy of subsequent mapping of monitoring data onto the model and damage localization.

[0063] The entire process of temporal noise separation and reconstruction is a continuous loop. The system processes the latest monitoring data continuously using a sliding window, generating a cleaned and corrected data stream that updates over time. This process effectively improves the signal-to-noise ratio of the original data, laying a solid foundation for confidence assessment based on the consistency of multi-sensor data. High-confidence monitoring data is a key prerequisite for the digital twin model to accurately reflect the true state of the physical pile foundation.

[0064] Example 2: See Figure 3 In the pile foundation monitoring system, all sensor groups output corrected monitoring data sequences under strictly synchronized timestamps. The system sets a unified time reference and coordinates the sampling time of each sensor node through a high-precision clock signal. Taking a specific monitoring cycle as an example, at time point T, the system acquires corrected monitoring data values ​​from 32 channels of 8 strain sensors, 8 acceleration sensors, 8 acoustic emission sensors, and 8 temperature sensors distributed at different depths of the pile foundation. These values ​​represent the measurement results of physical quantities after noise processing, such as micro-strain values, acceleration amplitude, acoustic emission energy values, and Celsius temperature values.

[0065] The calculation process first performs a dispersion analysis on all 32 data values ​​at time point T. The system calculates the arithmetic mean of these values, then calculates the sum of squared deviations of each data value from the mean, finally obtaining the standard deviation of the dataset at that time point. This standard deviation directly reflects the consistency of the monitoring data from various sensors at the current moment. For example, when the pile foundation structure is in a stable stress state, the strain readings of sensors at different depths should show a regular distribution, and the standard deviation is usually maintained at a low level. However, if abnormal deformation occurs in a local area, the readings of the relevant sensors will deviate from the overall trend, resulting in a significant increase in the standard deviation.

[0066] To eliminate the impact of transient interference, the system employs a moving average algorithm to process the standard deviation of continuous time series. A sliding window containing 60 consecutive time points (corresponding to a 5-minute monitoring duration) is set, and the arithmetic mean of all standard deviations within the window is calculated as the output value. This moving average smooths out accidental fluctuations at individual time points, such as data jumps caused by transient electromagnetic interference from a particular sensor, thus more reliably characterizing the overall trend of data dispersion. The window length can be adjusted according to the response characteristics of the pile foundation structure; a shorter window can be used for bridge pile foundations subject to frequent dynamic loads, while a longer window can be used for building foundation piles.

[0067] The preset confidence level mapping function is responsible for converting the standard deviation moving average into a confidence level index within the range of 0 to 1. This function employs a piecewise linear transformation rule: when the moving average is below the set lower threshold, the highest confidence level of 1.0 is output; when the moving average exceeds the upper threshold, the lowest confidence level of 0.1 is output; within the upper and lower threshold range, the confidence level decreases linearly as the moving average increases. The threshold parameters are determined based on the statistical distribution of historical monitoring data. For example, for a certain offshore wind power pile foundation project, the lower threshold is set to 0.8 μɛ (micro-strain), and the upper threshold is set to 4.5 μɛ.

[0068] The complete mathematical expression for the confidence mapping function is:

[0069]

[0070] in =Standard deviation moving average (unit: μɛ, matched to strain sensor accuracy); = Output confidence level (value range: 0.1-1.0); =Lower threshold (data stability threshold); = Upper limit threshold (critical value for data distortion). Standard deviation moving average This refers to the result of calculating the sliding window standard deviation and then performing 5 moving averages on 30 sets of data continuously collected by the sensor (sampling frequency 1Hz); threshold and The critical value characterizing the stability of pile foundation monitoring data directly affects the sensitivity of confidence assessment. Scenario-based value selection rules: Threshold. and The value is based on historical data statistics for different pile foundation engineering scenarios, building pile foundations: , (Applicable to scenarios where static loads are the primary factor); Bridge pile foundations: , (Considering the impact of vehicle dynamic loads); Offshore platform pile foundations: , (Considering the alternating characteristics of wave loads). Parameter calibration method: The threshold can be adjusted through on-site calibration experiments. Data is continuously collected for 24 hours under undamaged pile foundation conditions, and the lower limit of the 95% confidence interval is taken as the threshold. Take 1.5 times the maximum normal fluctuation value as Taking building pile foundations as an example, when the standard deviation moving average collected by the sensor... Time: Confidence level This result corresponds to a medium confidence level, and the system will trigger a secondary data verification process. The calculation of the standard deviation moving average must meet the following requirements: ① Sampling equipment: strain sensor with an accuracy class of 0.1; ② Environmental conditions: temperature 0-40℃, humidity ≤85%RH; ③ Data preprocessing: outliers exceeding 3 times the standard deviation must be removed before calculation.

[0071] The acoustic emission monitoring triggering logic employs a dual-condition judgment mechanism. The system continuously monitors the confidence value of the output pile foundation data. When this value remains below a preset confidence threshold (e.g., 0.65) for an extended period exceeding a preset duration threshold (e.g., 180 seconds), an acoustic emission monitoring start command is generated. This design avoids false triggering due to instantaneous data anomalies, such as vibration interference caused by a briefly passing heavy vehicle. In a cross-sea bridge pile foundation monitoring example, when a typhoon caused non-uniform swaying of the pile body, the data dispersion of each acceleration sensor continuously increased, and the confidence value dropped from 0.72 to 0.58 within 3 minutes, triggering the acoustic emission monitoring system.

[0072] Upon receiving the start command, the acoustic emission sensor array enters a high-sensitivity acquisition mode. The 12 piezoelectric sensors in the array synchronously acquire acoustic emission signals at a sampling rate of 1 MHz for 500 milliseconds. Bandpass filtering (50 kHz-400 kHz) is used to suppress environmental noise during the acquisition process, and hardware gain control is activated to adapt to acoustic emission events of varying intensities. In a case study of monitoring subsea tunnel pile foundations, when an abnormally low data confidence level was detected, the acoustic emission array successfully captured the characteristic acoustic emission signals generated by the propagation of microcracks within the pile concrete, acquiring up to 6 million valid waveform data points in a single acquisition.

[0073] The system incorporates a status recovery mechanism. Once acoustic emission monitoring is initiated, if the data confidence level continuously rises above the safety threshold and remains there for more than 300 seconds, acoustic emission acquisition is automatically terminated, and the system returns to normal monitoring mode. This mechanism has been applied in the monitoring of pile foundations in large ports. After abnormal vibrations caused by crane loading and unloading operations cease, the system shuts down the high-power acquisition mode of the acoustic emission sensor after confirming that data consistency has been restored. The entire data confidence assessment and decision-making process forms a closed-loop control, optimizing system energy consumption while ensuring monitoring reliability.

[0074] Example 3: See Figure 4 The digital twin feature library module constructs a refined three-dimensional geometric model of the pile foundation. This model includes the precise dimensions and cross-sectional shape of the pile, as well as the theoretical spatial coordinates of all sensors and positioning markers. The material parameter model stores constitutive relation data such as the elastic modulus, Poisson's ratio, density, and strength parameters of the concrete. Historical health status feature vectors are stored in time series, with each vector containing historical damage coordinates, stress gradient values, and the corresponding monitoring timestamp.

[0075] When the real-time diagnostic engine starts, it first processes the raw waveform data acquired by the acoustic emission sensor array. The system identifies reflected signals from the positioning markers on the pile foundation sidewall and accurately extracts their arrival times. For any two sensors... and It received from the same marker point The time difference of the reflected wave is This time delay difference is related to the speed of sound propagation in concrete. Together they are used to calculate path deviation:

[0076]

[0077] in: Indicates sensor and For the marked point The deviation between the measured path and the theoretical path. The speed at which sound waves propagate in concrete. For sensors and Received marker point Time difference of the reflected wave , , They represent sensors respectively. , and marker points The theoretical space coordinate vector.

[0078] Based on the path deviation calculation results of multiple marker points, a least squares optimization algorithm is used to correct the actual coordinates of all sensors in the digital twin model, establishing a spatial mapping relationship consistent with the physical pile foundation. After coordinate correction, the system maps the corrected monitoring data sequence provided by the multi-source data acquisition module onto the updated geometric model surface. Strain sensor data is converted into surface stress values, acceleration data is integrated to obtain the displacement field, and temperature data is used to correct for the thermal expansion effect of the material. These physical quantities are assigned to the corresponding nodes of the model, forming the initial boundary conditions of the pile surface.

[0079] The stress field distribution inside the pile is calculated using the constitutive relations provided by the material parameter model and the finite element method. The calculation process considers the pile-soil interaction boundary conditions, applying the resistance of the surrounding soil as an additional constraint to the model. After obtaining the stress tensor at each point on the pile, the system extracts the principal stress values ​​and calculates their gradient modulus, generating a color-coded stress distribution cloud map of the pile. The cloud map uses a continuous color spectrum to represent stress magnitude, transitioning from blue (low stress area) to red (high stress concentration area). In the generated stress distribution cloud map, the system uses a region growing algorithm to identify abnormal stress concentration areas. The algorithm uses stress points exceeding 80% of the material's yield strength as seed points, expanding to adjacent regions until the stress value falls below the threshold boundary, ultimately determining the geometric center coordinates of the abnormal area as the damage location coordinates. All identified damage coordinates and their corresponding stress gradient values ​​are sent to a digital twin feature library in real time to update the historical health status feature vector sequence.

[0080] The entire diagnostic process forms a closed loop: acoustic emission signals provide spatial calibration information, multi-source monitoring data provide physical field input, digital twin models provide the computational framework, and the final output is a visualized stress cloud map and quantified damage coordinates. This multimodal data fusion method significantly improves the spatial accuracy and reliability of pile foundation condition diagnosis, providing detailed spatial information for structural health monitoring.

[0081] Example 4: The status trend prediction module periodically extracts historical health status feature vector sequences from the digital twin feature library. These feature vectors are arranged in chronological order, and each vector contains the three-dimensional coordinates of the damage point, the stress gradient value of that point, and the corresponding timestamp. The module adds the latest damage location coordinates and stress gradient values ​​output by the real-time diagnostic engine as new feature vectors to the historical sequence, forming an expanded feature dataset. The expanded feature vector sequence is preprocessed, including feature dimensionality compression and temporal alignment. Dimensionality compression uses principal component analysis to project high-dimensional feature data into a low-dimensional space, retaining the main change information. Temporal alignment is performed using a dynamic time warping algorithm to align feature vectors from different time points on the time axis, eliminating the influence of inconsistent acquisition time intervals. The processed feature vector sequence exhibits a clearer temporal evolution pattern.

[0082] Based on the compressed and aligned eigenvector sequence, the module performs state evolution rate analysis. The analysis focuses on the spatial distribution changes of damage coordinates and the temporal trends of stress gradient values. By calculating the differences in eigenvectors between adjacent time points, the rate of change of damage parameters is obtained. Using time series analysis methods, a mathematical model of damage development is established to predict the damage state at future time points. The output results include the damage development rate and the predicted remaining life, where the damage development rate includes two components: spatial expansion rate and intensity change rate.

[0083] The health status grading module receives damage location coordinates from the real-time diagnostic engine and damage development rate from the status trend prediction module. The module first calculates the depth distribution density of the damage location coordinates, analyzing the distribution of damage points along the pile depth direction. Using kernel density estimation, it obtains the probability distribution function of damage points along the depth axis. The depth distribution density and damage development rate are then weighted and fused to calculate the overall pile foundation deterioration index. The weighting coefficients are determined based on the pile foundation design requirements and engineering experience; typically, depth distribution density has a higher weight because deeper damage has a more significant impact on pile foundation safety.

[0084] Table 1: Threshold standards for grading the health status of pile foundations.

[0085]

[0086] Referring to Table 1, the calculated overall deterioration index of the pile foundation is matched with the preset grading threshold ranges to output the corresponding pile foundation health status level code. The grading threshold ranges are formulated based on extensive engineering practice and structural safety standards, and are divided into five level ranges, each corresponding to a different status description and treatment recommendations. The health status level codes adopt a standardized coding format to facilitate data exchange and integration with other management systems.

[0087] The entire process forms a complete health status assessment chain: extracting characteristic patterns from historical data, combining current monitoring data to predict development trends, and finally providing a quantitative health status rating. This multi-dimensional data analysis-based method can comprehensively reflect the structural status of the pile foundation, providing detailed status information for engineering decisions. The system regularly outputs health status reports, including the current level code, the changing trends of major damage parameters, and recommended inspection and maintenance cycles. All assessment results are stored in a digital twin feature library, forming a complete historical record for long-term trend analysis and structural performance evolution studies.

[0088] In practical engineering applications, this grading method can adapt to different types of pile foundation structures. For bridge pile foundations, the focus is on the development of fatigue damage caused by vehicle loads; for building pile foundations, the emphasis is on the impact of long-term settlement; and for offshore platform pile foundations, special attention needs to be paid to the damage evolution under corrosive environments. By adjusting the weighting coefficients and grading thresholds, the system can perform customized assessments for specific engineering scenarios, ensuring the applicability and accuracy of the assessment results. The multi-source data fusion method and time series analysis technology used in the assessment process can effectively handle the uncertainty and volatility in monitoring data, providing stable condition assessment outputs.

[0089] Example 5: The maintenance strategy generation module receives the pile foundation health status level code output by the health status grading module. This code adopts a standardized discrete identifier form, such as the HL1 to HL5 level codes. The module has a built-in structured maintenance measure database, which stores the corresponding maintenance operation sets according to the health status level. Each maintenance measure entry contains structured fields such as operation type, standard execution cycle, resource requirements, and expected effects. The database design adopts a tree index structure, supporting quick retrieval of related measure sets by health status level code.

[0090] When the health status level code HL4 is entered, the system retrieves the corresponding set of measures, which includes three main operations: crack surface sealing treatment, local non-destructive testing, and rebar corrosion potential measurement. Each operation comes with standard execution parameters: crack sealing treatment is recommended to be completed within 30 days after the level is determined; non-destructive testing requires coverage within a 2-meter radius around the damage location coordinates; corrosion measurement requires setting up 9 measuring points at different elevations on the pile body.

[0091] The module synchronously receives the remaining life prediction value provided by the status trend prediction module. This value, in monthly units, represents the estimated time when the pile foundation will reach a critical state at the current rate of damage development. For the retrieved maintenance measures, the system initiates a time parameter optimization program. The optimization process considers the remaining life prediction value, the time required to implement the measures, and seasonal influencing factors. Taking a high-rise building pile foundation as an example, when the remaining life prediction value is 24 months and the current health level is HL3, the comprehensive inspection originally scheduled for 12 months is optimized and postponed to 18 months later, while the frequency of quarterly settlement observations is increased.

[0092] The calculation of the maintenance measure execution time window adopts a dynamic programming algorithm. The algorithm input includes: the standard cycle of the measure, the predicted remaining life, the current seasonal constraints (such as construction is not advisable during the frozen soil period), and the logical dependencies between measures. The output is the earliest start time, the latest completion time, and the recommended execution period for each measure. For the inspection operation of offshore platform pile foundations during the typhoon season, the system automatically limits the underwater inspection time window to May to September each year and generates an early warning prompt 3 months in advance.

[0093] When generating the pile foundation maintenance strategy list, the system integrates the following elements: measure name, technical specification description, implementation entity requirements, resource consumption estimate, priority level identifier, and time window parameters. The list is output in both machine-readable JSON format and human-readable tabular format. An example maintenance list for a subway support pile includes: Installing fiber optic grating sensors on pile number P-27, requiring a team with special engineering qualifications, consuming 35 meters of 8-core optical cable, with an execution time window from October 15th to November 20th, 2023, marked as a priority A task.

[0094] When the health status level changes or a new remaining life prediction is received, the system automatically triggers the inventory update process. The update process retains historical version records and marks the modified content. In the monitoring of pile groups of large bridges, the system generates differentiated maintenance strategies based on the independent assessment results of each pile. The main pier pile foundation adopts a carbon fiber reinforcement scheme, while the auxiliary pier pile foundation only requires anti-corrosion coating maintenance. All strategies are classified and archived by pile number to form a project-level maintenance manual.

[0095] The inventory output interface supports multi-platform adaptation and can be pushed to the engineering management system via a web service interface or generated as an industry-standard PDF document. Maintenance personnel can obtain personalized maintenance strategies by scanning the QR code on the pile with a mobile terminal, and the on-site execution status is fed back to the system through the mobile application to form a closed-loop record. This dynamic strategy generation mechanism realizes digital integration from status assessment to maintenance execution, providing operational guidelines for the entire lifecycle management of infrastructure.

[0096] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0097] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A real-time monitoring and diagnosis system for pile foundation status based on digital twin technology, characterized in that, include: The multi-source data acquisition module includes multiple sensor groups deployed at different depths in the pile foundation. Each sensor group contains a strain sensor, an acceleration sensor, an acoustic emission sensor, and a temperature sensor. The data confidence assessment module is used to receive the raw monitoring data of the multiple sensor groups, perform time-series noise separation and reconstruction on the raw monitoring data of each sensor, generate a corrected monitoring data sequence, and calculate the confidence of the pile foundation data at the current moment based on the distribution dispersion of the corrected monitoring data sequence of all sensors. The acoustic emission monitoring decision module is used to determine whether to start acoustic emission monitoring based on the confidence level of the pile foundation data output by the data confidence assessment module. When it is determined that acoustic emission monitoring should be started, the acoustic emission sensor array deployed on the top of the pile foundation is controlled to collect the acoustic emission signal of the pile foundation. The data confidence assessment module performs time-series noise separation and reconstruction on the raw monitoring data of each sensor to generate a corrected monitoring data sequence, including: For each sensor, extract continuous sampled values ​​of its raw monitoring data within a preset time window; Adaptive mode decomposition is performed on the continuous sampled values ​​to separate high-frequency noise components and low-frequency feature components; The high-frequency noise components are filtered based on a preset noise threshold, and the filtered high-frequency noise components and low-frequency feature components are reconstructed into a corrected monitoring data sequence. The data confidence assessment module calculates the confidence level of the pile foundation data at the current moment based on the distribution dispersion of the corrected monitoring data sequence from all sensors, including: Acquire corrected monitoring data from all sensors at the same timestamp; Calculate the standard deviation of all corrected monitoring data at the same timestamp; Calculate the moving average of the standard deviation based on a series of standard deviations from multiple consecutive timestamps; The standard deviation moving average is input into a preset confidence mapping function to output the confidence level of the pile foundation data; The complete mathematical expression for the confidence mapping function is: ； in =Standard deviation moving average, unit: μɛ, matched with the accuracy of strain sensor; = Output confidence level; =Lower threshold; =Upper threshold; The acoustic emission monitoring decision module determines whether to initiate acoustic emission monitoring based on the confidence level of the pile foundation data output by the data confidence assessment module, including: When the confidence level of the pile foundation data is lower than the preset confidence threshold and the duration exceeds the preset duration threshold, an acoustic emission monitoring start command is generated; The acoustic emission sensor array performs signal acquisition in response to the acoustic emission monitoring start command.

2. The real-time monitoring and diagnosis system for pile foundation condition based on digital twin technology according to claim 1, characterized in that, Each sensor group in the multi-source data acquisition module also includes positioning markers set on the sidewall of the pile foundation; The acoustic emission signal collected by the acoustic emission sensor array includes the reflected waveform of the positioning marker point.

3. The real-time monitoring and diagnosis system for pile foundation condition based on digital twin technology according to claim 2, characterized in that, Also includes: The digital twin feature library module is used to store the pile foundation geometric model, material parameter model, and historical health status feature vectors. The real-time diagnostic engine module is used to fuse the corrected monitoring data sequence from the multi-source data acquisition module with the acoustic emission signals acquired by the acoustic emission sensor array, and combine the pile foundation geometric model and material parameter model in the digital twin feature library module to generate a pile stress distribution cloud map and damage location coordinates.

4. The real-time monitoring and diagnosis system for pile foundation condition based on digital twin technology according to claim 3, characterized in that, The real-time diagnostic engine module generates a pile stress distribution cloud map and damage location coordinates, including: The sound wave propagation path deviation is calculated based on the time delay difference of the reflected waveforms of the positioning markers in the acoustic emission signal. The coordinates of the positioning markers in the pile foundation geometric model are corrected based on the deviation of the sound wave propagation path. The corrected monitoring data sequence of the multi-source data acquisition module is mapped onto the surface of the corrected pile foundation geometric model; The stress gradient distribution of the pile body is calculated using a material parameter model, and a stress distribution cloud map of the pile body is generated. Identify the coordinates of abnormal stress concentration areas in the pile stress distribution cloud map and use them as damage location coordinates.

5. The real-time monitoring and diagnosis system for pile foundation condition based on digital twin technology according to claim 4, characterized in that, Also includes: The status trend prediction module is used to extract historical health status feature vectors from the digital twin feature library module; Use the current damage location coordinates and stress gradient values ​​as new feature vectors; The fused historical health status feature vector sequence is compressed in terms of feature dimension and aligned in time sequence. State evolution rate analysis is performed based on the compressed and aligned feature vector sequence, and the damage development rate and remaining lifetime prediction are output.

6. The real-time monitoring and diagnosis system for pile foundation condition based on digital twin technology according to claim 5, characterized in that, Also includes: The health status grading module is used to receive the damage location coordinates from the real-time diagnostic engine module and the damage development rate from the status trend prediction module. The overall deterioration index of the pile foundation is calculated based on the weighted product of the depth distribution density of the damage location coordinates and the damage development rate. The overall deterioration index of the pile foundation is matched with the preset grading threshold range, and the pile foundation health status level code is output.

7. The real-time monitoring and diagnosis system for pile foundation condition based on digital twin technology according to claim 6, characterized in that, Also includes: The maintenance strategy generation module is used to receive the pile foundation health status level code from the health status grading module; A pre-defined maintenance measures library is established based on the pile foundation health status level code index; The remaining life prediction value of the state trend prediction module is combined to optimize the maintenance measure execution time parameters; Generate a list of pile foundation maintenance strategies that includes maintenance measure types and execution time windows.

Images