Real-time analysis system for pile driving process of pipe pile based on multi-sensor fusion

Abstract

The application relates to the field of foundation construction monitoring, and discloses a multi-sensor fusion pipe pile sinking process real-time analysis system, which comprises a sensor group, a data acquisition and preprocessing module connected with the sensor group, a real-time analysis module connected with the data acquisition and preprocessing module, a feature analysis submodule for generating a dynamic time sequence feature vector, a health diagnosis submodule for pile body health state diagnosis, a dynamic threshold determination submodule for determining a dynamic judgment threshold, a comprehensive judgment submodule, an early warning module connected with the real-time analysis module, and a data storage and visualization module connected with the real-time analysis module. A dynamic early warning threshold model is constructed according to real-time geological changes and construction stages, the problems of false alarms and missed alarms caused by the fact that traditional fixed threshold early warning cannot adapt to variable geology and working conditions are solved, the accuracy and reliability of early warning judgment are greatly improved, and the system can accurately identify real risks.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of foundation construction monitoring, in particular to a multi-sensor fusion pipe pile sinking process real-time analysis system. BACKGROUND

[0002] Pipe piles, especially prestressed high-strength concrete (PHC) pipe piles, as a kind of efficient and reliable foundation component, play a key role in the foundation treatment of large-scale projects such as ports, bridges and high-rise buildings. In these projects, pipe pile foundations often need to bear complex dynamic and static loads, so the sinking quality, especially the verticality of the pile body and the accurate arrival of the pile end bearing layer, is required to be extremely strict. In order to ensure the sinking quality, the industry generally uses dynamic monitoring technology based on stress wave theory to collect force and acceleration signals during hammering through sensors installed at the top of the pile, and then analyzes the pile body integrity and single pile vertical bearing capacity. This technology provides an important quantitative basis for pile foundation construction quality control and is one of the core technical means to ensure the quality of pile sinking projects.

[0003] However, as engineering projects develop towards more complex geological conditions and higher requirements for refinement and intelligent construction, the existing technology also shows further space for improvement in dealing with the dynamics and uncertainties of the construction process. On the one hand, traditional monitoring methods have limitations in information dimension. For example, focusing only on penetration depth cannot provide real-time insight into sudden changes in soil resistance around the pile, and analyzing only stress waves at the top of the pile can easily overlook the risk of local stress concentration caused by the inclination of the pile body posture, which are potential hazards leading to pile body damage or failure to reach the bearing layer at the pile end. On the other hand, field construction monitoring still largely relies on manual observation and offline analysis, which not only has a lag in responding to sudden conditions such as hammering overload, but more importantly, fails to effectively integrate the vast amount of real-time data generated during construction with geological survey information. Therefore, the sinking data of each pile, as the most authentic in-situ detection of local geology, has not been fully exploited to dynamically correct the geological understanding of the entire site area and to provide precise and forward-looking guidance for subsequent pile foundation work. SUMMARY

[0004] In view of the shortcomings of the prior art, the present application provides a multi-sensor fusion pipe pile sinking process real-time analysis system, which solves the problem that the existing pipe pile sinking monitoring technology has a single information dimension, a static analysis model and a lack of data closed-loop learning ability, making it difficult to accurately evaluate and real-time alert the health status of the pile body under complex working conditions, and unable to dynamically optimize the guidance of pile group construction in the site area.

[0005] To achieve the above object, the application provides a multi-sensor fusion pipe pile sinking process real-time analysis system, which comprises:

[0006] a sensor group for collecting hammering force signals, pile body acceleration signals, pile body inclination signals, pile body penetration signals and soil pressure signals around piles in a pipe pile sinking process;

[0007] a data acquisition and preprocessing module connected with the sensor group, for synchronously collecting and filtering and denoising the signals obtained by the sensor group in the pipe pile sinking process to obtain processed signals;

[0008] a real-time analysis module connected with the data acquisition and preprocessing module, wherein the real-time analysis module comprises:

[0009] a feature analysis submodule for performing dynamic time series feature analysis on the processed hammering force signals, pile body acceleration signals, pile body penetration signals and soil pressure signals around piles to generate a dynamic time series feature vector;

[0010] a health diagnosis submodule for diagnosing a pile body health state based on the dynamic time series feature vector;

[0011] a dynamic threshold determination submodule for determining a dynamic judgment threshold for pile sinking state judgment in combination with a preset field area geological model and a real-time pile sinking working condition;

[0012] a comprehensive judgment submodule for comprehensively judging a pile sinking state based on the results of the pile body health state diagnosis, the dynamic judgment threshold for the pile sinking state judgment and the pile body inclination signals;

[0013] wherein the field area geological model comprises initial depth-related soil layer boundary data and initial geological adjustment coefficients corresponding to each soil layer, and the real-time pile sinking working condition is a real-time updated data set, including a pile end depth and hammering energy corresponding to a current hammering and the results of the pile body health state diagnosis;

[0014] a warning module connected with the real-time analysis module, for sending a warning signal according to the judgment results of the pile sinking state;

[0015] Specifically, the system comprehensively captures multi-dimensional physical signals in the pile sinking process through a sensor group, and ensures high-quality synchronous input of data by using a data acquisition and preprocessing module. The real-time analysis module does not make isolated single-point judgments, but realizes precise and adaptive evaluation of the pile sinking state through four logically progressive steps of feature analysis, health diagnosis, dynamic threshold determination and comprehensive judgment. Finally, the analysis results are presented to the user in the form of direct warning signals through the warning module, thereby realizing full-process coverage from raw data acquisition to intelligent decision support, and greatly improving the safety, quality and efficiency of pile sinking construction.

[0016] Preferably, the feature analysis submodule is specifically used for:

[0017] Within the time window of each hammering event, a dynamic time sequence feature vector is constructed, which can fully characterize the excitation, response and feedback process of the hammering;

[0018] The dynamic time sequence feature vector specifically includes:

[0019] An energy input feature for characterizing external energy input;

[0020] A dynamic response feature for characterizing the immediate absorption of hammering energy and vibration response of the pile body;

[0021] A displacement feedback feature for characterizing the efficiency of conversion of hammering energy into effective penetration work;

[0022] Resistance response delay and resistance response peak for jointly characterizing pile-soil interaction characteristics;

[0023] Specifically, the dynamic time sequence feature vector decomposes each hammering event into an excitation, response and feedback process with clear physical logic. For this purpose, the feature vector constructed by the present scheme is not a simple list of multi-source data, but a structured portrait of this physical process. Among them, the energy input feature represents the excitation from the outside world; the dynamic response feature and the displacement feedback feature jointly depict the response process of the pile body system to the excitation, and respectively reflect the energy absorption and conversion efficiency; and the resistance response related features characterize the feedback to the pile body movement. By constructing such a feature vector that can fully map the physical process, the present scheme provides a highly condensed information, clear physical meaning and extremely robust data basis for subsequent pile health diagnosis, geological model inversion and other advanced analysis tasks, which is much better than direct analysis of the original time sequence signal.

[0024] Preferably, the dynamic time sequence feature vector is constructed in the following manner:

[0025] The energy input feature is obtained by integrating the hammering force signal within the time window of each hammering event.

[0026] the dynamic response feature is obtained by integrating the acceleration signal of the pile body and taking the peak value of the velocity in the time window;

[0027] the displacement feedback feature is obtained by calculating the rate of change of the pile body penetration signal and taking the peak value of the rate of change in the time window;

[0028] the resistance response delay is obtained by calculating the time difference between the time when the soil pressure signal around the pile reaches the peak value and the time when the hammering force signal reaches the peak value;

[0029] the resistance response peak value is obtained by taking the peak value of the soil pressure signal around the pile in the time window;

[0030] Specifically, it is clear that each feature component is derived from one or more specific original sensor signals and is realized through explicit mathematical operations (such as integration, peak value calculation, time difference calculation, etc.). For example, the energy input feature is accurately quantified by integrating the hammering force signal, and the dynamic response feature is captured by integrating the acceleration signal to capture the peak value of the velocity response. This design ensures that the construction process of the feature vector is objective, repeatable, and closely anchored to the physical reality. The final effect is to successfully and stably transform a complex, multi-physical field coupled hammering event into a low-dimensional, informationized digital fingerprint.

[0031] Preferably, the pile body health state diagnosis specifically includes:

[0032] A healthy hammering response baseline model is established at the initial stage of pile sinking, and a health deviation degree between the current dynamic time series feature vector and the healthy hammering response baseline model is calculated in real time.

[0033] Preferably, the healthy hammering response baseline model is established by statistically averaging the dynamic time series feature vectors of continuous multiple hammerings at the initial stage of pile sinking;

[0034] The health deviation degree is obtained by calculating the normalized Euclidean distance between the current dynamic time series feature vector and the healthy hammering response baseline model;

[0035] Specifically, the present scheme proposes a dynamic evaluation model based on baseline deviation. It does not rely on an absolute, preset damage threshold, but at the initial stage of pile sinking of each pile, it uses the hammer response of the pile body which is necessarily in a healthy state to learn online and establish a health hammer response baseline model that belongs to the pile. During the subsequent pile sinking process, the system quantifies the trend of the pile body's health status by calculating the distance between the current hammer response and the health baseline, i.e., the health deviation. The great advantage of this method is its high adaptability, which can effectively exclude the normal response differences caused by different piles and different geological conditions, thus extremely sensitively capturing the subtle changes in the dynamic response caused by abnormal pile body state (such as the presence of micro-cracks, joint loosening, etc.), achieving early and accurate warning of potential risks to the pile body.

[0036] Preferably, the determination method of the dynamic judgment threshold is:

[0037] multiplying a base threshold value by a geological adjustment coefficient determined by matching the current pile end depth with the field area geological model;

[0038] and multiplying a working condition adjustment coefficient determined by the real-time pile sinking stage to obtain the dynamic judgment threshold;

[0039] Specifically, the dynamic judgment threshold is generated in real time according to two dynamic adjustment coefficients, namely the geological adjustment coefficient and the working condition adjustment coefficient. The former enables the threshold to be adaptively adjusted according to the different geological strata (provided by the field area geological model) in which the pile end is located, for example, appropriately relaxed in hard soil layers and more stringent in weak interlayers; the latter responds to changes in real-time sinking conditions. In this way, the present invention allows the judgment standard to change dynamically, enabling it to closely match the actual geological conditions and construction conditions on site, thereby greatly improving the accuracy of the warning and effectively avoiding the frequent false positives or false negatives caused by fixed standards in traditional fixed threshold systems.

[0040] Preferably, the real-time analysis module further comprises a self-learning and optimization submodule, which specifically comprises:

[0041] a geological response map generation unit for generating a geological response map of the pile site after the single pipe pile sinking operation is completed;

[0042] a model online correction unit for online correction of the field area geological model based on the geological response map.

[0043] Preferably, the geological response map generation unit is specifically used for:

[0044] The dynamic time-series feature vectors generated by the real-time analysis module throughout the entire process of the pipe pile are invoked to form a dynamic time-series feature vector sequence;

[0045] Simultaneously, the pile tip depth, calculated based on the pile penetration signal and corresponding one-to-one with each dynamic time-series feature vector, is acquired to form a depth sequence corresponding to the dynamic time-series feature vector sequence.

[0046] An unsupervised clustering algorithm is used to process the dynamic time-series feature vector sequence to identify cluster centers that characterize different dynamic response modes and determine the cluster affiliation of each dynamic time-series feature vector.

[0047] Based on the clustering of each dynamic time-series feature vector and combined with the depth sequence, hammer impact event sequences that are continuous in depth and belong to the same cluster are merged to determine the data-driven geological response layer boundary.

[0048] Preferably, the online model correction unit is specifically used for:

[0049] The geological response layer boundary determined in the geological response map is taken as the observation boundary depth;

[0050] By combining the a priori boundary depth in the initial geological model of the site with the observed boundary depth, the corrected posterior boundary depth is calculated to update the soil layer boundary of the geological model of the site.

[0051] Calculate the mean of the dynamic time series feature vectors falling within each corrected geological layer, and update the geological adjustment coefficient corresponding to the geological layer based on the difference between the mean and the cluster center.

[0052] Preferably, the process by which the real-time analysis module determines the operating condition adjustment coefficient specifically includes:

[0053] First, multiple pile driving conditions are predefined and corresponding condition adjustment coefficient values ​​are preset for each pile driving condition. The multiple pile driving conditions include multiple conventional pile driving conditions and one risk observation condition.

[0054] The current pile driving condition is determined by a working condition switching rule. The working condition switching rule switches between the conventional pile driving conditions under normal pile driving conditions based on the real-time obtained pile tip depth or penetration rate of change. When a pile health status diagnosis result indicating that there is a risk in the pile body is received, the current pile driving condition is forcibly switched to the risk observation condition.

[0055] Finally, the preset working condition adjustment coefficient value corresponding to the current pile driving working condition determined by the working condition switching rules is selected.

[0056] This invention provides a real-time analysis system for the pipe pile driving process based on multi-sensor fusion, which has the following advantages:

[0057] This invention constructs a dynamic early warning threshold model that can automatically adjust the judgment criteria based on real-time geological changes and construction stages. This achieves the effect that the early warning standard line is no longer fixed, but can be intelligently adjusted according to whether the pile is driven into a hard soil layer or a soft soil layer, and whether it is in the initial driving or final hammering stage. This solves the problem of false alarms and missed alarms caused by the inability of traditional fixed threshold early warning methods to adapt to changing geological and working conditions. This greatly improves the accuracy and reliability of early warning judgment, enabling the system to accurately identify real risks.

[0058] This invention establishes a site self-learning optimization process by continuously refining the geological model based on completed pile foundation data and guiding subsequent construction. This process ensures that after each pile is driven, the system automatically reviews and learns, making the 3D geological map of the entire construction site increasingly accurate. It also intelligently recommends optimal construction parameters for the next pile. This solves the problem of traditional construction relying solely on static geological reports before construction begins, which cannot dynamically update understanding using new data during construction, leading to delayed construction guidance and recurring risks. This transforms the construction process from a series of isolated operations into a continuously iterating and self-optimizing intelligent process, enhancing the predictability of construction and improving the overall safety and operational efficiency of the project.

[0059] This invention integrates multiple sensor signals to construct a comprehensive health diagnostic index that fully characterizes a single hammer impact event. It moves beyond simply looking at isolated data such as hammer force or penetration depth, instead combining multiple key aspects of a single hammer impact, including energy transfer, pile response, and pile-soil interaction, into a single comprehensive health score. This solves the problem of traditional monitoring methods, which, due to their limited information dimensions, struggle to comprehensively assess complex pile-soil interactions and are prone to misjudging potential pile damage or abnormal obstruction. It significantly improves the accuracy and sensitivity of abnormal diagnosis for single hammer impact events, enabling earlier detection of minute anomalies that are difficult to detect using traditional methods, thus providing a higher level of safety assurance for preventing pile structural damage. Attached Figure Description

[0060] Figure 1 This is a logical structure block diagram of the system according to an embodiment of the present invention;

[0061] Figure 2 This is a sensor group deployment diagram according to an embodiment of the present invention;

[0062] Figure 3 This is a schematic diagram of the acquisition and preprocessing module according to an embodiment of the present invention;

[0063] Figure 4 This is a flowchart of the real-time analysis module in an embodiment of the present invention;

[0064] Figure 5 This is a schematic diagram of the early warning module according to an embodiment of the present invention;

[0065] Figure 6 This is a schematic diagram of the user interface of the data storage and visualization module in an embodiment of the present invention. Detailed Implementation

[0066] See Figure 1 , Figure 1 This is a logical structure block diagram of a multi-sensor fusion real-time analysis system for the pipe pile driving process according to an embodiment of the present invention. The embodiment of the present invention provides a multi-sensor fusion real-time analysis system for the pipe pile driving process, which may include: a sensor group 1, a data acquisition and preprocessing module 2, a real-time analysis module 3, an early warning module 4, and a data storage and visualization module 5.

[0067] Sensor group 1 is used to collect hammer impact force signals, pile acceleration signals, pile inclination signals, pile penetration signals, and soil pressure signals around the pile during the physical process of pipe pile driving. The output terminal of sensor group 1 is electrically or wirelessly connected to the input terminal of data acquisition and preprocessing module 2.

[0068] The data acquisition and preprocessing module 2 is used to receive multiple raw signals from sensor group 1. Internally, this module includes an analog-to-digital converter and a signal processor, performing the following main operations: First, it assigns a unified timestamp to all acquired signal data points for timestamp alignment; second, it filters and denoises the timestamp-aligned signal data, and the processed clean data streams are output to the real-time analysis module 3.

[0069] The input terminal of the real-time analysis module 3 is connected to the output terminal of the data acquisition and preprocessing module 2. For the driving process of a single pipe pile, the real-time analysis module 3 performs operations in the following order:

[0070] First, the received data stream is analyzed to identify each independent hammering event, and a dynamic temporal feature vector is constructed for each hammering event. Its expression is:

[0071]

[0072] In the formula, This is the dynamic time-series characteristic vector. As an energy input characteristic, As a dynamic response feature, For displacement feedback characteristics, For the resistance response delay, This represents the peak value of the resistance response.

[0073] Secondly, the real-time analysis module 3 diagnoses the health status of the pile body based on historical dynamic time-series feature vector sequences. Simultaneously, this real-time analysis module 3 retrieves a preset site geological model from the data storage and visualization module 5 and, combined with the current real-time pile driving conditions, determines a dynamic judgment threshold. .

[0074] Dynamic threshold judgment The calculation formula is:

[0075]

[0076] In the formula, To dynamically determine the threshold, Based on the threshold, This is the geological adjustment coefficient. This is the operating condition adjustment factor.

[0077] Geological adjustment coefficient The determination of the depth is accomplished by the real-time analysis module 3 through a table lookup operation. First, the real-time analysis module 3 obtains the pile penetration signal from the signal output by the data acquisition and preprocessing module 2, and then parses the real-time pile tip depth corresponding to the current hammering event from the pile penetration signal. Next, the real-time analysis module 3 uses the obtained real-time pile tip depth as the query basis to search within the site geological model stored in the data storage and visualization module 5. The site geological model defines a one-to-one correspondence between different depth intervals and pre-calibrated geological adjustment coefficient values ​​in its data structure. At this point, the real-time analysis module 3 matches the depth interval into which the real-time pile tip depth falls, thereby finding the corresponding geological adjustment coefficient value in the site geological model, and uses this geological adjustment coefficient value as the geological adjustment coefficient used in the calculation of the dynamic judgment threshold. .

[0078] Operating condition adjustment coefficient The determination of the driving conditions is dynamically accomplished by a finite state machine within the real-time analysis module 3. The finite state machine predefines multiple pile driving conditions, including the initial driving stage, the stable penetration stage, and the final hammer stage, and sets a corresponding adjustment coefficient value for each condition. Based on real-time acquired parameters such as pile tip depth and penetration rate of change, the real-time analysis module 3 automatically switches between these different pile driving conditions to select the adjustment coefficient that matches the current construction stage. In addition, the finite state machine includes a special risk observation phase, which forms a closed-loop feedback with the output of the health diagnosis submodule.

[0079] When the health deviation calculated by the health diagnosis submodule exceeds a preset fixed health risk threshold multiple times, the real-time analysis module 3 will force the finite state machine to switch to the risk observation stage and select the highest value of the working condition adjustment coefficient set for the risk observation stage.

[0080] Finally, the real-time analysis module 3 comprehensively judges the current pile driving status by combining the results of the pile health status diagnosis and the calculated dynamic judgment threshold, and outputs the judgment result data to the early warning module 4 and the data storage and visualization module 5.

[0081] The early warning module 4 has its input connected to the output of the real-time analysis module 3. When the received judgment result data meets the preset early warning trigger conditions, the early warning module 4 outputs one or more early warning signals. These early warning signals can be high or low level signals to drive the audible and visual alarm device, or specific data packets to notify the host computer control system.

[0082] The data storage and visualization module 5 communicates bidirectionally with the real-time analysis module 3. This module 5 includes non-volatile storage media and a graphics processing unit. It primarily receives and persistently stores all raw data, processed data, feature vectors, model parameters, and judgment results from the real-time analysis module 3. Simultaneously, it processes this data into graphical interface elements for visualization on a display device.

[0083] The system in this embodiment also includes a closed-loop workflow for site self-learning optimization. After the driving of a single pipe pile is completed, the real-time analysis module 3 begins to execute this workflow.

[0084] First, the real-time analysis module 3 integrates the dynamic time-series feature vector sequence of the entire process of the pile and the corresponding depth data to generate a high-precision geological response map of the pile location.

[0085] Subsequently, the real-time analysis module 3, based on the high-precision geological response map, performs online correction on the site geological model stored in the data storage and visualization module 5, and writes the corrected new version of the geological model data back to the storage unit of the data storage and visualization module 5. Simultaneously, based on the high-precision geological response map, the real-time analysis module 3 performs retrospective analysis to extract the optimal hammering strategy, and also stores the data of this optimal hammering strategy in the data storage and visualization module 5.

[0086] Before the next pipe pile is driven, the real-time analysis module 3 first reads the latest version of the site geological model and the optimal hammering strategy for adjacent pile positions from the data storage and visualization module 5. In subsequent real-time analysis, the real-time analysis module 3 uses this updated geological model to calculate the geological adjustment coefficient and displays the optimal hammering strategy on the visualization interface, thereby using the construction information of the previous pile to guide the operation of the next pile.

[0087] like Figure 2 As shown, Figure 2 This is a schematic diagram of sensor group deployment according to an embodiment of the present invention. The system provided by the embodiment of the present invention is based on acquiring a multi-dimensional, high-fidelity dataset that can comprehensively characterize the complex dynamic behavior of the pile driving process through an optimized sensor group 1.

[0088] In one specific embodiment, the sensor group 1 is configured and deployed as follows:

[0089] Pressure sensor 101 is used to acquire hammer impact force signal. In this embodiment, a piezoelectric or strain gauge pressure sensor with high frequency response characteristics (e.g., a natural frequency greater than 20 kHz) and high dynamic range is selected. This pressure sensor 101 is installed between the hammer pad and the pile cap of the pile hammer. This deployment location is chosen to allow for the most direct and minimally distorted measurement of the initial impact energy input at the very beginning of the hammer impact energy transmission path, before the energy is absorbed and dissipated by the pile cap and pile body. The acquired hammer impact force signal... It serves as the direct data source for subsequent energy input feature calculations.

[0090] Accelerometer 102 is used to acquire the acceleration signal of the pile body. In this embodiment, at least two micro-electro-mechanical systems (MEMS) triaxial accelerometers 102 are used. These two accelerometers 102 are arranged symmetrically along the radial direction of the pile (e.g., installed 180 degrees apart on the same horizontal section). The installation position is set at a distance greater than twice the pile diameter to avoid the complex near-field reflection region caused by boundary effects of the hammering stress wave near the pile top, thereby measuring a more stable plane wave propagation signal. The symmetrical arrangement is to eliminate the interference of pile bending vibration on the longitudinal acceleration measurement. During hammering, the pile will bend, causing the two symmetrically positioned accelerometers 102 to generate acceleration components of equal magnitude but opposite direction in the horizontal direction. By averaging the longitudinal acceleration signals measured by the two sensors 102, the signal components introduced by bending vibration can be effectively canceled out, thereby extracting a purer acceleration that only reflects the overall longitudinal penetration motion of the pile. The calculation formula is:

[0091]

[0092] In the formula, This is the processed longitudinal acceleration signal of the pile body used for subsequent analysis; The longitudinal acceleration signal measured by the first accelerometer; This is the longitudinal acceleration signal measured by the second accelerometer.

[0093] Inclination sensor 103 is used to acquire the pile inclination signal. A biaxial inclination sensor is selected and is fixed to the top of the geometric center of the pile cap. Inclination sensor 103 measures in real time the inclination angle of the pile relative to the vertical line in two mutually perpendicular orthogonal planes (e.g., the XZ plane and the YZ plane, where Z is the vertical direction). and The pile inclination signal is a key parameter for evaluating construction quality and preventing excessive pile inclination from causing a decrease in bearing capacity or structural damage.

[0094] Laser displacement sensor 104 is used to acquire pile penetration signal. A non-contact laser displacement sensor 104 is selected and mounted on a highly stable reference base independent of the pile driving equipment, which is directly set on the stable ground. The laser beam of the laser displacement sensor 104 is aligned with a high-reflectivity target fixed to the pile cap. The technical purpose of this deployment method is to eliminate errors in displacement measurement caused by vibrations or minor settlements of the pile driving equipment (e.g., pile frame) under the hammer impact reaction force. By establishing an independent measurement benchmark, the laser displacement sensor 104 can accurately obtain the absolute penetration depth of the pile relative to the ground, which is the basis for calculating the pile bearing capacity and analyzing pile-soil interaction.

[0095] Earth pressure sensor 105 is used to acquire earth pressure signals around the pile. Based on the geological survey report, vibrating wire or piezoresistive earth pressure cells are selected and pre-installed at the design elevation of the top interface of one or more key bearing soil layers (e.g., gravel or dense sand layers) before construction. The deployment of earth pressure sensors 105 provides the system with a direct observation window into the interaction between the ground and the pile. As the pile tip approaches and penetrates the bearing layer, the earth pressure sensors 105 can directly measure the changes in soil compressive stress caused by pile penetration. The soil pressure signal around the pile is crucial for calculating the resistance response delay and peak resistance response characteristics, directly reflecting the stratum's response speed and resistance strength to hammer energy input.

[0096] like Figure 3 As shown, Figure 3 This is a schematic diagram of a data acquisition and preprocessing module according to an embodiment of the present invention. The data acquisition and preprocessing module 2 is mainly used to solve the time synchronization problem of signals from multiple heterogeneous sensors and to clean up the original signals containing noise, so as to provide high-quality and high-reliability data input for the subsequent real-time analysis module 3.

[0097] In one specific embodiment, the data acquisition and preprocessing module 2 is implemented in hardware as a multi-channel synchronous data acquisition system. This system is a data acquisition card with multiple parallel analog-to-digital converter (ADC) channels. To achieve high-precision time synchronization between channels, the acquisition card integrates a high-precision clock source, such as a temperature-compensated crystal oscillator (TCXO) or a GPS-disciplined oscillator (GPSDO) that is time-disciplined by receiving signals from the Global Positioning System (GPS). The uniform, in-phase sampling clock signal generated by this clock source is fed in parallel to each ADC channel. Therefore, when multiple analog signals such as impact force, acceleration, and penetration from sensor group 1 are input to the acquisition card, they are sampled and converted into digital signals at physically simultaneous moments. This hardware-based synchronization mechanism ensures that the timestamp alignment error between different signal data streams is on the order of microseconds, thus providing an accurate data foundation for subsequent analysis of the timing relationship between signals (such as resistance response delay).

[0098] After synchronous acquisition is completed, the signal processor inside the data acquisition and preprocessing module 2, such as a digital signal processor (DSP), performs targeted filtering and noise reduction operations on the digitized signal stream.

[0099] For pile acceleration signal and pile penetration signal This embodiment employs the Kalman filter algorithm for processing. The Kalman filter algorithm is an optimized recursive data processing algorithm based on a state-space model. Its technical advantage lies in its ability to fuse multi-source information and effectively suppress process noise and measurement noise. In this embodiment, the vertical motion of the pile is established as a linear dynamic system, and its state-space model is defined as follows:

[0100] The system's state vector At any moment Defined as:

[0101]

[0102] In the formula, Let be the state vector of the system. For the pile body at all times The vertical position; For the pile body at all times The vertical velocity.

[0103] The system's state transition equation describes how the state vector changes from time 1 to 2. Evolution to time :

[0104]

[0105] In the formula, Let be the state vector of the system. The state transition matrix is ​​defined based on kinematic principles. ; The sampling time interval; To control the input matrix, it is defined as follows: ; To control the input, this is at time [time]. Measured pile acceleration ; The process noise is assumed to be white noise following a zero-mean Gaussian distribution, and its covariance matrix is... .

[0106] The system's observation equations describe the relationship between the measured values ​​and the state vector:

[0107]

[0108] In the formula, For at any time The measurement vector, here representing the pile penetration signal measured by the laser displacement sensor 104. The observation matrix is ​​defined as follows: , indicating that the position component in the state vector is directly measured; To measure the noise, we assume it is white noise following a zero-mean Gaussian distribution, with its covariance matrix being... ; Let be the state vector of the system.

[0109] By performing standard Kalman filter prediction and update iteration on the above model, the data acquisition and preprocessing module 2 can output the optimal estimate sequence of the true position and velocity of the pile body. This optimal estimate sequence effectively filters out sensor measurement noise and high-frequency vibration interference caused by hammer impact.

[0110] For the pile perimeter earth pressure signal and the pile inclination signal, this embodiment employs a moving average filtering algorithm. Both of these signals are slowly varying or quasi-static signals; the analysis focuses on the macroscopic trend rather than the instantaneous high-frequency fluctuations caused by hammering. The formula for calculating the moving average filter is:

[0111]

[0112] In the formula, For the first The filtered output value of each sampling point; The size of the sliding window is a preset integer whose value is a trade-off between smoothing effect and signal response delay. For the first The algorithm extracts the original input values ​​from each sampling point. It effectively smooths signal curves and filters out short-term random noise with low computational cost. For pile-perimeter earth pressure signals, it extracts the true pressure baseline reflecting the soil consolidation state; similarly, for tilt signals, it effectively suppresses high-frequency fluctuations in measurements caused by instantaneous hammer impact, providing the operator with a stable and reliable attitude reading.

[0113] For the hammer impact signal, this embodiment employs a Butterworth low-pass filter algorithm. This signal is a typical transient impact signal, with its key characteristic information concentrated in the extremely short main pulse. However, the signal is highly susceptible to contamination from high-frequency mechanical ringing and electromagnetic noise. The amplitude-frequency response characteristic of the Butterworth filter is defined by the following formula:

[0114]

[0115] In the formula, Signal energy transmittance, The frequency of the current signal; Frequency threshold; The degree of rigor of the filtering.

[0116] This algorithm was chosen because of its maximum passband flatness, which ensures that the original waveform and key peaks of the hammer impact force main pulse are preserved without distortion. This algorithm can efficiently filter out peaks higher than [a specific value, likely a threshold] while fully retaining this crucial impact information. The high-frequency noise is eliminated, and a smooth, clean force signal waveform is finally output, providing a reliable basis for the accurate calculation of various dynamic parameters.

[0117] After the above processing, the data acquisition and preprocessing module 2 packages the multi-channel synchronous, clean data streams with precise timestamps into structured data frames, which are then transmitted in real time to the real-time analysis module 3 via the internal high-speed bus.

[0118] like Figure 4 As shown, Figure 4 This is a flowchart of the real-time analysis module according to an embodiment of the present invention. The real-time analysis module 3 first receives a synchronous, clean data stream from the data acquisition and preprocessing module 2, and performs a series of defined calculation steps to achieve real-time, quantitative analysis and judgment of the pile driving process.

[0119] In one specific embodiment, the workflow of the real-time analysis module 3 is as follows:

[0120] First, continuously monitor the input hammer force signal. .when The amplitude exceeds a preset trigger threshold used to distinguish valid hammer blows from background noise during a rising edge. When a new hammering event occurs, the system determines that a new event has taken place. The real-time analysis module 3 records this moment as the start time of the event. And from this starting point, a fixed duration is extracted. The data segment, this duration The settings are sufficient to fully cover the entire process of a single hammer impact, from the start of the impact to the basic decay of pile vibration. All subsequent calculations are performed within this time window. The data is processed within the system.

[0121] This data segment, which fully records the entirety of a single hammer strike event, forms the basis for all subsequent analyses. Based on this, the real-time analysis module 3 immediately initiates its computation process, the main steps of which include:

[0122] Step 301: Construct the dynamic time series feature vector through the feature analysis submodule 31, as follows:

[0123] Within a defined time window, the feature analysis submodule 31 performs calculations on multiple signals to construct a dynamic temporal feature vector that can characterize the hammering event from multiple physical dimensions. The specific calculation method for each component of this vector is as follows:

[0124] Energy input characteristics It represents the effective impact energy transmitted to the pile body in a single hammer blow, through the hammer force signal within a time window. The result is obtained by numerical integration.

[0125]

[0126] In the formula, For the first Energy input characteristics of the second hammer strike; This is a hammer impact force signal; For the first The start time of the second hammering event; To analyze the duration of the time window; This represents an extremely small time segment. In the actual implementation, this integral is calculated using numerical integration methods such as the trapezoidal rule.

[0127] Dynamic response characteristics This feature characterizes the peak velocity response of the pile under hammer impact. Firstly, it is determined by analyzing the pile acceleration signal within a time window. Numerical integration is performed to obtain the pile velocity signal. The initial conditions for integration It is reset to zero at the start of each hammering event. Then, in Search for and determine its maximum value in the time series.

[0128]

[0129] In the formula, The dynamic response characteristics of the i-th hammer strike; This is the pile acceleration signal; For the first The start time of the second hammering event; To analyze the duration of the time window; same Both are extremely small time segments; It means the maximum value. for The scope of an operator.

[0130] Displacement feedback characteristics This feature characterizes the maximum penetration rate of the pile under hammer impact. It is obtained by analyzing the pile penetration rate signal within a time window. Numerical difference was performed to obtain the pile penetration rate signal, and its peak value was taken. The corresponding specific calculation formula is shown below:

[0131]

[0132] In the formula, For the first Displacement feedback characteristics of the second hammer blow; This represents the pile penetration depth signal. In practice, this differential is calculated using numerical differentiation methods such as backward difference.

[0133] Drag response delay This feature characterizes the response time of the soil around the pile to the transmission of hammer stress waves. It is obtained by calculating the time difference between the peak value of the soil pressure signal around the pile and the peak value of the hammer force signal.

[0134]

[0135] In the formula, For the first The resistance response delay of the second hammer blow; This is the earth pressure signal around the pile; This is a hammer impact force signal; For reflected wave signals The moment when it reaches its peak; Hammering force signal The moment when it reaches its peak.

[0136] Peak resistance response This feature characterizes the maximum resistance of the soil around the pile under hammer impact, obtained by capturing the soil pressure signal around the pile within a time window. The peak value was obtained.

[0137]

[0138] In the formula, For the first Peak resistance response of the second hammer strike; This is the earth pressure signal around the pile.

[0139] Step 302: Non-invasive diagnosis of the health status of the pile body is performed through the health diagnosis submodule 32, as follows:

[0140] This step is based on a healthy hammer impact response baseline model. The healthy hammer impact response baseline model is established during the initial stabilization stage of pipe pile driving (e.g., after the pile verticality has stabilized, the penetration depth is between 2 and 5 meters).

[0141] Before establishment, the system continuously collects and processes a series (e.g., N=10 to 20) of valid hammer impact events confirmed to be in a healthy state. For each hammer impact, the system executes step 301 to extract its dynamic temporal feature vector through the health diagnosis submodule 32. After obtaining a set of healthy sample vectors, the system calculates the element-wise mean of the sample set to obtain a baseline feature vector representing the most typical healthy response, and simultaneously calculates the standard deviation vector of each feature element to quantify the normal fluctuation range. These two together constitute a complete health baseline model.

[0142] After the health baseline model is established, the real-time analysis module 3 also calculates its feature vector for each subsequent new hammer blow. Following that, by... The health deviation is calculated by comparing it with a health baseline model. The health deviation is the normalized Euclidean distance, and its specific calculation formula is as follows:

[0143]

[0144] In the formula, For the first Health deviation of the second hammer blow; The dimension of the feature vector is 5 in this embodiment; For the first The th eigenvector of the th feature vector One component; The first of the baseline model mean vectors One component; For the baseline model, the first The standard deviation of each characteristic component. The purpose of normalization using standard deviation is to eliminate the influence of differences in physical units and numerical ranges among the characteristic components, making them more uniform. It becomes a dimensionless, directly comparable indicator that comprehensively reflects the degree of deviation from multi-dimensional responses.

[0145] Step 303: The adaptive threshold model and closed-loop feedback coupling geology and working conditions are determined by the dynamic threshold determination submodule 33, as follows:

[0146] The dynamic threshold determination submodule 33 is used to generate a dynamic judgment threshold that can adapt to field conditions. The dynamic threshold determination submodule 33 first determines the threshold based on the real-time pile penetration signal. The system queries a structured geological model of the site, stored in the data storage and visualization module 5. The structured geological model is a data structure that maps depth ranges to geological units and corresponding parameters. The dynamic threshold determination submodule 33 retrieves the corresponding geological adjustment coefficient based on the geological unit where the current pile tip depth is located. Meanwhile, the dynamic threshold determination submodule 33 maintains a finite state machine describing the pile driving stage, whose states include at least:

[0147] Initial attack phase (corresponding coefficient) );

[0148] Stable penetration phase (corresponding coefficient) (1.0).

[0149] Final hammer stage (corresponding coefficient) );

[0150] Risk observation phase (corresponding coefficient) The real-time analysis module 3 determines and switches the current state based on information such as the penetration rate of change, thereby determining the working condition adjustment coefficient. .

[0151] This step, together with the diagnostic results of step 302, forms a closed-loop feedback (i.e., step 304), as follows: If the health deviation... If the number of consecutive M instances (M being a preset integer) exceeds a preset fixed health risk threshold related to the safety of the pile structure, the risk is considered to be M times consecutively. The finite state machine is then forced to transition to the risk observation phase, thereby selecting the operating condition adjustment coefficient with the highest value. Ultimately, the threshold is determined dynamically. The following formula is used to calculate:

[0152]

[0153] In the formula, For dynamic threshold determination; This is a preset base threshold; This is the geological adjustment coefficient; This is the operating condition adjustment factor.

[0154] Step 304: Perform comprehensive judgment and output through the comprehensive judgment submodule 34, as follows:

[0155] The comprehensive judgment submodule 34 will calculate the health deviation. With dynamic judgment threshold The comparison is performed. Simultaneously, the comprehensive judgment submodule 34 also receives the pile inclination signal from the data acquisition and preprocessing module 31 (or directly from the sensor group). It compares the current pile inclination with a preset pile inclination safety threshold (e.g., 2 degrees).

[0156] If health deviation Greater than the dynamic judgment threshold Alternatively, if the pile inclination exceeds a preset inclination safety threshold (i.e., the pile posture is abnormal), the current hammering event is determined to be abnormal; otherwise, it is normal. Finally, the comprehensive judgment submodule 34 generates a result containing a judgment result code and a feature vector. Health deviation Dynamically determine threshold The structured data package containing the pile's components and coefficients, as well as the current pile inclination and its judgment results, is output to the early warning module 4 and the data storage and visualization module 5.

[0157] Step 305: Perform post-pile analysis and online correction of the site geological model through the self-learning and optimization submodule 35, as detailed below:

[0158] After all the pile driving operations for a single pipe pile are completed, the self-learning and optimization submodule 35 also performs a post-pile analysis and model optimization function, the specific process of which is as follows:

[0159] The geological response map generation unit 351 generates a high-precision geological response map. The geological response map generation unit 351 first retrieves and integrates all the structured data packets output and stored by the comprehensive judgment submodule 34 during the entire pile driving process from the data storage and visualization module 5.

[0160] By parsing these data packets, the geological response map generation unit 351 constructs two complete time series: one is the dynamic time series feature vector sequence; the other is the pile tip depth information corresponding to each vector (this depth information is calculated and recorded in real time based on the pile penetration signal), thereby constructing a depth sequence that matches the dynamic time series feature vector sequence.

[0161] Subsequently, an unsupervised clustering algorithm (such as K-means clustering or Gaussian mixture model) is used to process the feature vector sequence to identify cluster centers representing different dynamic response modes and determine the clustering result corresponding to each dynamic time-series feature vector. Based on the clustering result of each dynamic time-series feature vector, the system combines the depth sequence to merge hammer impact event sequences that are continuous in depth and belong to the same cluster (here, a hammer impact event sequence refers to a group of single hammer impact events that are continuous in depth and have the same clustering). The start and end depths are then determined as the boundaries of the data-driven geological response layer, thereby generating a high-precision geological response map of the pile location.

[0162] The online correction of the site geological model is based on the high-precision geological response map generated by the geological response map generation unit 351, and then the site geological model is corrected online by the model online correction unit 352.

[0163] First, the geological response layer boundaries determined in the high-precision geological response map are used as high-confidence observation boundary depths. The system uses a fusion algorithm (such as Bayesian update or weighted average) to combine the prior boundary depths in the initial site geological model with these observation boundary depths to calculate the corrected posterior boundary depths, and uses these depths to update the soil layer boundary information of the site geological model.

[0164] Finally, for each corrected geological layer, the system calculates the statistical mean of all dynamic time-series feature vectors falling within that depth range, and recalibrates or updates the geological adjustment coefficient corresponding to that geological layer based on this mean vector.

[0165] To achieve this goal, after completing the above corrections, the self-learning and optimization submodule 35 will re-output the complete site geological model, including the corrected soil layer boundary information and updated geological adjustment coefficients, to the data storage and visualization module 5, and overwrite and update the old version model stored therein. This realizes the self-learning and iterative optimization of the site geological model, providing more accurate prior guidance for subsequent pile foundation construction.

[0166] See Figure 5 , Figure 5This is a schematic diagram of an early warning module according to the present invention. This section elaborates on the internal logic and specific implementation of the early warning module 4. The role of the early warning module 4 in the system is to convert the quantitative, high-dimensional analysis and judgment results output by the real-time analysis module 3 into early warning signals with clear levels and physical forms, which can be directly perceived by operators or directly executed by automated equipment.

[0167] In one specific embodiment, the early warning module 4 can be implemented in hardware as an independent microcontroller unit (MCU) or as an independent software task with high real-time priority running in the system's main controller. The input terminal of the early warning module 4 is connected to the output terminal of the real-time analysis module 3 via an internal data bus, and is used to receive data including the judgment result code and health deviation degree. Structured data packets containing information such as [list of information]. The output of the early warning module 4 includes multiple physical channels, which are connected to the audible and visual alarm device and the control system of the pile driving equipment, respectively.

[0168] The early warning module 4 is internally implemented as a multi-level early warning logic processor. This processor includes internal state variables, such as a counter to record the number of consecutive anomalies, and decides which early warning signal to output based on a set of preset, tiered triggering rules. The specific early warning logic is divided as follows:

[0169] Level 1 Warning (Single Hammer Strike Anomaly Alert): This is the lowest level of alert, informing the operator that the response to a single hammer strike has deviated from the normal range. After receiving the data packet from the real-time analysis module 3, the warning module 4 parses its judgment result. When the judgment result is abnormal (i.e., the condition of health deviation is met), the warning module will issue a warning. When the condition is met, this level of warning is immediately triggered. Upon triggering, the warning module 4 sends a predefined first-level warning code to the data storage and visualization module 5 via its digital communication interface. This warning code causes the status alarm indicator 401 on the user interface to momentarily turn yellow. Simultaneously, the warning module 4 outputs a square wave signal with a pulse width of, for example, 100 milliseconds through an I / O port, driving a low-decibel buzzer to emit a brief audio prompt. After the action is completed, if the next received judgment result is normal, this warning state is automatically deactivated.

[0170] The second-level warning (persistent pile driving anomaly warning) indicates that the pile driving process is facing a persistent obstacle, requiring operator attention and intervention. The warning module 4 maintains a first-level anomaly counter internally. Whenever the triggering conditions for a Level 1 warning are met, The value is incremented by 1; when the received judgment result is normal, The value is cleared to zero. The value reaches a preset integer threshold. (For example, When a certain condition is detected, this level of warning is triggered. Upon triggering, the warning module 4 sends a second-level warning code to the data storage and visualization module 5, causing the status alarm indicator 401 to display a continuously flashing red color. Simultaneously, the warning module 4 outputs a continuous pulse square wave signal (e.g., at a frequency of 2kHz) through another I / O port, driving a high-decibel buzzer to emit a continuous, high-frequency warning sound. This warning state will continue until a continuous warning signal is received. For example, Only after a normal judgment result is obtained will the status be downgraded or lifted.

[0171] Level 3 Warning (Emergency Alert for Health Risk of Pile Structure): This is the highest level of alert, indicating that the dynamic response mode of the pile has consistently exhibited characteristics that may lead to structural damage, requiring immediate action. The triggering conditions for this level are independent of the first two levels. Warning module 4 maintains a second risk counter internally. The early warning module 4 will receive the health deviation. With an independent, fixed health risk threshold preset based on the mechanical properties of the pile material and structural safety standards. Compare. When the condition is met. hour, Increment the value by 1; otherwise Reset to zero. When The value reaches a preset integer threshold. (For example, When the alarm is triggered, this level of alarm is activated. Upon triggering, the warning module 4 sends the highest priority level 3 warning code to the data storage and visualization module 5, causing a prominent, non-closable red alarm window to pop up on the user interface. Simultaneously, the audible and visual alarm device emits a clearly distinguishable, intermittent emergency alarm sound (e.g., a cycle of 1 second sound followed by a 0.5-second pause). The warning module 4 outputs a stable 24V DC high-level signal through a dedicated hardware I / O port isolated by an optocoupler. This port is directly connected to the emergency stop input of the pile driving equipment's PLC (Programmable Logic Controller). This high-level signal directly triggers the pile driving equipment's emergency stop procedure, immediately stopping the hammering operation to prevent structural damage to the pile. The purpose of using optocoupler isolation is to ensure reliable transmission of the warning signal and avoid electrical interference between the control system and the power system.

[0172] like Figure 6 As shown, Figure 6This is a schematic diagram of the user interface of a data storage and visualization module according to an embodiment of the present invention. This section elaborates on the internal data structure, working mechanism, and human-computer interaction interface implementation of the data storage and visualization module 5. This module performs a dual function in the system: on the one hand, it serves as the system's data persistence center, providing structured data support for all historical analysis, state backtracking, and field self-learning; on the other hand, it serves as the sole window for human-computer interaction, presenting complex background analysis results to the user in an intuitive and quantitative graphical manner.

[0173] In one specific embodiment, the data storage and visualization module 5 includes, in hardware, a non-volatile storage medium (e.g., an industrial-grade solid-state drive SSD) for high-speed data read and write operations, a graphics processing unit (GPU) for processing graphics rendering tasks, and a video output interface connected to a display device.

[0174] Furthermore, to ensure data integrity and consistency, and to support efficient query operations, the data storage and visualization module 5 employs a time-series database and a relational database model to manage all data. The time-series database stores the massive amounts of high-frequency raw and preprocessed waveform data output by module 2. This type of data is characterized by dense timestamps, frequent writes, and queries primarily based on time ranges. Using a time-series database enables extremely high write throughput and efficient time-series querying, aggregation, and downsampling, providing high-performance data support for subsequent refined waveform review and algorithm verification.

[0175] Relational databases are generally used to store structured and semi-structured data. Furthermore, a relational database must contain at least the following normalized tables:

[0176] The single-pile information table stores the static foundation information for each pipe pile. Its fields include: a unique identifier for the pile, a primary key, geographical coordinates, start date, and completion date.

[0177] The hammer impact event table is used to record each valid hammer impact event in chronological order. Its fields include: a unique identifier for the event, a primary key, a foreign key, a link to the single pile information table, a timestamp of the event, and the pile depth at the time the event occurred.

[0178] The feature vector table, which forms a one-to-one relationship with the hammer impact event table, is used to store the quantified features of each hammer impact event. Its fields include: primary key and foreign key, as well as the dynamic time series feature vector.

[0179] The analysis results table is also linked one-to-one with the hammer impact event table, storing the analysis and judgment results for each hammer impact. Its fields include: primary key and foreign key, health deviation degree, dynamic judgment threshold, geological adjustment coefficient, working condition adjustment coefficient, and judgment result code.

[0180] The geological model table is used to store and version control the geological models of the site. Its fields include: a unique identifier for the model version, a primary key, the ID of the pile on which the model was based, the version generation time, and structured data stored in JSON or XML format, describing the soil layer boundary depths and corresponding physical parameters.

[0181] In this way, the constraints of primary keys and foreign keys ensure strong consistency and correlation between data of different dimensions (e.g., the characteristics, results, and depth of a single hammer strike), providing a solid data foundation for subsequent complex data queries, statistical analysis, and field self-learning algorithms.

[0182] Furthermore, this data storage and visualization module can also realize a multi-dimensional visualization interface. The graphics rendering engine within data storage and visualization module 5 generates a user interface by performing real-time queries on the aforementioned database. This interface dynamically presents different views depending on the different stages of the operation.

[0183] During the real-time driving of a single pile, the main display area is the real-time monitoring area, which includes the following components:

[0184] The real-time waveform display window retrieves the latest raw data of the hammer force signal and pile acceleration signal from the database after each hammering event and plots them in a coordinate system with a shared time axis. This side-by-side display allows technicians to intuitively observe the instantaneous correspondence between the hammer force (input) and the pile acceleration (response).

[0185] The key indicator value display area displays the dynamic time-series characteristic vector of the latest hammer impact by querying the analysis results table and the feature vector table. Component values, health deviation and dynamic judgment threshold Displayed in a clear numerical format.

[0186] The health deviation trend chart plots the pile penetration depth on the horizontal axis and the health deviation on the vertical axis. It displays two curves in real time: the first curve shows the health deviation after each hammer blow. The first line represents the state curve formed by the sequence points; the second line represents the corresponding dynamic judgment threshold. The threshold envelope formed by the sequence points. This visualization method, which overlays and compares the state variables with their dynamic judgment benchmarks in the same view, provides operators with a highly intuitive quantitative basis for judging the safety margin of the pile driving process.

[0187] The status alarm indicator is directly driven by the warning code received from the warning module 4, providing users with the most direct risk level prompts.

[0188] After the completion of single-pile construction or when site-level analysis is required, users can switch to the site-level iterative optimization view, which includes:

[0189] The geological model comparison window simultaneously queries and renders both the initial geological model and the latest version of the geological model from the geological model table. By displaying the two geological profiles side-by-side and highlighting the differences, users can clearly see the specific corrections made to the geological model by the system through self-learning.

[0190] The optimal strategy recommendation curve window is activated when construction guidance for the next pile is needed. The module executes a background analysis program that queries all historical data of completed piles within the site, using a specific algorithm (e.g., under certain conditions) to determine the optimal strategy. < Under constraints, find the set of points with the highest penetration efficiency, extract an optimal curve relating hammer energy to depth, and plot it in this window.

[0191] The data storage and visualization module 5 is not an isolated terminal. It engages in bidirectional data communication with the real-time analysis module 3. In the real-time analysis process, the data storage and visualization module 5 passively receives the data packets calculated by the real-time analysis module 3 and performs storage and visualization updates. Simultaneously, when the real-time analysis module 3 needs to calculate geological adjustment coefficients, it proactively initiates a query request to the data storage and visualization module 5 to obtain the latest version of the site geological model data, thus forming a closed-loop workflow where data drives model updates and the model guides real-time analysis.

[0192] Specific embodiments have been used to illustrate the principles and implementation methods of this invention. The descriptions of the embodiments above are only for the purpose of helping to understand the method and core ideas of this invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this invention. Therefore, the content of this specification should not be construed as a limitation of this invention.

Claims

1. A real-time analysis system for the pile driving process using multi-sensor fusion, characterized in that, include: The sensor array is used to collect hammer force signals, pile acceleration signals, pile inclination signals, pile penetration signals, and soil pressure signals around the pile during the pile driving process. The data acquisition and preprocessing module is connected to the sensor group and is used to synchronously acquire and filter the signals obtained by the sensor group during the pile driving process to obtain the processed signals. A real-time analysis module is connected to the data acquisition and preprocessing module, and the real-time analysis module includes: The feature analysis submodule is used to perform dynamic time series feature analysis on the processed hammer impact force signal, pile acceleration signal, pile penetration signal and pile perimeter earth pressure signal, and generate dynamic time series feature vectors. The health diagnosis submodule is used to diagnose the health status of the pile body based on the dynamic time-series feature vector. The dynamic threshold determination submodule is used to determine the dynamic judgment threshold for judging the pile driving status by combining the preset site geological model and real-time pile driving conditions. The comprehensive judgment submodule is used to make a comprehensive judgment on the pile driving status based on the results of the pile health status diagnosis, the dynamic judgment threshold of the pile driving status, and the pile inclination signal. The site geological model includes initial, depth-related soil layer boundary data, and initial geological adjustment coefficients corresponding to each soil layer; the real-time pile driving condition is a real-time updated dataset, including the pile tip depth and hammer energy corresponding to the current hammer blow, as well as the results of the pile health status diagnosis. An early warning module, connected to the real-time analysis module, is used to issue an early warning signal based on the judgment result of the pile driving status. The feature analysis submodule is specifically used for: Within the time window of each hammering event, a dynamic temporal feature vector is constructed that can fully characterize the entire process of excitation, response, and feedback of that hammering event. Specifically, the dynamic time-series feature vector includes: Energy input characteristics used to characterize external energy input; The dynamic response characteristics used to characterize the pile's instantaneous absorption and vibration response to hammer impact energy; Displacement feedback characteristics are used to characterize the efficiency of converting hammering energy into effective penetration work. The resistance response delay and resistance response peak are used to jointly characterize the pile-soil interaction properties.

2. The multi-sensor fusion real-time analysis system for pipe pile driving process according to claim 1, characterized in that, The construction methods for the dynamic temporal feature vectors include: The energy input characteristics are obtained by integrating the hammer force signal within the time window of each hammering event. The dynamic response characteristics are obtained by integrating the pile acceleration signal to obtain the velocity, and taking the peak value of the velocity within the time window. The displacement feedback characteristic is obtained by calculating the rate of change of the pile penetration signal and taking the peak value of the rate of change within the time window; The resistance response delay is obtained by calculating the time difference between the peak time of the pile perimeter soil pressure signal and the peak time of the hammer impact force signal. The peak value of the resistance response is obtained by taking the peak value of the soil pressure signal around the pile within the time window.

3. The multi-sensor fusion real-time analysis system for pipe pile driving process according to claim 1, characterized in that, The diagnosis of the health status of the pile body specifically includes: In the early stage of pile driving, a healthy hammer impact response baseline model is established, and the health deviation between the current dynamic time-series feature vector and the healthy hammer impact response baseline model is calculated in real time.

4. The multi-sensor fusion real-time analysis system for pipe pile driving process according to claim 3, characterized in that, The healthy hammer impact response baseline model is established by statistically averaging the dynamic time-series characteristic vectors of multiple consecutive hammer impacts in the initial stage of pile driving. The health deviation is obtained by calculating the normalized Euclidean distance between the current dynamic time-series feature vector and the health hammer impact response baseline model.

5. The multi-sensor fusion real-time analysis system for pipe pile driving process according to claim 1, characterized in that, The dynamic judgment threshold is determined as follows: Multiply a base threshold by a geological adjustment coefficient determined by matching the current pile tip depth with the geological model of the site area; Multiply by a working condition adjustment coefficient determined by the real-time pile driving stage to obtain the dynamic judgment threshold.

6. The multi-sensor fusion real-time analysis system for pipe pile driving process according to claim 1, characterized in that, The real-time analysis module further includes a self-learning and optimization submodule, which specifically includes: The geological response map generation unit is used to generate a geological response map of the pile location after the single pipe pile driving operation is completed. The online model correction unit is used to correct the geological model of the site area online based on the geological response map.

7. The multi-sensor fusion real-time analysis system for pipe pile driving process according to claim 6, characterized in that, The geological response map generation unit is specifically used for: The dynamic time-series feature vectors generated by the real-time analysis module throughout the entire process of the pipe pile are invoked to form a dynamic time-series feature vector sequence; Simultaneously, the pile tip depth, calculated based on the pile penetration signal and corresponding one-to-one with each dynamic time-series feature vector, is acquired to form a depth sequence corresponding to the dynamic time-series feature vector sequence. An unsupervised clustering algorithm is used to process the dynamic time-series feature vector sequence to identify cluster centers that characterize different dynamic response modes and determine the cluster affiliation of each dynamic time-series feature vector. Based on the clustering of each dynamic time-series feature vector and combined with the depth sequence, hammer impact event sequences that are continuous in depth and belong to the same cluster are merged to determine the data-driven geological response layer boundary.

8. The multi-sensor fusion real-time analysis system for pipe pile driving process according to claim 7, characterized in that, The online model correction unit is specifically used for: The geological response layer boundary determined in the geological response map is taken as the observation boundary depth; By combining the a priori boundary depth in the initial geological model of the site with the observed boundary depth, the corrected posterior boundary depth is calculated to update the soil layer boundary of the geological model of the site. Calculate the mean of the dynamic time series feature vectors falling within each corrected geological layer, and update the geological adjustment coefficient corresponding to the geological layer based on the difference between the mean and the cluster center.

9. The multi-sensor fusion real-time analysis system for pipe pile driving process according to claim 5, characterized in that, The process by which the real-time analysis module determines the operating condition adjustment coefficient specifically includes: First, multiple pile driving conditions are predefined and corresponding condition adjustment coefficient values ​​are preset for each pile driving condition. The multiple pile driving conditions include multiple conventional pile driving conditions and one risk observation condition. The current pile driving condition is determined by a working condition switching rule. The working condition switching rule switches between the conventional pile driving conditions under normal pile driving conditions based on the real-time obtained pile tip depth or penetration rate of change. When a pile health status diagnosis result indicating that there is a risk in the pile body is received, the current pile driving condition is forcibly switched to the risk observation condition. Finally, the preset working condition adjustment coefficient value corresponding to the current pile driving working condition determined by the working condition switching rules is selected.

Images