A TRD wall forming quality intelligent detection system based on multi-source data fusion

Abstract

This invention relates to the field of construction quality inspection technology, specifically a TRD (Tracking, Deposition, and Recognition) intelligent inspection system for wall construction quality based on multi-source data fusion. The system includes: a data acquisition and alignment module, a virtual twin modeling module, a logical deduction and identification module, and a suspicious point location and output module. The data acquisition and alignment module collects time-series data on the cutting box depth and deep soil displacement, aligning them with timestamps to generate a multi-source construction status dataset with a unified time reference. The virtual twin modeling module constructs a virtual twin model containing the three-dimensional spatial trajectory of the cutting box, cutting frequency parameters, and soil stress-strain state. The logical deduction and identification module deduces the continuity of the wall construction and identifies unplanned stagnation or regression anomalies in the vertical direction. The suspicious point location and output module maps the anomaly locations to construction drawings and generates an initial list of quality suspicious points. This system can achieve spatiotemporal fusion of multi-source construction data and real-time identification and precise location of construction anomalies.

Description

Technical Field

[0001] This invention relates to the field of building construction quality inspection technology, and in particular to a TRD intelligent inspection system for wall quality based on multi-source data fusion. Background Technology

[0002] Currently, quality control of TRD (Tracking Deposition and Reduction) wall construction mainly adopts post-construction sampling inspection. During construction, cutting box depth data and deep soil displacement data are collected independently. There is no unified time correlation standard for the two types of monitoring data. Cutting box depth data is only used as an intuitive reference for equipment operation, and deep soil displacement data is only used for site settlement early warning. In this construction scenario, virtual twin technology only simulates single equipment parameters and does not combine the characteristics of TRD construction method to build a model that couples equipment operation and soil response.

[0003] Multi-source monitoring data are in a state of separation, making it impossible to form a correlation between equipment actions and soil deformation. During construction, it is impossible to extrapolate the continuity of the wall through modeling. Unplanned stagnation and retreat behavior of the cutting box in the vertical direction cannot be identified in real time. Abnormal construction sections cannot be mapped to the construction design drawings. The investigation of suspected defective pile sections can only be carried out after construction is completed, and it is impossible to generate a corresponding list of quality doubts at the same time.

[0004] This invention aims to achieve timestamp alignment of multi-source construction data and construction of a corresponding virtual twin model, while also completing the identification of abnormal intervals based on model deduction and the output of a list of quality doubts under drawing mapping. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of existing technologies and propose a TRD-based intelligent detection system for wall quality based on multi-source data fusion.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a TRD-based intelligent detection system for wall lining quality based on multi-source data fusion, comprising:

[0007] The data acquisition and alignment module acquires the cutting box depth time series data output in real time by the TRD method construction machinery during construction, and simultaneously acquires the deep soil displacement time series data output by the inclinometer displacement sensors arranged around the construction site. The cutting box depth time series data and the deep soil displacement time series data are timestamped to generate a multi-source construction status dataset with a unified time reference.

[0008] The virtual twin modeling module constructs a virtual twin model of the TRD construction process based on the multi-source construction state dataset. The virtual twin model includes the three-dimensional spatial trajectory of the cutting box, the cutting frequency parameters of the cutting head, and the stress-strain state of the soil response.

[0009] The logical deduction and identification module uses the virtual twin model to logically deduce the continuity of the TRD wall structure and identify abnormal intervals in which the cutting box stops or retracts unplanned in the vertical direction.

[0010] The suspicious point location output module maps the physical location corresponding to the abnormal interval to the construction design drawings, and generates an initial list of quality suspicious points containing suspected defective pile segments.

[0011] As a further aspect of the present invention, the acquisition of the cutting box depth time-series data output in real time by the TRD method construction machinery during construction includes:

[0012] The hydraulic pressure transmitter installed on the power head of the TRD construction machinery collects the working pressure value of the hydraulic system during the propulsion operation of the power head.

[0013] Simultaneously read the current feedback value of the cutting box traction motor recorded in the electrical control system of the TRD construction machinery;

[0014] The working pressure value of the hydraulic system and the current feedback value of the cutting box traction motor are normalized to eliminate the influence of dimensions caused by differences in equipment models.

[0015] The normalized hydraulic system working pressure value and the current feedback value of the cutting box traction motor are input into the pre-trained working condition discriminator, and the actual travel status label of the cutting box at the current moment is output.

[0016] By combining the actual travel status tag and the position pulse count fed back by the mechanical encoder, the cutting box depth time sequence data is obtained by integral calculation.

[0017] As a further aspect of the present invention, based on the multi-source construction status dataset, a virtual twin model of the TRD method construction process is constructed, including:

[0018] The cutting box depth time series data in the multi-source construction status dataset is analyzed to extract the dwell time characteristics of the cutting box at different strata interfaces;

[0019] Retrieve the pre-set stratum lithology database and match the standard penetration depth of the strata and the shear modulus of the soil layer at the corresponding depth based on the coordinate information of the construction area.

[0020] The dwell time characteristics, standard penetration of the formation, and soil shear modulus are imported into the finite element analysis engine to establish a nonlinear dynamic cutting simulation model that considers tool wear.

[0021] In the nonlinear dynamic cutting simulation model, the displacement time series data of the deep soil is introduced as a boundary constraint condition, and the pressure diffusion cloud map of the soil-mixed grout inside the TRD wall is obtained by solving.

[0022] The three-dimensional spatial trajectory of the cutting box, the cutting frequency parameters of the cutting head, and the pressure diffusion cloud map of the soil-mixed slurry are encapsulated to form the virtual twin model.

[0023] As a further aspect of the present invention, the continuity of the TRD wall structure is logically deduced using the virtual twin model to identify abnormal intervals in which the cutting box experiences unplanned stagnation or retraction in the vertical direction, including:

[0024] In the virtual twin model, the normal travel speed threshold range and normal torque fluctuation range of the cutting box are set;

[0025] The simulated travel speed and simulated torque values ​​output by the virtual twin model are compared in real time with the actual sensor values ​​collected from the construction machinery using the TRD method, and residual calculations are performed.

[0026] When the calculated residual value continuously exceeds the preset fault tolerance range, the cutting box is determined to be in an abnormal operating state.

[0027] Tracing back to the time point when the abnormal operation state occurred, the cutting box depth value in the virtual twin model corresponding to the time point is retrieved;

[0028] If the cutting box depth values ​​of two adjacent time nodes decrease, the segment with the decreased cutting box depth value is marked as the abnormal interval of the unplanned stagnation or regression.

[0029] As a further aspect of the present invention, it also includes:

[0030] The uniformity assessment module is used to evaluate the uniformity of cement-soil mixing in TRD wall construction.

[0031] An acoustic emission sensor is installed at the cutting head of the TRD construction machinery to collect the original acoustic emission waveform signal generated during the cutting and mixing operation;

[0032] Spectral analysis was performed on the original acoustic emission waveform signal to extract the characteristic frequency band energy values ​​that reflect the intensity of cement-soil mixing and crushing.

[0033] The energy value of the characteristic frequency band is dynamically compared with the standard energy curve set by the construction process, and the deviation index of the energy curve is calculated.

[0034] Based on the deviation index of the energy curve, and combined with the pressure diffusion cloud map of the soil-mixed slurry at the corresponding depth in the virtual twin model, it is determined whether there is a mixing blind zone in the cement-soil at the depth range.

[0035] Add the depth information of the stirring blind zone to the initial quality doubt list and mark it as a material segregation risk point.

[0036] As a further aspect of the present invention, spectral analysis is performed on the original acoustic emission waveform signal to extract characteristic frequency band energy values ​​reflecting the intensity of cement-soil mixing and crushing, including:

[0037] The original acoustic emission waveform signal is denoised to filter out low-frequency mechanical vibration interference in the construction environment;

[0038] Wavelet packet decomposition is performed on the original acoustic emission waveform signal after denoising to decompose the signal into different frequency band subspaces;

[0039] Based on the material of the cutting tools of the TRD construction machinery and the properties of the soil, the target frequency band that is strongly correlated with the rock-breaking behavior of the cutting tools is determined.

[0040] The sum of squares of the signal amplitudes of all frequency band subspaces within the target frequency band is calculated to obtain the energy value of the characteristic frequency band.

[0041] As a further aspect of the present invention, it also includes:

[0042] The core sampling verification module is used to verify and correct the initial quality doubt list using core samples.

[0043] At the construction site, a predetermined location in the high-risk section of the initial quality doubt list was selected, and a full-section core sampling operation was carried out to obtain physical core samples.

[0044] The physical core sample was subjected to indoor tests to determine its unconfined compressive strength and permeability coefficient, and a physical performance test report was generated.

[0045] The measured unconfined compressive strength and permeability coefficient in the physical performance test report are compared with the material mechanical parameters predicted by the virtual twin model at the depth segment.

[0046] If the measured unconfined compressive strength is less than 70% of the predicted value, or the permeability coefficient is one order of magnitude higher than the predicted value, then the depth segment is confirmed as a real defect point, and its risk level in the initial quality doubt list is increased.

[0047] If the measured unconfined compressive strength and permeability coefficient are both within the allowable error range, the corresponding depth segment will be removed from the initial list of quality issues, thus completing the list correction.

[0048] As a further aspect of the present invention, it also includes:

[0049] The thickness inversion calculation module is used to invert and calculate the overall wall thickness of TRD wall structures.

[0050] Using the pressure diffusion cloud map of the soil-soil mixture output by the virtual twin model, combined with the groundwater level monitoring data of the construction site, the diffusion radius of the cement-soil slurry in the groundwater body is calculated.

[0051] Based on the wall-forming tool length of the TRD construction machinery and the overlap coefficient specified in the construction process, a geometric calculation model for the wall thickness is established.

[0052] The diffusion radius is used as the boundary disturbance variable in the geometric calculation model. Substitute it into the calculation model to solve for the corrected wall thickness value that takes into account the influence of groundwater seepage.

[0053] The difference between the calculated corrected wall thickness value and the theoretical wall thickness value in the design drawings is calculated to generate a wall thickness deviation distribution map.

[0054] Areas with negative deviations exceeding the allowable threshold in the wall thickness deviation distribution map are added to the initial quality doubt list.

[0055] As a further aspect of the present invention, in the nonlinear dynamic cutting simulation model, the displacement time series data of the deep soil is introduced as a boundary constraint condition, and the pressure diffusion cloud map of the soil-slurry mixture inside the TRD wall is obtained by solving, including:

[0056] The time-domain filtering process is performed on the deep soil displacement time series data to remove abnormal displacement abrupt points caused by construction vibration and environmental noise, and the preprocessed displacement data sequence is obtained.

[0057] A cylindrical coordinate system is established with the three-dimensional spatial trajectory of the cutting box as the central axis, and the preprocessed displacement data sequence is mapped to each grid node under the cylindrical coordinate system.

[0058] In the nonlinear dynamic cutting simulation model, a computational domain is defined, and the displacement values ​​corresponding to each mesh node in the cylindrical coordinate system are used as displacement boundary conditions for the corresponding points on the boundary of the undetermined computational domain and applied to the finite element analysis engine.

[0059] In the nonlinear dynamic cutting simulation model, the balance between the cement-soil slurry injection pressure and the soil shear resistance is dynamically calculated based on the cutting frequency parameters of the cutting head and the real-time depth of the cutting box.

[0060] Using the equilibrium relationship as the internal control equation and the displacement boundary condition as the external boundary constraint, the finite element analysis engine is called to solve the seepage equation of the grout in the soil pores, and the pressure value of each grid node in the computational domain is obtained.

[0061] Spatially interpolate and render the pressure values ​​at all grid nodes to generate a two-dimensional or three-dimensional pressure diffusion cloud map that reflects the distribution of pressure within the wall space.

[0062] As a further aspect of the present invention, the simulated travel speed and simulated torque values ​​output by the virtual twin model are compared in real time with the residual values ​​collected by the actual sensors of the TRD construction machinery to perform residual calculation, including:

[0063] In the virtual twin model, the simulated speed and simulated torque values ​​are calculated in real time based on the input parameters at the current moment.

[0064] The data acquisition and alignment module synchronously acquires the measured values ​​of the actual travel speed sensor and the actual torque sensor from the TRD construction machinery, thus obtaining the numerical pair of measured travel speed and measured torque values.

[0065] The simulated travel speed output by the virtual twin model and the measured travel speed collected by the actual sensor are numerically differentially calculated to obtain the travel speed residual.

[0066] The simulated torque value output by the virtual twin model and the measured torque value collected by the actual sensor are calculated by numerical difference to obtain the torque value residual.

[0067] Calculate the time series statistics of the travel speed residual and the torque value residual, respectively. The statistics include the residual mean, the moving average of the absolute value of the residual, and the variance of the residual change within a preset time window.

[0068] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0069] The cutting box depth time series data and deep soil displacement time series data are timestamped to form a multi-source construction status dataset with a unified time reference. The TRD construction method virtual twin model built on this dataset can fully carry the information of the cutting box three-dimensional spatial trajectory, the cutting frequency parameters of the cutting head, and the stress-strain state of the soil response. The two types of originally separate monitoring data are linked and bound in the spatiotemporal dimension. The correspondence between the equipment operating status and the soil deformation response can be fully presented in the model. Discrete construction data are transformed into integrated construction status representation content. The model can synchronously record the dynamic correlation information between equipment parameter changes and soil mechanical response.

[0070] The virtual twin model can be directly applied to the logical deduction and calculation of the continuity of the TRD wall structure. Unplanned stagnation or regression anomalies in the vertical direction of the cutting box can be directly identified through the deduction process. The anomaly identification process is completed autonomously by the model, without the need for manual verification of construction sequence parameters segment by segment. The physical location corresponding to the anomaly section can be directly mapped to the construction design drawings. The relevant information of suspected defective pile sections can be integrated to form an initial list of quality doubts. The generation process of quality doubts is synchronized with the on-site construction progress. The location output of abnormal construction sections does not need to wait for the later inspection process after construction is completed. The identification and location of abnormal working conditions form a coherent execution process, and the generation of the doubt list does not require additional manual processing. Attached Figure Description

[0071] Figure 1 This is a timing diagram of a TRD intelligent detection system for wall quality based on multi-source data fusion, as described in this invention.

[0072] Figure 2 A flowchart for constructing a virtual twin model of the TRD construction process;

[0073] Figure 3 A time-series variation and abnormal interval identification diagram of the TRD construction cutting box depth;

[0074] Figure 4 The curves show the energy and deviation of acoustic emission in the characteristic frequency bands.

[0075] Figure 5 This is a verification diagram showing the depth distribution of the permeability coefficient of the TRD wall and the defect section. Detailed Implementation

[0076] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0077] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0078] See Figure 1The data acquisition and alignment module is responsible for acquiring real-time cutting box depth data output by the TRD construction machinery during construction, and simultaneously acquiring deep soil displacement time-series data output by inclinometer displacement sensors around the construction site. The two types of data are timestamped to generate a multi-source construction status dataset with a unified time reference. Based on this multi-source construction status dataset, the virtual twin modeling module constructs a virtual twin model of the TRD construction process, including the cutting box's three-dimensional spatial trajectory, cutting frequency parameters of the cutting head, and the soil's stress-strain response. The logical deduction and identification module uses this virtual twin model to logically deduce the continuity of the TRD wall structure, identifying abnormal intervals where the cutting box experiences unplanned stagnation or retreat in the vertical direction. The suspicious point location and output module maps the physical locations corresponding to the identified abnormal intervals to the construction design drawings, generating an initial list of quality suspicious points containing suspected defective pile segments.

[0079] In one embodiment of the present invention, the TRD wall-forming quality intelligent detection system based on multi-source data fusion acquires cutting box depth time-series data through a data acquisition and alignment module. During implementation, a hydraulic pressure transmitter installed on the power head of the TRD construction machinery acquires the hydraulic system working pressure value during the propulsion operation. Simultaneously, it reads the cutting box traction motor current feedback value recorded in the TRD construction machinery's electrical control system. The hydraulic system working pressure value and the cutting box traction motor current feedback value are synchronously acquired and stored at the same time interval. In some embodiments, the hydraulic system working pressure value output by different models of TRD construction machinery may range from 0 to 30 MPa, and the cutting box traction motor current feedback value may range from 0 to 200 amperes. The data acquisition and alignment module directly receives these raw signals. These raw signals have different dimensions due to differences in equipment models. For example, the full-scale hydraulic system working pressure value of model A equipment is 25 MPa, while that of model B equipment is 30 MPa. Similar differences exist in the cutting box traction motor current feedback value, which introduces bias when directly comparing or processing the raw data. In practical implementation, the working pressure value of the hydraulic system and the current feedback value of the cutting box traction motor are normalized to eliminate the influence of dimensions. The normalization process maps the original values ​​to a unified dimensionless interval. The normalization formula is expressed as:

[0080]

[0081] in: This represents the normalized working pressure value of the hydraulic system or the current feedback value of the cutting box traction motor. This indicates the original hydraulic system working pressure value or the original cutting box traction motor current feedback value. This indicates the minimum range of the corresponding sensor. This indicates the maximum range of the corresponding sensor. The minimum and maximum ranges are obtained from the sensor calibration parameters or the equipment nameplate data.

[0082] It is understandable that after normalization, the hydraulic system working pressure values ​​and the cutting box traction motor current feedback values ​​of different equipment models are converted into values ​​between 0 and 1. Data comparison shows that the original hydraulic system working pressure values ​​of model A and model B are 20 MPa and 25.5 MPa respectively under the same working conditions. After normalization, they are approximately 0.8 and 0.85 respectively, indicating that normalization eliminates the difference in dimensions and makes the data comparable. In some embodiments, the normalized hydraulic system working pressure value and the cutting box traction motor current feedback value are input to a pre-trained working condition discriminator. The working condition discriminator is generated based on historical data, receives the normalized hydraulic system working pressure value and the cutting box traction motor current feedback value as input features, and outputs the actual travel status label of the cutting box at the current moment. The actual travel status label includes forward, stationary, or backward. After the working condition discriminator outputs the actual travel status label, the system records the label and the corresponding timestamp. Optionally, the working condition discriminator can be implemented using a support vector machine or neural network model. During model training, the labeled historical hydraulic system working pressure value, the current feedback value of the cutting box traction motor, and the corresponding actual travel state of the cutting box are used as the training set. After training, the model can infer the actual travel state of the cutting box based on real-time input.

[0083] In practical implementation, the cutting box depth timing data is obtained by integrating the position pulse counts fed back by the actual travel status label and the mechanical encoder. The mechanical encoder is installed on the transmission component of the TRD construction machinery and outputs position pulse counts proportional to the cutting box displacement. The depth change is obtained by integrating the cumulative position pulse counts and multiplying them by the pulse equivalent. The pulse equivalent represents the vertical displacement of the cutting box corresponding to a single position pulse count. For example, the pulse equivalent is 0.1 mm per pulse. The integration calculation starts from the initial depth and is expressed by the formula: Depth = Initial Depth + ∑(Position Pulse Count × Pulse Equivalent). The initial depth is set to zero or a known elevation at the start of construction. It can be understood that the actual travel status label is used for verification and correction during the integration calculation. For example, when the actual travel status label indicates reversal, the position pulse counts are accumulated as negative values ​​in the integration calculation. Data comparison shows that using only the position pulse count integration may produce cumulative errors due to encoder slippage. However, combining the actual travel status label can identify abnormal reversal intervals and correct the depth value. For example, when the actual travel status label indicates a standstill, the depth timing data should remain stable. If the position pulse count integration shows a depth change, the system triggers the correction logic. Optionally, the integral calculation is performed at fixed time intervals, such as calculating the cutting box depth value every 0.1 seconds, generating a continuous cutting box depth time series data sequence. The cutting box depth time series data sequence is stored as a time-depth correspondence table for the virtual twin modeling module to call.

[0084] In one embodiment of the present invention, see [reference] Figure 2 The virtual twin modeling module constructs a virtual twin model by parsing a multi-source construction status dataset. The module reads the cutting box depth time-series data generated by the data acquisition and alignment module and extracts the dwell time characteristics of the cutting box at different strata interfaces. The dwell time characteristics are obtained by analyzing the curve of the cutting box depth changing over time. When the curve slope is close to zero and the duration exceeds a preset threshold, the system determines that the cutting box is in a dwell state and records the start and end times of that period. The dwell time is the difference between the start and end times. The system records all identified dwell periods and their corresponding depth information as a dwell time feature set. In some embodiments, the extracted dwell time feature data shows that the cutting box stayed for 120 seconds at a depth of 15.3 meters and for 95 seconds at a depth of 22.1 meters. These data are related to the location of the interface between the silty clay layer and the sand layer revealed in the geological exploration report.

[0085] In practical implementation, the system retrieves a pre-set stratigraphic lithology database, which stores stratigraphic information corresponding to different geographical coordinates. Based on the coordinates of the construction area, the system matches the standard penetration depth (SPT) and soil shear modulus at the corresponding depth. In practical implementation, the extracted dwell time features, the matched SPT and soil shear modulus are imported into the finite element analysis engine to establish a nonlinear dynamic cutting simulation model considering tool wear. The tool wear coefficient is obtained by looking up tables based on the cumulative cutting distance of the cutting box and the tool material, and is used as a state variable in the model calculation. In practical implementation, the nonlinear dynamic cutting simulation model needs to incorporate deep soil displacement time-series data as boundary constraints. The deep soil displacement time-series data undergoes time-domain filtering, using a low-pass filter to remove abnormal displacement abrupt points caused by high-frequency construction vibrations and environmental noise, resulting in a pre-processed displacement data sequence. This displacement data sequence includes timestamps and horizontal displacement values ​​at different depths of the corresponding inclinometer boreholes.

[0086] In the specific implementation, a cylindrical coordinate system is established with the three-dimensional spatial trajectory of the cutting box as the central axis. The Z-axis of the cylindrical coordinate system coincides with the vertical centerline of the cutting box, the radial direction is R, and the circumferential direction is θ. The preprocessed displacement data sequence is mapped to each grid node in the cylindrical coordinate system. The mapping process involves coordinate transformation based on the horizontal relative position of the inclinometer hole and the cutting box. In the specific implementation, a computational domain is defined in the nonlinear dynamic cutting simulation model, and the displacement values ​​corresponding to each grid node in the cylindrical coordinate system are used as displacement boundary conditions for the corresponding positions on the boundary of the undetermined computational domain. These displacement boundary conditions are defined on the outer boundary of the model in the form of nodal loads. In the specific implementation, in the nonlinear dynamic cutting simulation model, the balance relationship between the cement-soil slurry injection pressure and the soil shear resistance is dynamically calculated based on the cutting frequency parameters of the cutting head and the real-time depth of the cutting box. This balance relationship is described by a set of coupled equations. In practice, the equilibrium relationship is used as the internal control equation and the displacement boundary condition is used as the external boundary constraint. The finite element analysis engine is called to solve the seepage equation of the grout in the soil pores. The seepage equation is established based on Darcy's law and the law of conservation of mass. After solving, the pressure value of each grid node in the computational domain is obtained.

[0087] In some embodiments, the finite element analysis engine outputs grid node pressure values ​​after solving the problem. Data comparison shows that the node pressure value near the cutting box wall at a depth of 15.3 meters is 150 kPa, while the node pressure value at the same depth, 1 meter away from the cutting box, decreases to 50 kPa. In specific implementations, the pressure values ​​at all grid nodes are spatially interpolated and rendered, and different pressure values ​​are mapped using color gradients to generate a two-dimensional or three-dimensional pressure diffusion cloud map reflecting the pressure distribution within the wall structure. The cloud map is output as an image or animation. Optionally, the pressure diffusion cloud map can be overlaid on the three-dimensional model of the construction design drawings for visualization. It can be understood that the pressure diffusion cloud map reflects the penetration and pressure transmission effect of cement slurry in the soil. Areas with gentle pressure gradients in the cloud map indicate uniform slurry diffusion, while areas with steep pressure gradients may indicate impeded penetration or the presence of weak zones. In practice, the three-dimensional spatial trajectory of the cutting box, the cutting frequency parameters of the cutting head, and the pressure diffusion cloud map of the soil-mixed slurry are encapsulated to form a virtual twin model data package containing geometric, state, and field information, which is then called by the logic deduction and recognition module.

[0088] In one embodiment of the present invention, the logic deduction and identification module uses a virtual twin model to logically deduce the continuity of the TRD wall structure. The virtual twin model sets a normal travel speed threshold range and a normal torque fluctuation range for the cutting box. The normal travel speed threshold range is set according to the construction process specifications, for example, a lower limit of 0.02 meters per minute and an upper limit of 0.08 meters per minute. The normal torque fluctuation range is determined based on the rated torque and load characteristics of the power head, for example, a lower limit of 30% of the rated torque and an upper limit of 85% of the rated torque. In specific implementation, the simulated values ​​output by the virtual twin model are compared with the actual sensor values ​​in real time, and residual calculations are performed. The virtual twin model calculates in real time based on the input parameters at the current moment to obtain numerical pairs of simulated travel speed and simulated torque values. Simultaneously, the data acquisition and alignment module synchronously acquires the measured values ​​of the actual travel speed sensor and the actual torque sensor from the TRD construction machinery to obtain numerical pairs of measured travel speed and measured torque values.

[0089] In practical implementation, the simulated travel speed output by the virtual twin model and the measured travel speed collected by the actual sensors are numerically differentiated to obtain the travel speed residual. Similarly, the simulated torque value output by the virtual twin model and the measured torque value collected by the actual sensors are numerically differentiated to obtain the torque value residual. The formula for calculating the residual is as follows:

[0090]

[0091] in: This represents the residual speed or torque value. This represents the simulated speed or simulated torque value output by the virtual twin model. This represents the actual measured travel speed or torque value collected by the sensor. It can be understood that the travel speed residual and torque residual are calculated separately, resulting in two independent time series. In specific implementations, time series statistics for the travel speed residual and torque residual are calculated separately. These statistics include the mean of the residuals, the moving average of the absolute values ​​of the residuals, and the variance of the residual changes within a preset time window. The preset time window length is, for example, set to 10 seconds, and the system calculates the statistics in 1-second increments. In some embodiments, data comparison shows that within five consecutive calculation steps, the mean of the travel speed residual exceeds 0.03 meters per minute, the moving average of the absolute values ​​of the residuals consistently exceeds 0.02 meters per minute, and the variance of the residual changes increases.

[0092] In practical implementation, the system traces back to the time point where the abnormal operation state occurred, retrieves the cutting box depth value in the virtual twin model for the corresponding time point, and extracts the cutting box depth value from the 3D spatial trajectory data output by the virtual twin model. Optionally, the system records the start and end timestamps of the abnormal operation state. In practical implementation, if the cutting box depth value decreases between two adjacent time points, the segment with the decreased cutting box depth value is marked as an abnormal interval of unplanned stagnation or regression. For example, if the cutting box depth value recorded in the virtual twin model at time point T1 is 15.30 meters, and at the next time point T2 the depth value is 15.25 meters, the depth value decreases by 0.05 meters. The system then marks the segment with a depth from 15.30 meters to 15.25 meters as an abnormal interval. It can be understood that a decrease in the cutting box depth value directly indicates that the cutting box has regressed, while a depth value remaining unchanged for a certain period of time may indicate stagnation. Optionally, the information of the marked abnormal interval includes the start depth, end depth, start time, and end time. In practice, all marked abnormal intervals are compiled into a list and output to the suspicious point location output module for further processing.

[0093] See Figure 3 In the time-series analysis of cutting box depth during TRD construction, the identification of abnormal intervals relies on the morphological characteristics of the depth time-series curve and the deductive logic of the virtual twin model. Specifically, the original cutting box depth time-series data is represented as a one-dimensional time series curve consisting of time and depth, reflecting the real-time movement of the cutting box in the vertical direction. Abnormal intervals are determined jointly through a combination of local morphological changes in the curve and residual analysis of the virtual twin model: the first-order difference calculation is performed on the depth time-series curve to extract the depth change rate at adjacent time nodes. When the change rate is continuously negative or approaches 0 and continuously exceeds a preset duration threshold, it is marked as a potential abnormal segment. The virtual twin model performs the same operation simultaneously, using the residual statistics (such as residual mean, moving average, and variance) between the simulated depth curve and the measured curve as the anomaly confidence criterion. When the residual statistics continuously exceed the tolerance range, the time period is confirmed as an abnormal interval of unplanned stagnation or regression. The shaded areas in the figure represent two abnormal intervals identified by the system, corresponding to the periods when the cutting box depth curve showed local regression and stagnation, respectively. This abnormal interval information will serve as the core basis for the continuity defects of the wall and will be entered into the initial list of quality doubts for subsequent location and verification.

[0094] In one embodiment of the invention, a uniformity assessment module evaluates the uniformity of cement-soil mixing in TRD-based wall construction. An acoustic emission sensor is installed at the cutting head of the TRD construction machinery. The sensor is fixed to the cutting head near the cutting surface via a magnetic base or threaded interface and is used to collect the original acoustic emission waveform signal generated during the cutting and mixing operation. This original acoustic emission waveform signal is a voltage sequence that varies over time. The signal acquisition frequency is set to 1 MHz to capture high-frequency particle breakage and friction signals. In a specific implementation, spectral analysis is performed on the original acoustic emission waveform signal to extract the characteristic frequency band energy values ​​reflecting the intensity of cement-soil mixing and breakage. First, the original acoustic emission waveform signal is denoised to filter out low-frequency mechanical vibration interference from the construction environment. A high-pass digital filter with a cutoff frequency of 5 kHz is used to filter the original waveform signal, removing low-frequency vibration noise mainly generated by the engine and hydraulic pump, resulting in a denoised acoustic emission waveform signal. In some embodiments, a 1-second acoustic emission waveform signal is acquired and, after denoising, its amplitude fluctuates in the range of -2.5 volts to +2.5 volts, while the original signal is mixed with low-frequency interference signals with an amplitude of up to 5 volts.

[0095] In the specific implementation, wavelet packet decomposition is performed on the original acoustic emission waveform signal after denoising, decomposing the signal into different frequency band subspaces. The 'db4' wavelet basis function is selected for wavelet packet decomposition, with 5 decomposition levels, resulting in 32 frequency band subspace signals from low to high frequencies, each corresponding to a frequency range. In the specific implementation, based on the material of the cutting tools and the soil properties of the TRD method construction machinery, the target frequency band strongly correlated with the cutting tool's rock-breaking behavior is determined. The cutting tool material is cemented carbide, and the soil is gravelly sand. Historical calibration data shows that the effective cutting and breaking signals are mainly concentrated in the 80 kHz to 220 kHz frequency band. Therefore, the frequency band subspace covering this frequency range after wavelet packet decomposition is determined as the target frequency band strongly correlated with the cutting tool's rock-breaking behavior. In the specific implementation, the sum of the squares of the amplitudes of all frequency band subspace signals within the target frequency band is calculated to obtain the characteristic frequency band energy value. The calculation formula is expressed as:

[0096]

[0097] in: This represents the calculated energy value of the characteristic frequency band. This indicates the total number of frequency band subspaces included in the target frequency band. Indicates the first frequency band within the target frequency band The frequency band subspace signal at discrete time points amplitude, This represents the total number of signal sampling points. It can be understood as the energy value of the characteristic frequency band. This is a scalar value whose magnitude reflects the total energy of the acoustic emission signal carried by the target frequency band within the analysis period. This energy is related to the severity of the cutting and breaking. In practice, the characteristic frequency band energy value is dynamically compared with the standard energy curve set by the construction process. The standard energy curve is a baseline curve generated based on the acoustic emission data collected from successfully constructed sections in uniform soil layers. The ratio or difference between the characteristic frequency band energy value and the expected energy value of the standard energy curve at each analysis time point is dynamically calculated. In practice, the deviation index of the energy curve is calculated. The deviation index is defined as the normalized root mean square error between the characteristic frequency band energy value sequence and the standard energy curve sequence. The larger the value, the greater the difference between the current mixing state and the standard uniform mixing state. In some embodiments, refer to Table 1 for the characteristic frequency band energy value, standard energy curve value, and calculated deviation index.

[0098] Table 1: Comparison of Energy Values ​​in Characteristic Frequency Bands and Standard Energy Curves

[0099] 1.0 1250 1200 0.040 2.0 980 1150 0.155 3.0 1150 1180 0.025 4.0 750 1160 0.353

[0100] In practical implementation, based on the deviation index of the energy curve and combined with the pressure diffusion cloud map of the soil-soil mixture at the corresponding depth in the virtual twin model, it is determined whether there is a mixing blind zone in the cement-soil mixture at that depth. The pressure diffusion cloud map reflects the spatial uniformity of the slurry pressure distribution. It can be understood that when the deviation index of a certain depth segment is consistently high, and the virtual twin model shows a local low-pressure area or an abnormal pressure gradient area in the pressure diffusion cloud map of that depth segment, the system determines that there is a mixing blind zone at that depth segment, and the cement-soil mixture may be uneven. In practical implementation, the information of depth segments with mixing blind zones is added to the initial quality suspicion list. The depth segment information includes the starting depth, ending depth, and the judgment criteria. The judgment criteria are recorded as "acoustic emission energy deviation exceeds the standard and the pressure cloud map shows abnormalities," and are marked as material segregation risk points in the initial quality suspicion list. Optionally, material segregation risk points will be highlighted in the list with different colors or icons. Optionally, the uniformity assessment module continuously evaluates the entire cutting and mixing process, using each 0.5-meter drilling depth as an analysis unit.

[0101] See Figure 4In the TRD wall-forming cement-soil mixing uniformity assessment module, the deviation analysis of the acoustic emission characteristic frequency band energy from the standard energy curve is the core basis for identifying mixing blind zones. Specifically, the figure uses the characteristic frequency band energy value (solid dotted line), the standard energy curve value (dashed box line), and the deviation index (solid triangular line) as the core observation objects, while introducing a deviation threshold (0.3) (dashed line) as an anomaly judgment benchmark: At the 1.0 second analysis point, the characteristic frequency band energy value (1250 volts²) and the standard energy curve value (1200 volts²) are highly consistent, and the deviation index is only 0.040, far below the threshold, indicating that the current mixing state is consistent with the standard uniform working condition. At the 2.0 second analysis point, the characteristic frequency band energy value (980 volts²) is significantly lower than the standard energy curve value (1150 volts²), and the deviation index rises to 0.155, indicating that the mixing and breaking strength has undergone a phased decay, but is still within a controllable range. At the 3.0-second analysis point, the characteristic frequency band energy value (1150 volts²) approached the standard energy curve value (1180 volts²) again, and the deviation index dropped back to 0.025, indicating that the mixing state briefly returned to normal. At the 4.0-second analysis point, the characteristic frequency band energy value plummeted to 750 volts², showing a significant deviation from the standard energy curve value (1160 volts²), and the deviation index climbed to 0.353, exceeding the preset threshold of 0.3. Combined with the slurry pressure diffusion cloud map of this depth segment in the virtual twin model, it can be determined that there is a mixing blind zone here, and the cement-soil mixing uniformity has significant defects, which needs to be marked as a high-risk point for material segregation. The figure intuitively presents the fluctuation law of the characteristic frequency band energy over time, the degree of deviation from the standard working condition, and the abnormal triggering logic, providing a quantitative and visual basis for continuous monitoring and suspicious point location of TRD wall quality.

[0102] In one embodiment of the present invention, the core sampling verification module is used to verify and correct the initial quality doubt list using core samples. At the construction site, a predetermined location in the high-risk section of the initial quality doubt list is selected for full-section core sampling. The initial quality doubt list marks an abnormal interval with a depth of 15.0 meters to 15.5 meters as a high-risk section. Based on this, the system determines a core sampling point at the corresponding pile axis position. A core drilling rig is used to perform full-section core sampling at this point to obtain continuous physical core samples from the ground to the design depth. The physical core samples are numbered according to the core sampling sequence and sealed for preservation. In practice, physical core samples undergo indoor testing to determine their unconfined compressive strength and permeability coefficient, generating a physical performance test report. After curing the physical core samples under standard conditions for 28 days, they are processed into standard cylindrical specimens. The unconfined compressive strength is determined using a pressure testing machine, and the permeability coefficient is determined using a variable head permeability test. For example, specimens prepared from physical core samples corresponding to high-risk sections show an unconfined compressive strength of 1.2 MPa and a permeability coefficient of 5.0 × 10⁻⁻⁻⁶. 7Centimeters per second—these measured values ​​were recorded in the physical performance test report.

[0103] In practice, the measured unconfined compressive strength and permeability coefficient in the physical performance test report are compared with the material mechanical parameters predicted by the virtual twin model at the corresponding depth. Based on construction parameters and geological data, the virtual twin model predicts that the unconfined compressive strength of the cement-soil at this depth is 1.8 MPa and the permeability coefficient is 3.0 × 10⁻⁻⁻⁶. 8 Centimeters per second. In specific implementations, if the measured unconfined compressive strength is less than 70% of the predicted value, or the permeability coefficient is one order of magnitude higher than the predicted value, the corresponding depth segment is confirmed as a real defect point, and its risk level in the initial quality suspicion list is increased. In some embodiments, the measured unconfined compressive strength of 1.2 MPa is less than 70% of the predicted value of 1.8 MPa (i.e., 1.26 MPa), and the measured permeability coefficient is 5.0 × 10⁻⁻⁻⁶. 7 Centimeters per second higher than predicted by 3.0 × 10⁻ 8 The rate of change is on the order of centimeters per second. Since both conditions are met, the system confirms that the depth segment is a genuine defect point and upgrades its risk level in the initial quality suspect list from "high risk" to "confirmed defect." In practice, if the measured unconfined compressive strength and permeability coefficient are both within the allowable error range, the corresponding depth segment is removed from the initial quality suspect list, thus correcting the list. The allowable error range is, for example, set at ±20% of the predicted value.

[0104] In practical implementation, the thickness inversion and estimation module inverts and estimates the overall wall thickness of the TRD wall. Using the pressure diffusion cloud map of the soil-soil mixture output by the virtual twin model, combined with the groundwater level monitoring data of the construction site, the diffusion radius of the cement-soil grout in the groundwater is calculated. The groundwater level monitoring data comes from the water level observation wells arranged in the site. The pressure diffusion cloud map provides the grout pressure distribution. Combined with Darcy's law and the groundwater seepage velocity, the diffusion distance of the grout front can be estimated.

[0105] In practical implementation, the diffusion radius is used as a boundary disturbance variable in the geometric calculation model. Substituted into the calculation model, the corrected wall thickness value considering the influence of groundwater seepage is obtained. The revised calculation formula is expressed as follows:

[0106]

[0107] in: This indicates the corrected wall thickness value that takes into account the impact of groundwater seepage. This indicates the length of the wall-forming cutter in the TRD construction machinery. This indicates the overlap coefficient specified in the construction process. This indicates the reduction in effective wall thickness due to groundwater seepage. With diffusion radius and groundwater influence coefficient Related, calculated as Groundwater influence coefficient Determined based on the permeability of the formation.

[0108] See Figure 5 In the logarithmic coordinate visualization of the depth distribution of the permeability coefficient in TRD wall construction quality, the core is to accurately identify and verify the quality of defective sections by comparing the measured data from borehole core sampling with the judgment threshold. Specifically, the graph uses depth (meters) as the vertical axis (downward is positive) and the logarithm of the permeability coefficient as the horizontal axis. 10 The graph uses (cm / s) as the horizontal axis to visually represent the variation of measured permeability coefficients from the surface to a depth of 30 meters. The measured permeability coefficients in the shallow (0–15m) and deep (15.5–30m) regions are generally distributed within the logarithmic range of -7.6 to -7.4, with stable values ​​and significantly lower than the judgment threshold (+1 order of magnitude) marked by the dashed line on the right, indicating that the wall quality in this range meets the seepage prevention requirements. The 15.0–15.5m depth segment is marked as a shaded anomaly segment. The measured permeability coefficient in this segment shifts significantly towards the judgment threshold, exceeding the predicted value by an order of magnitude compared to the virtual twin model. Simultaneously, the unconfined compressive strength of the corresponding core sample is lower than 70% of the predicted value. These two indicators jointly verify that this depth segment is a confirmed defect point, posing a quality risk of insufficient seepage prevention performance and material strength degradation. The use of logarithmic coordinates effectively amplifies the order-of-magnitude differences in permeability coefficients, making even minor permeability variations clearly visible in the graph, providing a visual basis for delineating the boundaries of defective segments and assessing risk levels.

[0109] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A TRD wall forming quality intelligent detection system based on multi-source data fusion, characterized in that, The system includes: The data acquisition and alignment module acquires the cutting box depth time series data output in real time by the TRD method construction machinery during construction, and simultaneously acquires the deep soil displacement time series data output by the inclinometer displacement sensors arranged around the construction site. The cutting box depth time series data and the deep soil displacement time series data are timestamped to generate a multi-source construction status dataset with a unified time reference. The virtual twin modeling module constructs a virtual twin model of the TRD construction process based on the multi-source construction state dataset. The virtual twin model includes the three-dimensional spatial trajectory of the cutting box, the cutting frequency parameters of the cutting head, and the stress-strain state of the soil response. The construction of a virtual twin model of the TRD construction process based on the multi-source construction status dataset includes: The cutting box depth time series data in the multi-source construction status dataset is analyzed to extract the dwell time characteristics of the cutting box at different strata interfaces; Retrieve the pre-set stratum lithology database and match the standard penetration depth of the strata and the shear modulus of the soil layer at the corresponding depth based on the coordinate information of the construction area. The dwell time characteristics, standard penetration of the formation, and soil shear modulus are imported into the finite element analysis engine to establish a nonlinear dynamic cutting simulation model that considers tool wear. In the nonlinear dynamic cutting simulation model, the displacement time series data of the deep soil is introduced as a boundary constraint condition, and the pressure diffusion cloud map of the soil-mixed grout inside the TRD wall is obtained by solving. The three-dimensional spatial trajectory of the cutting box, the cutting frequency parameters of the cutting head, and the pressure diffusion cloud map of the soil-mixed slurry are encapsulated to form the virtual twin model; The logical deduction and identification module uses the virtual twin model to logically deduce the continuity of the TRD wall structure, identifying abnormal intervals where the cutting box experiences unplanned stagnation or retraction in the vertical direction, including: In the virtual twin model, the normal travel speed threshold range and normal torque fluctuation range of the cutting box are set; The simulated travel speed and simulated torque values ​​output by the virtual twin model are compared in real time with the actual sensor values ​​collected from the construction machinery using the TRD method, and residual calculations are performed. When the calculated residual value continuously exceeds the preset fault tolerance range, the cutting box is determined to be in an abnormal operating state. Tracing back to the time point when the abnormal operating state occurred, the cutting box depth value of the corresponding time point in the virtual twin model is retrieved; If the cutting box depth value decreases between two adjacent time nodes, the segment where the cutting box depth value decreases is marked as the abnormal interval of the unplanned stagnation or rollback. The suspicious point location output module maps the physical location corresponding to the abnormal interval to the construction design drawings, and generates an initial list of quality suspicious points containing suspected defective pile segments.

2. The TRD wall quality intelligent detection system based on multi-source data fusion according to claim 1, characterized in that, The acquisition of real-time cutting box depth time-series data output by the TRD method construction machinery during construction includes: The hydraulic pressure transmitter installed on the power head of the TRD construction machinery collects the working pressure value of the hydraulic system during the propulsion operation of the power head. Simultaneously read the current feedback value of the cutting box traction motor recorded in the electrical control system of the TRD construction machinery; The working pressure value of the hydraulic system and the current feedback value of the cutting box traction motor are normalized to eliminate the influence of dimensions caused by differences in equipment models. The normalized hydraulic system working pressure value and the current feedback value of the cutting box traction motor are input into the pre-trained working condition discriminator, and the actual travel status label of the cutting box at the current moment is output. By combining the actual travel status tag and the position pulse count fed back by the mechanical encoder, the cutting box depth time sequence data is obtained by integral calculation.

3. The TRD wall quality intelligent detection system based on multi-source data fusion according to claim 2, characterized in that, Also includes: The uniformity assessment module is used to assess the uniformity of cement-soil mixing in TRD wall construction. An acoustic emission sensor is installed at the cutting head of the TRD construction machinery to collect the original acoustic emission waveform signal generated during the cutting and mixing operation; Spectral analysis was performed on the original acoustic emission waveform signal to extract the characteristic frequency band energy values ​​that reflect the intensity of cement-soil mixing and crushing. The energy value of the characteristic frequency band is dynamically compared with the standard energy curve set by the construction process, and the deviation index of the energy curve is calculated. Based on the deviation index of the energy curve, and combined with the pressure diffusion cloud map of the soil-mixed slurry at the corresponding depth in the virtual twin model, it is determined whether there is a mixing blind zone in the cement-soil at the depth range. Add the depth information of the stirring blind zone to the initial quality doubt list and mark it as a material segregation risk point.

4. The TRD walling quality intelligent detection system based on multi-source data fusion according to claim 3, characterized in that, Spectral analysis was performed on the original acoustic emission waveform signal to extract the characteristic frequency band energy values ​​reflecting the mixing and crushing intensity of cement-soil, including: The original acoustic emission waveform signal is denoised to filter out low-frequency mechanical vibration interference in the construction environment; Wavelet packet decomposition is performed on the original acoustic emission waveform signal after denoising to decompose the signal into different frequency band subspaces; Based on the material of the cutting tools of the TRD construction machinery and the properties of the soil, the target frequency band that is strongly correlated with the rock-breaking behavior of the cutting tools is determined. The sum of squares of the signal amplitudes of all frequency band subspaces within the target frequency band is calculated to obtain the energy value of the characteristic frequency band.

5. The TRD walling quality intelligent detection system based on multi-source data fusion according to claim 4, characterized in that, Also includes: The core sampling verification module is used to verify and correct the initial list of quality issues based on the core sampling samples. At the construction site, a predetermined location in the high-risk section of the initial quality doubt list was selected, and a full-section core sampling operation was carried out to obtain physical core samples. The physical core sample was subjected to indoor tests to determine its unconfined compressive strength and permeability coefficient, and a physical performance test report was generated. The measured unconfined compressive strength and permeability coefficient in the physical performance test report are compared with the material mechanical parameters predicted by the virtual twin model at the depth segment. If the measured unconfined compressive strength is less than 70% of the predicted value, or the permeability coefficient is one order of magnitude higher than the predicted value, then the depth segment is confirmed as a real defect point, and its risk level in the initial quality doubt list is increased. If the measured unconfined compressive strength and permeability coefficient are both within the allowable error range, the corresponding depth segment will be removed from the initial list of quality issues, thus completing the list correction. 6.The TRD wall quality intelligent detection system based on multi-source data fusion according to claim 5, characterized in that, Also includes: The thickness inversion calculation module is used to invert and calculate the overall wall thickness of TRD wall structures. Using the pressure diffusion cloud map of the soil-soil mixture output by the virtual twin model, combined with the groundwater level monitoring data of the construction site, the diffusion radius of the cement-soil slurry in the groundwater body is calculated. Based on the wall-forming tool length of the TRD construction machinery and the overlap coefficient specified in the construction process, a geometric calculation model for the wall thickness is established. The diffusion radius is used as the boundary disturbance variable in the geometric calculation model. Substitute it into the calculation model to solve for the corrected wall thickness value that takes into account the influence of groundwater seepage. The difference between the calculated corrected wall thickness value and the theoretical wall thickness value in the design drawings is calculated to generate a wall thickness deviation distribution map. Areas with negative deviations exceeding the allowable threshold in the wall thickness deviation distribution map are added to the initial quality doubt list.

7. The TRD walling quality intelligent detection system based on multi-source data fusion according to claim 6, characterized in that, In the nonlinear dynamic cutting simulation model, the displacement time series data of the deep soil is introduced as a boundary constraint condition, and the pressure diffusion cloud map of the soil-grout mixture inside the TRD wall is obtained by solving, including: The time-domain filtering process is performed on the deep soil displacement time series data to remove abnormal displacement abrupt points caused by construction vibration and environmental noise, and the preprocessed displacement data sequence is obtained. A cylindrical coordinate system is established with the three-dimensional spatial trajectory of the cutting box as the central axis, and the preprocessed displacement data sequence is mapped to each grid node under the cylindrical coordinate system. In the nonlinear dynamic cutting simulation model, a computational domain is defined, and the displacement values ​​corresponding to each mesh node in the cylindrical coordinate system are used as displacement boundary conditions for the corresponding points on the boundary of the undetermined computational domain and applied to the finite element analysis engine. In the nonlinear dynamic cutting simulation model, the balance between the cement-soil slurry injection pressure and the soil shear resistance is dynamically calculated based on the cutting frequency parameters of the cutting head and the real-time depth of the cutting box. Using the equilibrium relationship as the internal control equation and the displacement boundary condition as the external boundary constraint, the finite element analysis engine is called to solve the seepage equation of the grout in the soil pores, and the pressure value of each grid node in the computational domain is obtained. Spatially interpolate and render the pressure values ​​at all grid nodes to generate a two-dimensional or three-dimensional pressure diffusion cloud map that reflects the distribution of pressure within the wall space. 8.The TRD wall quality intelligent detection system based on multi-source data fusion according to claim 7, characterized in that, The simulated travel speed and simulated torque values ​​output by the virtual twin model are compared in real time with the residual values ​​collected by the actual sensors of the construction machinery using the TRD method. This includes: In the virtual twin model, the simulated speed and simulated torque values ​​are calculated in real time based on the input parameters at the current moment. The data acquisition and alignment module synchronously acquires the measured values ​​of the actual travel speed sensor and the actual torque sensor from the TRD construction machinery, thus obtaining the numerical pair of measured travel speed and measured torque values. The simulated travel speed output by the virtual twin model and the measured travel speed collected by the actual sensor are numerically differentially calculated to obtain the travel speed residual. The simulated torque value output by the virtual twin model and the measured torque value collected by the actual sensor are calculated by numerical difference to obtain the torque value residual. Calculate the time series statistics of the travel speed residual and the torque value residual, respectively. The statistics include the residual mean, the moving average of the absolute value of the residual, and the variance of the residual change within a preset time window.

Images