Monitoring and feedback system for embankment compaction process under high liquid limit soil based on intelligent sensor

Abstract

The invention discloses a monitoring feedback system for an embankment compaction process under high liquid limit soil based on an intelligent sensor, and relates to the technical field of high liquid limit soil roadbed construction, and the monitoring feedback system comprises an intelligent sensor node buried in embankment filling soil, a data communication module connected with the node, and a feedback control terminal for receiving data; the intelligent sensor node is characterized in that the intelligent sensor node comprises a dynamic self-calibration anchoring mechanism which is composed of a deformable memory alloy net-shaped anchoring sleeve, and the sleeve is in a contraction state during initial embedding, expands into a multidirectional barb net-shaped structure and is embedded into a soil body when the temperature is increased to a preset phase change temperature range through rolling. According to the monitoring and feedback system for the compaction process of the embankment under the high liquid limit soil based on the intelligent sensor, the problem of data distortion caused by displacement and inclination of an embedded sensor is solved; and through intelligent linkage of an anchoring program and compaction parameter feedback, full-closed-loop control of displacement suppression, data fidelity and failure repair is formed, and the accuracy of compaction state evaluation and the reliability of feedback regulation and control are ensured.

Description

Technical Field

[0001] This invention relates to the field of high liquid limit soil subgrade construction technology, specifically to a monitoring and feedback system for the compaction process of high liquid limit soil embankments based on intelligent sensors. Background Technology

[0002] In road engineering, the compaction quality control of high liquid limit soils is crucial due to their high water content, strong plasticity, and low permeability. Traditional compaction testing methods rely on destructive sampling after the fact, which cannot provide real-time continuous monitoring and immediate feedback, making it difficult to guide dynamic compaction and easily leading to insufficient or excessive compaction, affecting embankment stability and work efficiency. Existing technologies introduce embedded intelligent sensor monitoring systems, which acquire parameters such as soil pressure and pore water pressure by deploying sensors and combine them with surface vibration response data to build an online perception and assessment capability for compaction status, thereby adjusting compaction parameters. However, existing systems face unresolved challenges in high liquid limit soil compaction monitoring: under vibratory compaction, high liquid limit soils exhibit significant and uneven plastic flow deformation, causing unexpected spatial orientation changes such as settlement and tilting of the embedded intelligent sensors and connecting components. This orientation inaccuracy severely distorts the correspondence between measured data and the actual soil state. For example, tilting the soil pressure sensor changes the direction of force, and settlement changes the burial depth, making the readings unable to reflect the pressure at the preset position; other sensors also lose representativeness due to positional offset. Data distortion accumulates and amplifies with each compaction pass, polluting the data source for monitoring and causing the assessment of compaction status and feedback control to lose their reliable basis. This may lead to misjudgments and ineffective control, reducing the value of the intelligent monitoring system.

[0003] Existing technologies primarily focus on sensor performance and placement strategies, but fail to address the physical interference problem of sensor misalignment caused by plastic deformation in high-liquid-limit soils and its fatal impact on the reliability of monitoring data. Therefore, the urgent technical challenge is to overcome the spatial misalignment of embedded sensors due to severe plastic deformation in high-liquid-limit soils, ensuring that monitoring data accurately reflects the state of the target soil. Summary of the Invention

[0004] To achieve the above objectives, the present invention is implemented through the following technical solution: a monitoring and feedback system for the compaction process of embankment under high liquid limit soil based on intelligent sensors, including intelligent sensor nodes buried in the embankment fill, a data communication module connected to the nodes, and a feedback control terminal for receiving data;

[0005] The smart sensor node includes:

[0006] The dynamic self-calibrating anchoring mechanism is composed of a deformable memory alloy mesh anchoring sleeve. The sleeve is initially in a contracted state when buried, and expands into a multi-directional barbed mesh structure and embeds itself into the soil when the rolling temperature rises to the preset phase transformation temperature range. The preset phase transformation temperature range is set based on the phase transformation characteristics of the memory alloy material and the temperature conditions of the engineering rolling construction.

[0007] The pose detection unit is fixedly installed inside the node and includes a three-axis accelerometer, a three-axis gyroscope and a three-axis magnetometer, which outputs three-dimensional spatial pose offset parameters in real time.

[0008] The dual-modal data fusion unit has its input end connected to the sensor probe and the pose detection unit, and its output end connected to the data communication module.

[0009] The feedback control terminal is configured to generate a compaction parameter adjustment command only after receiving the verification data output by the dual-modal data fusion unit.

[0010] Preferably, in the dynamic self-calibrating anchoring mechanism, the barb structure of the shape memory alloy mesh anchoring sleeve is asymmetrically distributed, and the long axis of the barbs is perpendicular to the embankment filling layer.

[0011] Preferably, the sampling frequency of the pose detection unit is a preset sampling frequency range, and the pose offset parameters include displacement vector, tilt angle and rotation amount; the preset sampling frequency range is set based on the pose dynamic detection accuracy requirements and data processing capabilities.

[0012] Preferably, the dual-modal data fusion unit performs the following operations:

[0013] a. Dynamically reconstructing the soil stress field distribution model based on pose offset parameters;

[0014] b. Input the original earth pressure data into the model for inversion calculation, and output the corrected data after pose compensation;

[0015] c. Apply an alternating current within a preset frequency range to the soil using a microelectrode array to obtain bioimpedance spectrum characteristics; the preset alternating current frequency range is set based on the soil electrical property detection requirements;

[0016] d. When the deviation between the corrected data and the bioimpedance spectrum characteristics in the compacted state mapping space exceeds a threshold, anchorage diagnosis is triggered.

[0017] Preferably, in the dual-modal data fusion unit, the electrode spacing of the microelectrode array is within a preset electrode spacing range, and the electrode surface is covered with an anti-ion adsorption coating; the preset electrode spacing range is set based on the requirements of bioimpedance spectroscopy detection resolution and signal stability.

[0018] Preferably, the anchorage diagnosis includes: calculating the correlation coefficient between the rate of change of pose offset parameters and the rate of change of bioimpedance spectrum; generating an anchorage failure signal when the correlation coefficient is lower than a preset correlation coefficient threshold, wherein the preset correlation coefficient threshold is set based on the reliability requirements of anchorage condition diagnosis.

[0019] Preferably, the system responds to the anchoring failure signal by executing:

[0020] The high-frequency, low-amplitude vibration command is sent to the road roller via the data communication module.

[0021] The built-in miniature piezoelectric actuator of the start-up node generates local micro-vibration to stimulate the sleeve for secondary anchoring.

[0022] Preferably, the feedback control terminal generates rolling parameter adjustment instructions including: roller travel path optimization instructions, vibration frequency adjustment instructions, amplitude adjustment instructions, and area rolling pass reset instructions.

[0023] Preferably, the data communication module adopts an anti-rollover frequency band frequency hopping communication protocol, and the communication frequency band avoids the peak range of the road roller's operating frequency.

[0024] Preferably, the compaction state mapping space is established through machine learning, with the inputs being the corrected earth pressure value, water content value, and bioimpedance spectrum characteristic value, and the outputs being the compaction degree grade and uniformity index.

[0025] This invention provides a monitoring and feedback system for the compaction process of embankments in high liquid limit soil based on intelligent sensors. It has the following beneficial effects:

[0026] This intelligent sensor-based monitoring and feedback system for the compaction process of embankments in high liquid limit soil uses a thermally triggered shape memory alloy directional anchoring mechanism to physically suppress the spatial orientation shift of the sensor caused by plastic deformation of the soil. By combining orientation perception with dual-modal data fusion, it solves the data distortion problem caused by displacement and tilting of embedded sensors. The intelligent linkage between the re-anchoring procedure and the compaction parameter feedback forms a closed-loop control of "displacement suppression - data fidelity preservation - failure repair", ensuring the accuracy of compaction status assessment and the reliability of feedback control.

[0027] This intelligent sensor-based monitoring and feedback system for the compaction process of embankments in high liquid limit soil utilizes dynamic anchoring technology to ensure the long-term stable operation of sensors under intense rolling, reducing the frequency of manual intervention. A self-evolving compaction state model dynamically optimizes output based on real-time working conditions, avoiding rework caused by over- or under-compaction. Precise control of roller vibration parameters and path achieves optimal resource allocation during the high liquid limit soil compaction process. Communication anti-interference mechanisms and node degradation strategies ensure continuous system operation in extreme environments, improving the first-pass yield and project progress controllability of embankment construction under complex geological conditions. Attached Figure Description

[0028] Figure 1 This is a schematic diagram of the overall module interaction of the present invention;

[0029] Figure 2 This is a schematic diagram illustrating the interaction between the perception layer, transmission layer, and execution layer of the present invention. Detailed Implementation

[0030] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0031] Please see Figure 1 and Figure 2 The present invention provides a technical solution: a monitoring and feedback system for the compaction process of embankment under high liquid limit soil based on intelligent sensors, including intelligent sensor nodes buried in the embankment fill, a data communication module connected to the nodes, and a feedback control terminal for receiving data;

[0032] The smart sensor node includes:

[0033] The dynamic self-calibrating anchoring mechanism is composed of a deformable memory alloy mesh anchoring sleeve. The sleeve is initially in a contracted state when buried, and expands into a multi-directional barbed mesh structure and embeds itself into the soil when the temperature rises to 50°C to 55°C during compaction.

[0034] The pose detection unit is fixedly installed inside the node and includes a three-axis accelerometer, a three-axis gyroscope and a three-axis magnetometer, which outputs three-dimensional spatial pose offset parameters in real time.

[0035] The dual-modal data fusion unit has its input end connected to the sensor probe and the pose detection unit, and its output end connected to the data communication module.

[0036] The feedback control terminal is configured to generate rolling parameter adjustment instructions only after receiving the verification data output by the dual-modal data fusion unit.

[0037] It should be further explained that, in the specific implementation process, the intelligent sensor node is buried within the high liquid limit soil embankment fill layer. In the initial state, its external deformable shape memory alloy mesh anchor sleeve remains in a contracted shape, reducing the embedment resistance. When the road roller performs vibratory compaction, the soil friction heats up to the critical range of 50°C to 55°C, triggering the shape memory alloy to undergo an austenitic phase transformation, and the sleeve expands into a multi-directional barbed mesh structure. At this time, the long axis of the barbs automatically orients to the direction perpendicular to the embankment bedding, embedding itself into the surrounding soil in an asymmetrical shape, forming a distributed mechanical anchor that resists plastic deformation, significantly suppressing the displacement and tilting of the sensor body caused by soil flow.

[0038] The nine-axis inertial measurement unit (IMU) fixedly installed inside the node operates in real time with a sampling frequency of 100Hz to 200Hz: a three-axis accelerometer captures linear displacement, a three-axis gyroscope monitors angular changes, and a three-axis magnetometer corrects the orientation reference, together generating pose offset parameters containing three-dimensional displacement vectors, tilt angles, and rotation amounts. These parameters are transmitted via hardwired to a dual-modal data fusion unit for synchronous processing with raw data collected by soil pressure sensors, moisture sensors, etc.

[0039] The dual-modal data fusion unit first dynamically reconstructs the soil stress field distribution model at the sensor location based on real-time pose offset parameters. When pose offset is detected, the original soil pressure data is input into the model for stress field inversion calculation, and the corrected pressure value to compensate for the pose error is output. Simultaneously, the unit controls the microelectrode array on the sensor surface to apply an alternating current of 100Hz to 10kHz to the soil, acquiring bioimpedance spectrum characteristics independent of the sensor pose. If there is a conflict between the corrected pressure value and the impedance spectrum characteristics in the preset compaction state mapping space, for example, the corrected pressure shows that the compaction is up to standard but the impedance spectrum reflects that the soil is loose, then the risk of anchorage failure is determined, and the following response mechanism is immediately triggered:

[0040] Calculate the correlation coefficient between the rate of change of pose offset and the rate of change of impedance spectrum. If the correlation coefficient is below 0.8 for 3 consecutive seconds, the anchorage failure is confirmed.

[0041] The anti-interference communication module sends a command to the road roller to implement a high-frequency, low-amplitude vibration of 30Hz to 35Hz above the failure point, which promotes the migration of fine soil to fill the voids in the anchoring zone.

[0042] The synchronous activation node has a built-in micro piezoelectric actuator that generates local micro-vibration to stimulate the memory alloy sleeve for secondary anchoring.

[0043] The feedback control terminal strictly adheres to the "verification first" principle: it only generates compaction parameter adjustment commands after receiving the anchoring stability compliance signal and verification data from the dual-modal data fusion unit. If the system is in the anchoring repair period, the control command output for that area is frozen to avoid making incorrect decisions based on distorted data. The command content is dynamically optimized based on the real-time compaction status, including replanning the roller's travel path to avoid the anchoring repair area, increasing the vibration frequency in under-compacted areas to 35Hz, reducing the amplitude in over-compacted areas to 2mm, and resetting the compaction pass counter for specific areas.

[0044] In the dynamic self-calibrating anchoring mechanism, the barb structure of the shape memory alloy mesh anchoring sleeve is asymmetrically distributed, and the long axis of the barbs is perpendicular to the embankment filling layer.

[0045] It should be further explained that, in the specific implementation process, the barb structure of the shape memory alloy mesh anchor sleeve is pre-formed with asymmetric geometric features through a hot-pressing process: the thickness of the barb root is 0.8mm to 1.2mm, gradually thinning to 0.2mm to 0.3mm towards the tip; the major axis of the barb is 1.8 to 2.2 times the minor axis, and the major axis direction automatically orients itself to the direction perpendicular to the embankment stratification during sleeve expansion. When the sleeve expands in response to the soil temperature rising to 50° to 55°, the barbs perpendicular to the stratification direction embed into the soil at an elevation angle of 70° to 80°, forming anchor points with a depth of 8mm to 12mm, while the barbs in the horizontal direction embed shallowly at an elevation angle of 40° to 50°. This asymmetric layout makes the vertical pull-out resistance more than 1.8 times that in the horizontal direction, specifically resisting the vertical settlement and horizontal shear rheology generated during the compaction of high liquid limit soil.

[0046] When the sensor node is subjected to oblique tensile force due to plastic deformation of the soil, the asymmetric stress characteristics of the barbs trigger an adaptive reinforcement mechanism: the barbs in the direction perpendicular to the bedding bear the main load, and their large-angle embedding shape enables the soil to generate a synergistic resistance of normal compressive stress and tangential frictional resistance to the barbs; the barbs in the horizontal direction play a limiting role to prevent the node from rotating or deflecting.

[0047] If the nine-axis inertial measurement unit detects that the angle between the displacement vector and the bedding plane is greater than 60°, it indicates that an abnormal lateral force has been subjected to the system. The system immediately activates the micro piezoelectric actuator to apply high-frequency micro-vibration of 5kHz to 8kHz to the horizontal weak area of ​​the sleeve, causing the barbs in the area to bounce and deepen their embedment, and rebalance the anchoring forces in all directions.

[0048] Under extreme conditions, such as a sudden turn of the roller causing localized soil impact, the barbed structure provides system protection through controlled fracture: a pre-set stress concentration groove at the barb root ensures that when the tensile force exceeds critical values ​​(150N vertically and 80N horizontally), the horizontal barb fractures first to release stress, preventing the sleeve from detaching entirely; the vertical barb retains its complete anchoring capacity, ensuring no abrupt changes in sensor orientation. The fracture signal is detected and uploaded by micro-strain gauges embedded in the sleeve. The feedback control terminal marks this node as entering degraded mode, and the frequency of bioimpedance spectroscopy verification is increased to twice that of normal conditions, compensating for the reliability decrease caused by anchoring force loss.

[0049] The sampling frequency of the pose detection unit is 100Hz to 200Hz, and the pose offset parameters include displacement vector, tilt angle, and rotation. It should be further noted that in the specific implementation, the nine-axis inertial measurement unit is fixed to the center of the sensor node via a rigid substrate. The three-axis accelerometer uses a piezoresistive MEMS device to detect linear acceleration in the range of 0.05g to 8g. The three-axis gyroscope uses a fiber optic structure to measure angular velocity of ±2000° / s. The three-axis magnetometer is equipped with a permalloy magnetic shielding layer to resist electromagnetic interference from the road roller. The unit operates at a sampling frequency of 100Hz to 200Hz. It activates an anti-interference mode each time the vibrating wheel of the road roller approaches. Specifically, when the accelerometer detects a vibration impact with an instantaneous value greater than 5g, it automatically blocks the data from the gyroscope and magnetometer during that period. During the vibration interval, when the acceleration value is less than 0.1g, it synchronously acquires raw data from 9 channels and performs temperature drift compensation. Through Kalman filter fusion processing, it generates pose offset parameters that include the three-dimensional displacement vector magnitude, tilt angle deviation, and rotational angular velocity.

[0050] The displacement vector synthesis employs a layered verification mechanism: first, the initial displacement is calculated using accelerometer integration; then, rotation artifacts are corrected using gyroscope angular velocity; and finally, the horizontal reference is verified using magnetometer azimuth. If the angle between the displacement vector direction and the embankment layer normal exceeds 45°, it is determined to be abnormal slip and a three-level response is triggered, including the following:

[0051] Level 1 response, i.e., angle 45°-60°: increase the sampling frequency to 200Hz for 5 seconds to eliminate transient interference;

[0052] Secondary response, i.e., angle 60°-75°: Freeze the current data output and start the microelectrode array bioimpedance spectroscopy verification;

[0053] Level 3 response, i.e., the included angle is greater than 75°: the marker pose is invalidated, and the re-anchoring procedure is forcibly started.

[0054] To address the unique creep characteristics of high liquid limit soils, the system incorporates a displacement accumulation threshold monitoring feature: when the cumulative vertical displacement reaches 15mm or the cumulative horizontal displacement reaches 8mm during a single compaction cycle, anchorage failure is determined regardless of the included angle. This determination takes precedence over the angle response mechanism, avoiding monitoring blind spots caused by slow plastic deformation.

[0055] The dual-modal data fusion unit performs the following operations:

[0056] a. Dynamically reconstructing the soil stress field distribution model based on pose offset parameters;

[0057] b. Input the original earth pressure data into the model for inversion calculation, and output the corrected data after pose compensation;

[0058] c. Bioimpedance spectrum characteristics were obtained by applying an alternating current of 100Hz to 10kHz to the soil through a microelectrode array;

[0059] d. When the deviation between the corrected data and the bioimpedance spectrum characteristics in the compacted state mapping space exceeds a threshold, anchorage diagnosis is triggered.

[0060] It should be further explained that, in the specific implementation process, after receiving the pose offset parameters, the dual-modal data fusion unit first determines whether the displacement vector magnitude exceeds 5mm or the tilt angle is greater than 10°. If the limits are not exceeded, the original sensor data is directly output; if the limits are exceeded, physical field mapping correction is initiated: based on the real-time displacement and tilt angle parameters, an ellipsoidal soil stress field model centered on the sensor is reconstructed. The major axis of this ellipsoid is aligned with the displacement vector direction, and the minor axis is set according to the Poisson's ratio compressibility of the soil. The original soil pressure data is input into the model for stress field inversion, and the distortion of the measurement reference surface caused by the pose offset is eliminated through iterative calculation to generate the corrected pressure value P'.

[0061] Simultaneous bioimpedance spectroscopy verification was initiated: the microelectrode array was controlled to switch the transmission frequency at 200Hz intervals, including 100Hz, 1kHz, 5kHz, and 10kHz, to acquire the real and imaginary parts of the complex impedance in multiple frequency bands. When an anomaly was detected in a certain frequency band, the data was immediately sampled three times again in that band, and the median value was taken. P' and the impedance spectral features were input into a pre-trained compaction state mapping space, which divides the compaction process of high liquid limit soil into three regions: loose state, transition state, and dense state. If P' is located in the dense state while the 10kHz real part of the impedance is in the loose state, a data conflict is identified, and the following graded response is triggered:

[0062] Level 1 conflict, i.e., single parameter deviation: increase the sampling frequency of that point to twice the normal value and re-verify;

[0063] Level 2 conflict, i.e., dual parameter deviation + pose change rate > 0.5 mm / s: freeze the control command output of this node and start the anchorage diagnosis program;

[0064] Level 3 conflict, namely: impedance spectrum full-band anomaly: mark sensor failure, switch to nearby node data replacement.

[0065] During anchorage diagnosis, the system continuously compares the rate of change of pose with the rate of change of the real part of impedance at 10kHz. When the correlation coefficient between the two is below 0.8 for five consecutive sampling periods, the anchorage is deemed to have failed. In this case, horizontal re-anchoring is prioritized: an 8kHz standing wave vibration is generated in the XY plane by a piezoelectric actuator, causing the shape memory alloy sleeve's horizontal barbs to re-embed; if the correlation coefficient still does not recover after 10 seconds, vertical reinforcement anchoring is added, triggering a road roller to perform a 32Hz / 2.5mm fixed-point vibration above the node for 20 seconds.

[0066] In the dual-modal data fusion unit, the electrode spacing of the microelectrode array is 0.5 mm to 1 mm, and the electrode surface is covered with an anti-ion adsorption coating. It should be further noted that in the specific implementation, the microelectrode array uses a platinum-iridium alloy substrate, and an iridium oxide-ruthenium dioxide composite coating is sputtered onto the electrode surface to form a nanoporous anti-adsorption layer with a thickness of 5 μm to 8 μm. This coating effectively suppresses the physical adsorption of clay minerals on the electrodes and the cation polarization effect in high liquid limit soil by increasing surface hydrophilicity and providing anion exchange sites. The electrode spacing is strictly controlled between 0.5 mm and 1 mm, less than 60% of the average particle size of the high liquid limit soil, ensuring that the alternating current penetrates the soil pores rather than circulating around the particle surface.

[0067] When an alternating current of 100Hz to 10kHz is applied, the system monitors the phase angle of the electrode-soil interface impedance in real time. If the absolute value of the phase angle in the low-frequency range (100Hz-500Hz) is detected to be >70°, ion adsorption blockage is determined to have occurred, and dual-mode switching is immediately initiated, including the following:

[0068] Normal mode, i.e., phase angle ≤ 70°: apply a continuous sine wave and scan the entire frequency band to obtain the real and imaginary parts of the complex impedance;

[0069] Pulse mode, i.e., phase angle > 70°: switches to pulse current with a duty cycle of 1:5, positive pulse width of 10ms combined with negative deionization pulse, forcibly stripping the adsorption layer on the electrode surface.

[0070] In pulse mode, after every three measurements, the electrode automatically performs self-cleaning: a reverse 1.5V DC voltage is applied for 200ms to electrolyze and remove residual ions. If the phase angle is still greater than 65° after self-cleaning, the coating is considered to have failed, triggering the following protection mechanisms: the weight of this electrode is downgraded to 30% of the normal value, and the proportion of other electrode data is increased; spatial interpolation compensation is performed using electrode data from adjacent nodes at the same layer; and an electrode maintenance alarm is uploaded to the feedback control terminal to prompt manual intervention.

[0071] To address electrode-soil contact fluctuations caused by compaction vibration, the system incorporates a contact quality index, Qc: calculated based on the impedance modulus fluctuation rate at 1kHz. When Qc is less than 0.6, the system automatically increases the transmission current intensity to 1.8 times the standard value to maintain signal penetration. If Qc still fails to meet the standard after increasing the current, the system freezes the electrode's data output, operates solely on pose correction in a single mode, and marks it as reliability degraded.

[0072] Anchorage diagnosis includes calculating the correlation coefficient between the rate of change of pose offset parameters and the rate of change of bioimpedance spectrum. An anchorage failure signal is generated when the correlation coefficient is below 0.8. It should be further noted that in the specific implementation, anchorage stability diagnosis is achieved by continuously comparing the rate of change of pose with the change in the real part of impedance at 10kHz; where the rate of change of pose is dV / dt, and the rate of change of the real part of impedance is dZ / dt. The system calculates the rolling correlation coefficient R with a period of 0.5 seconds. When the R value first falls below 0.8, a three-level verification process is initiated, including the following:

[0073] Level 1 verification: Freeze the current crushing command output, increase the sampling frequency to 500Hz for 1 second to eliminate instantaneous interference;

[0074] Secondary verification: If R is still < 0.8 and the magnitude of the pose displacement vector is > 3 mm after resampling, activate the horizontal piezoelectric micro-vibration for 3 seconds;

[0075] Level 3 verification: After vibration, the rate of R-value recovery is detected. If it does not recover to above 0.85 within 5 seconds, the anchorage is deemed to have failed.

[0076] Differentiated diagnostics were implemented for high liquid limit soils under different moisture contents. The diagnostic process is as follows:

[0077] Normal range, i.e., moisture content 18%-22%: Level 3 verification of the standard is implemented;

[0078] High plasticity zone, i.e., moisture content > 22%: relax the correlation coefficient threshold to 0.75, but add monitoring of cumulative displacement. When the horizontal displacement > 5mm, it is directly judged as a failure.

[0079] Critical region, i.e., moisture content <18%: tighten the R threshold to 0.85, while requiring the imaginary part of the impedance to change abnormally synchronously.

[0080] After confirming anchorage failure, the system prioritizes horizontal re-anchoring: the horizontal barbs of the sleeve resonate and bounce using XY plane piezoelectric standing waves, with each bounce deepening the embedment by 0.2mm to 0.3mm. If the correlation coefficient R is still below 0.75 after three consecutive bounces, the system switches to vertical reinforcement mode: a 32Hz / 2.5mm fixed-point vibration command is sent to the roller, with the duration calculated according to the formula T = 10 × (1-R) ​​seconds, a minimum of 10 seconds, and a maximum of 30 seconds; simultaneously, the austenitic phase transformation self-repair of the shape memory alloy sleeve is triggered, and a 2A DC pulse is applied to heat it to 60℃, so that the deformed barbs restore their initial geometry and are re-anchored.

[0081] The system responds to the anchoring failure signal by executing:

[0082] The data communication module sends a high-frequency, low-amplitude vibration command to the road roller, with a vibration frequency of 30Hz to 35Hz and a vibration amplitude of 2mm to 3mm.

[0083] The built-in miniature piezoelectric actuator of the start-up node generates local micro-vibration to stimulate the sleeve for secondary anchoring.

[0084] It should be further explained that during the specific implementation process, when the re-anchoring command is executed, the system first detects the real-time distance between the roller and the faulty node. When the roller is within 15 meters of the node, a high-frequency, low-amplitude vibration command is immediately sent: the vibration frequency is locked at 32Hz to 35Hz, matching the resonance frequency band of high liquid limit soil particles, and the amplitude is strictly controlled between 2mm and 3mm to avoid excessive disturbance to the compacted layer. The duration is dynamically calculated according to the anchoring stability index deficit ratio, with an increase of 5 seconds for every 0.1 below the threshold, a minimum of 10 seconds, and a maximum of 30 seconds. During vibration, the node's pose change rate is monitored in real time. If abnormal acceleration occurs, i.e., the vertical displacement rate is greater than 0.5mm / s, the vibration is immediately stopped and switched to pure piezoelectric micro-vibration mode.

[0085] The node's built-in micro piezoelectric actuator adopts a ring array layout. Upon receiving a re-anchoring command, it performs directional reinforcement, which includes two stages: Horizontal reinforcement stage: 8kHz standing wave vibration is excited in the XY plane, with the antinodes aligned with the horizontal barb area of ​​the shape memory alloy sleeve, causing the barbs to generate elastic flutter at a frequency of 500Hz, with each flutter deepening the embedment by 0.1mm to 0.2mm; Vertical supplement stage: If the correlation coefficient does not meet the standard after horizontal vibration, 5kHz simple harmonic vibration in the Z-axis direction is added, with an amplitude of 10μm to 20μm, to promote micro-slip and re-engagement between the vertical barbs and the soil.

[0086] When encountering intractable failure, i.e., when the anchoring index remains less than 0.75 after three consecutive vibration coordinations, the system activates an emergency degradation protocol, including the following:

[0087] Level 1 Degradation: Increase the weight of bioimpedance spectrum verification to twice the normal value, and reduce the weight of physical field correction to 50% to compensate for the data reliability loss due to insufficient anchoring;

[0088] Level 2 Degradation: If the pose offset continues to increase, i.e., the displacement vector magnitude is greater than 10mm, the control authority of the node will be automatically deactivated and a virtual monitoring point will be constructed by the adjacent node group.

[0089] Final action: When the horizontal displacement exceeds 15mm, a self-locking mechanism is triggered, and the sleeve releases a biodegradable retarder to wrap the sensor body, preventing dragging and damage to the cable. At the same time, it uploads accurate positioning information for manual excavation and maintenance.

[0090] The spatiotemporal coordination of roller vibration and node micro-vibration follows the principle of "macro-first, micro-second": macro-vibration is responsible for adjusting the soil skeleton and must be implemented when the roller passes directly above the node; micro-vibration is responsible for optimizing the anchoring interface and is delayed until 2 seconds after the vibrating wheel leaves to avoid signal superposition interference. A 2-second silence period is set between the two vibration modes to allow the nine-axis unit to collect clean pose data to evaluate the re-anchoring effect.

[0091] The feedback control terminal generates compaction parameter adjustment instructions, including: roller travel path optimization instructions, vibration frequency adjustment instructions, amplitude adjustment instructions, and area compaction pass reset instructions. It should be further noted that, in the specific implementation process, when the feedback control terminal generates compaction parameter adjustment instructions, it first constructs a compaction degree heatmap based on verified compaction state data. When the compaction degree improvement rate of three consecutive compactions in a certain area is lower than the preset threshold of 0.5% per pass, a four-dimensional optimization strategy is automatically activated, including the following:

[0092] Vibration frequency adaptation: For under-compacted areas, i.e., compaction degree <92, calculate the characteristic peak frequency Fp of the soil bioimpedance spectrum in the area, set the vibration frequency of the roller to 1.2×Fp, in the range of 25-40Hz, to excite soil particle resonance.

[0093] Amplitude gradient control: In the over-compression risk zone, i.e., when the compaction degree is >95% and the moisture content is >20%, a step-down amplitude reduction strategy is adopted. The amplitude is 3mm for the first pass and 0.5mm for each subsequent pass, down to a minimum of 1.5mm.

[0094] Reset the number of compaction passes: Pause compaction in areas with a compaction uniformity index > 0.8, and add two passes of directional compaction in areas with an index < 0.6;

[0095] Dynamic obstacle avoidance: The coordinates of anchorage repair nodes are calibrated in real time, and a temporary restricted area with a radius of 2m is generated. The roller path automatically detours and reassembles into a closed loop.

[0096] To address the rheological properties of high liquid limit soils, the system implements a combined compaction-intermittent control: when a single continuous compaction exceeds five passes, a 10-minute static interval is forcibly inserted, during which the pore water pressure dissipation rate is monitored via a microelectrode array. If the dissipation rate is <30% / 10min, the interval is automatically extended to 20 minutes and vertical drainage micro-vibration is triggered.

[0097] The compaction quality assessment adopts a dual-track system, including the following:

[0098] Normal mode: The earth pressure value P' after correction and the real part of the 10kHz impedance are used for joint judgment, and both are required to meet the dense state standard;

[0099] Anchorage degradation mode: When a node is in a first-level degradation state, the determination is based on bioimpedance spectrum, combined with spatial interpolation compensation of neighboring node data.

[0100] Failure mode: After a node is decoupled, a virtual compaction model for that point is constructed based on historical data, and the model parameters are updated every 5 new sets of neighboring node data.

[0101] The path optimization follows the principle of minimum disturbance: the turning point of the roller is always set in the qualified area, and sharp turns in the under-compacted area are prohibited. When the centrifugal force of the turn causes the horizontal displacement of the node to be greater than 3mm, the turning radius in that area is automatically reduced by 50%, and the turning speed is reduced to 0.5m / s.

[0102] The data communication module employs a frequency-hopping communication protocol for the anti-rollover band, avoiding the peak operating frequency range of the road roller. It should be further noted that in specific implementation, the data communication module uses a three-band frequency-hopping protocol: a low-frequency channel of 868MHz for routine data transmission, a mid-frequency channel of 2.4GHz for carrying pose parameters and alarm signals, and a high-frequency channel of 5.8GHz dedicated to video stream transmission during the re-anchoring phase. When the road roller activates its vibratory drum, the system collects its operating frequency f in real time. v Dynamically avoid [f v -5Hz, f v The harmonic interference frequency band is +5Hz. The communication timing is strictly synchronized with the rolling vibration: when the vibrating wheel of the road roller is more than 10 meters away from the node, the full-band high-speed transmission is started; after entering the 10-meter warning zone, it switches to pulse mode and only uses the gap between the vibrating wheel and the ground, that is, in a 20ms-50ms window per cycle, to send key position and orientation verification data.

[0103] Channel quality assessment employs a two-parameter diagnostic approach, including the following:

[0104] Bit error rate-driven diagnosis: When the bit error rate of 3 consecutive packets is >10 -3 If electromagnetic interference is detected, the system will immediately switch to a preset clean channel.

[0105] Signal strength drop diagnosis: If the received signal strength RSSI decreases by more than 40% within 0.5 seconds, it is determined to be a mechanical blockage, triggering relay node relay transmission.

[0106] To address signal attenuation caused by changes in water content in high liquid limit soils, adaptive power compensation is implemented, including the following:

[0107] Normal range, i.e., moisture content 18%-22%, maintains standard transmission power;

[0108] In the high humidity zone, i.e., with a moisture content > 25%, the power is increased to 1.8 times the standard value, while the modulation method is downgraded from QPSK to BPSK.

[0109] Critical region, i.e., moisture content <18%: Enable multi-node spatial diversity reception, and merge the signals of the three nearest nodes to improve the signal-to-noise ratio.

[0110] Under extreme interference, such as when multiple road rollers are operating in parallel, a Level 4 emergency protocol will be activated, including:

[0111] Level 1 response, i.e., interference duration < 30 seconds: data is cached locally and retransmitted after the interference subsides;

[0112] Secondary response, i.e., interference for 30-120 seconds: activate the surface acoustic wave sensor to establish a temporary acoustic communication link and transmit a 200kHz ultrasonic modulated signal through the soil;

[0113] Level 3 response, i.e., interference > 120 seconds: switch to ultra-low power LoRa mode, reduce the data rate to 1kbps to maintain life signal transmission;

[0114] The ultimate solution is: if communication is interrupted for more than 5 minutes, the node enters a sleep state and sends a 1ms pulse every 30 seconds for location searching to avoid running out of energy.

[0115] During the communication recovery phase, data integrity verification is performed: missing data packets are automatically requested for retransmission by comparing timestamps. If the verification failure rate is greater than 20%, a full-node channel reassembly is triggered. All communication data is labeled with construction coordinates to ensure that the feedback control terminal is accurately associated with the physical location.

[0116] The compaction state mapping space is established through machine learning. The inputs are the corrected earth pressure value, water content value, and bioimpedance spectrum characteristic value, while the outputs are the compaction degree grade and uniformity index. It should be further explained that in the specific implementation process, the compaction state mapping space is dynamically constructed using a deep convolutional-temporal hybrid neural network. The input layer simultaneously receives three types of key parameters, including physical field correction data, electrical characteristic parameters, and environmental state quantities. Specifically: physical field correction data includes the pose-compensated earth pressure value P' and pore water pressure value U'; electrical characteristic parameters include the real and imaginary parts of the impedance at 100Hz / 1kHz / 10kHz frequencies; and environmental state quantities include the real-time water content ω, soil temperature T, and anchorage stability index K.

[0117] The initial model is pre-trained during the laboratory calibration phase, and a dual evolutionary mechanism is initiated after field deployment, including the following:

[0118] Short-term online fine-tuning: After each single rolling pass is completed, high-confidence data is automatically selected, i.e., K≥0.9 and communication error rate<10. -4 Update the weights of the fully connected layer;

[0119] Long-term structural reconstruction: When a sudden change in soil quality or seasonal transition is detected, the convolution kernel dimension is reorganized to adapt to the new feature space.

[0120] The model output is expressed probabilistically, including compaction grade and uniformity index. The compaction grade is the probability distribution of loose state (0-0.3), transitional state (0.3-0.7), and dense state (0.7-1.0). The uniformity index is calculated based on the standard deviation of compaction within a 0.5m×0.5m grid. The smaller the value, the more uniform the compaction.

[0121] Model credibility is ensured through closed-loop verification, which includes the following:

[0122] Cross-validation: Randomly select one group from every 50 prediction results and compare it with the measured value of the manual ring cutter method at that point. If the absolute error is >5%, initiate a special optimization of the model for that area;

[0123] Anchoring linkage verification: When the K value drops to the range of 0.75-0.8, the probability threshold of the output result of this node is automatically increased, and the dense state judgment standard is increased from 0.7 to 0.8;

[0124] Conflict resolution mechanism: If the physical field data indicates a dense state P' > 150 kPa, while the impedance spectrum indicates a loose state with a real part of 10 kHz < 50 Ω·m, the impedance spectrum result is accepted first and the P' data is marked as invalid.

[0125] A three-level response is implemented to address the risk of model failure, including the following:

[0126] Level 1 response, i.e., prediction confidence < 0.6: freeze control command, increase the data weight of adjacent nodes to 150% for compensation;

[0127] Level 2 response, i.e.: prediction error > 8% for 3 consecutive points: roll back to the previous stable version of the model and start offline training simultaneously;

[0128] The third-level response, namely: the abrupt change in moisture content leads to distortion of the characteristic space: switch to a simplified impedance-compaction univariate model, and gradually restore the full model after the environment stabilizes.

[0129] Spatial data is bound to the construction coordinate system, and the specific model parameters for that area are automatically loaded when the road roller enters a grid. After compaction is completed, a model iteration report is generated, recording variables whose feature weight change rate is greater than 10%, for reference in subsequent projects.

[0130] The intelligent sensor node is embedded within the fill layer of a high liquid limit soil embankment. Its core innovation lies in the collaborative operation of a dynamic self-calibrating anchoring mechanism and a dual-modal data fusion unit. Initially, the shape memory alloy mesh sleeve encasing the node is in a contracted state, significantly reducing insertion resistance. When the soil temperature rises to a specific critical range due to roller compaction, the shape memory alloy undergoes a solid-state phase transition, and the sleeve expands into a multi-directional barbed structure. At this point, the barbs automatically orient themselves perpendicular to the embankment fill layer, embedding themselves asymmetrically into the surrounding soil to form distributed anchoring points that resist plastic deformation. The vertical barb embedding depth and pull-out resistance are significantly higher than those in the horizontal direction, specifically suppressing vertical settlement and horizontal rheological displacement during the compaction process of high liquid limit soil.

[0131] A nine-axis inertial measurement unit (IMU) fixedly installed inside the node continuously monitors its spatial status. When the vibrating roller approaches, the unit activates an anti-interference mode: shielding angular velocity and magnetic field data during periods of strong vibration, and only collecting valid signals during vibration intervals. Through multi-sensor data fusion processing, it outputs pose offset parameters in real time, including three-dimensional displacement vectors, tilt deviations, and rotational amounts. When the angle between the displacement direction and the bedding plane normal exceeds a threshold, the system automatically increases the sampling frequency or initiates an emergency check. To address the unique creep characteristics of high liquid limit soils, a cumulative displacement monitoring mechanism is added; exceeding the vertical or horizontal displacement limit during a single compaction cycle determines an anchorage risk.

[0132] The dual-modal data fusion unit constitutes the system's decision-making center. When the pose deviation exceeds the limit, the unit first reconstructs the soil stress field distribution model based on real-time displacement and tilt angle, eliminates measurement reference plane distortion through inversion calculation, and generates corrected earth pressure values. A synchronously operating bioimpedance spectroscopy verification mechanism applies multi-band alternating current to the surrounding soil to obtain electrical characteristic parameters independent of sensor pose. The corrected data and impedance spectral characteristics are jointly input into the compaction state mapping space for consistency comparison: if both point to the same compaction state, verification data is output; if the physical field data shows compaction while the impedance spectrum reflects looseness, anchorage failure diagnosis is triggered; if the impedance spectrum is abnormal across the entire frequency band, it is determined to be a sensor fault, and data from neighboring nodes is switched.

[0133] Anchorage failure diagnosis is achieved through the dynamic correlation between the rate of change of pose and the rate of change of impedance characteristics. When the correlation coefficient remains below a threshold, the system initiates a tiered re-anchoring procedure: first, horizontal barb micro-flutter is excited by the piezoelectric actuator at the node to achieve precise and deeper embedding; if the expected effect is not achieved, the roller is coordinated to implement high-frequency, low-amplitude fixed-point vibration in the target area, simultaneously triggering the phase change self-healing of the shape memory alloy sleeve. The entire re-anchoring process follows the principle of spatiotemporal decoupling: the macroscopic vibration of the roller is responsible for adjusting the soil skeleton and must be executed when passing directly above the node; the micro-vibration of the node is responsible for optimizing the anchorage interface and is delayed until after the vibrating wheel leaves.

[0134] The feedback control terminal strictly adheres to the verification-first principle. It only generates compaction parameter adjustment commands after receiving a signal indicating that the anchoring stability has met the standards and dual-modal verification data. The command generation is based on a real-time constructed compaction heatmap, incorporating a four-dimensional optimization strategy: dynamically setting the roller vibration frequency for under-compacted areas according to the characteristic peak frequency of the bioimpedance spectrum; implementing amplitude step-decreasing control for over-compacted risk areas; resetting the number of compaction passes in the area based on the uniformity index; and dynamically avoiding anchoring repair nodes and reorganizing the compaction path.

[0135] The communication system employs a frequency band switching and spatiotemporal coordinated anti-interference mechanism. Low-frequency channels transmit routine data, high-frequency channels are dedicated to video streams, and mid-frequency channels carry critical alarm signals. Transmission timing is strictly synchronized with compaction vibration: full-speed communication is maintained when the roller is far from the node; after entering the warning zone, core parameters are transmitted only using the ground clearance of the vibrating wheel. In the event of continuous strong interference, emergency responses are initiated sequentially, including acoustic soil transmission, low-power mode, and sleep pulse positioning.

[0136] The compaction state mapping space is an autonomously evolving model. The initial version was pre-trained and calibrated in the laboratory, and then continuously optimized through a dual mechanism after field deployment, including short-term online fine-tuning and long-term structural reconstruction. Short-term online fine-tuning involves updating the weight parameters with high-confidence data after each compaction pass; long-term structural reconstruction involves adjusting the model architecture when soil conditions change abruptly or the environment changes drastically.

[0137] The model outputs a compaction probability distribution and a uniformity index, and its reliability is ensured through a triple closed-loop system: periodic manual sampling and comparison, anchorage status correlation verification, and priority adoption of impedance spectrum results for conflicting data. When the model's confidence is insufficient, the system automatically increases the weight of neighboring node data or switches to a simplified model to maintain system robustness.

[0138] The rolling heat effect triggers a phase transformation in shape memory alloys, forming an asymmetric directional anchoring structure. The embedding depth and angle of the barbs are optimized according to the embankment bedding direction, making the vertical pull-out resistance higher than the horizontal direction, specifically suppressing displacement in the main deformation direction of high liquid limit soil. A pre-set stress concentration structure at the barb root enables controllable fracture in the horizontal direction, ensuring that the core anchoring capacity is preserved under extreme working conditions.

[0139] Physical field correction dynamically reconstructs the stress field model based on pose parameters, eliminating measurement reference distortion; bioimpedance spectroscopy verification obtains intrinsic soil characteristics independent of sensor displacement through multi-band scanning. The cross-validation of the two in the compaction state mapping space provides double protection, especially when the pose correction results conflict with the impedance spectrum characteristics, automatically triggering the anchoring system depth diagnosis.

[0140] The macroscopic vibration and nodal micro-vibration of the roller are separated by a quiet period: macroscopic vibration is only implemented when the roller passes directly overhead, and is responsible for adjusting the soil skeleton; micro-vibration is started later and focuses on optimizing the anchorage interface. In the horizontal direction, high-frequency micro-flutter is preferred to achieve gradual deepening of the barbs, avoiding secondary displacement caused by vibration.

[0141] The compaction state mapping space uses the anchorage stability index as the core weighting factor for model confidence, overcoming the limitation of the algorithm neglecting sensor reliability. Model parameters are bound to the construction coordinate grid, and combined with online fine-tuning and structural reconstruction capabilities, dynamically adapt to soil differences and environmental changes in different sections of the embankment.

[0142] By using a thermally triggered shape memory alloy directional anchoring mechanism, the spatial orientation shift of the sensor caused by soil plastic deformation is suppressed from a physical level; by combining orientation perception with dual-modal data fusion, the data distortion problem caused by displacement tilt of the embedded sensor is solved; the intelligent linkage between the re-anchoring program and the compaction parameter feedback forms a closed-loop control of "displacement suppression - data fidelity preservation - failure repair", ensuring the accuracy of compaction status assessment and the reliability of feedback control.

[0143] Dynamic anchoring technology ensures the long-term stable operation of sensors under intense compaction, reducing the frequency of manual intervention; the self-evolving compaction state model dynamically optimizes output based on real-time working conditions, avoiding rework caused by over- or under-compaction; precise control of roller vibration parameters and path achieves optimal resource allocation in the compaction process of high liquid limit soils; communication anti-interference mechanisms and node degradation strategies ensure continuous operation of the system in extreme environments, improving the first-pass yield and project progress controllability of embankment construction under complex geological conditions.

[0144] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0145] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A monitoring and feedback system for the compaction process of embankments in high liquid limit soil based on intelligent sensors, characterized in that, It includes intelligent sensor nodes buried in the embankment fill, data communication modules connected to the nodes, and feedback control terminals for receiving data; The intelligent sensor node is characterized by comprising: The dynamic self-calibrating anchoring mechanism is composed of a deformable memory alloy mesh anchoring sleeve. The sleeve is initially in a contracted state when buried, and expands into a multi-directional barbed mesh structure and embeds itself into the soil when it is rolled and heated to the preset phase change temperature range. The pose detection unit is fixedly installed inside the node and includes a three-axis accelerometer, a three-axis gyroscope and a three-axis magnetometer, which outputs three-dimensional spatial pose offset parameters in real time. The dual-modal data fusion unit has its input end connected to the sensor probe and the pose detection unit, and its output end connected to the data communication module. The feedback control terminal is configured to generate a compaction parameter adjustment command only after receiving the verification data output by the dual-modal data fusion unit.

2. The monitoring and feedback system for the compaction process of embankments under high liquid limit soil based on intelligent sensors according to claim 1, characterized in that: In the dynamic self-calibrating anchoring mechanism, the barb structure of the shape memory alloy mesh anchoring sleeve is asymmetrically distributed, and the long axis of the barbs is perpendicular to the embankment filling layer.

3. The monitoring and feedback system for the compaction process of embankments under high liquid limit soil based on intelligent sensors according to claim 1, characterized in that: The sampling frequency of the pose detection unit is within a preset sampling frequency range, and the pose offset parameters include displacement vector, tilt angle, and rotation amount.

4. The monitoring and feedback system for the compaction process of embankments under high liquid limit soil based on intelligent sensors according to claim 1, characterized in that: The dual-modal data fusion unit performs the following operations: a. Dynamically reconstructing the soil stress field distribution model based on pose offset parameters; b. Input the original earth pressure data into the model for inversion calculation, and output the corrected data after pose compensation; c. Apply an alternating current within a preset frequency range to the soil using a microelectrode array to obtain bioimpedance spectrum characteristics; d. When the deviation between the corrected data and the bioimpedance spectrum characteristics in the compacted state mapping space exceeds a threshold, anchorage diagnosis is triggered.

5. The monitoring and feedback system for the compaction process of embankment under high liquid limit soil based on intelligent sensors according to claim 4, characterized in that: In the dual-modal data fusion unit, the electrode spacing of the microelectrode array is within a preset electrode spacing range, and the electrode surface is covered with an anti-ion adsorption coating.

6. The monitoring and feedback system for the compaction process of embankments under high liquid limit soil based on intelligent sensors according to claim 4, characterized in that: The anchorage diagnosis includes: calculating the correlation coefficient between the rate of change of pose offset parameters and the rate of change of bioimpedance spectrum, and generating an anchorage failure signal when the correlation coefficient is lower than a preset correlation coefficient threshold.

7. The monitoring and feedback system for the compaction process of embankments under high liquid limit soil based on intelligent sensors according to claim 6, characterized in that: The system responds to the anchoring failure signal by executing: The high-frequency, low-amplitude vibration command is sent to the road roller via the data communication module. The built-in miniature piezoelectric actuator of the start-up node generates local micro-vibration to stimulate the sleeve for secondary anchoring.

8. The monitoring and feedback system for the compaction process of embankment under high liquid limit soil based on intelligent sensors according to claim 1, characterized in that: The feedback control terminal generates rolling parameter adjustment instructions, including: roller travel path optimization instructions, vibration frequency adjustment instructions, amplitude adjustment instructions, and area rolling pass reset instructions.

9. The monitoring and feedback system for the compaction process of embankment under high liquid limit soil based on intelligent sensors according to claim 1, characterized in that: The data communication module adopts an anti-rollover frequency band frequency hopping communication protocol, and the communication frequency band avoids the peak range of the road roller's operating frequency.

10. The monitoring and feedback system for the compaction process of embankment under high liquid limit soil based on intelligent sensors according to claim 4, characterized in that: The compaction state mapping space is established through machine learning. The inputs are the corrected earth pressure value, water content value, and bioimpedance spectrum characteristic value, and the outputs are the compaction degree grade and uniformity index.

Images