Response monitoring method of embedded self-sensing cement-based sensor in cement concrete pavement

Abstract

The application discloses a response monitoring method of a buried self-sensing cement-based sensor in a cement concrete pavement, and comprises the following steps: S1, self-sensing cement-based sensor preparation and performance test; S2, buried sensor packaging treatment; S3, sensor piezoresistive performance pre-calibration; S4, sensor pavement burying installation and implementation monitoring; S5, monitoring data preprocessing; S6, pavement damage discrimination method; S7, SVM intelligent damage identification and early warning, and the application realizes the collaborative optimization of mechanical properties and conductive properties; through sensor pre-calibration, an exclusive piezoresistive model is established, intelligent analysis of the pavement health state is realized in combination with an SVM algorithm, real-time monitoring of pavement stress and strain and structure damage identification can be accurately completed, technical difficulties such as traditional sensor interface separation and signal drift are effectively solved, and the application is suitable for long-term health monitoring of cement concrete pavements.

Description

Technical Field

[0001] This invention belongs to the field of intelligent building materials and road health monitoring technology, specifically relating to a method for monitoring the response of an embedded self-sensing cement-based sensor on a cement concrete pavement. Background Technology

[0002] Cement concrete pavement, as a key component of domestic transportation infrastructure, is widely used in important pavement structures such as highways and airport runways. Due to its excellent load-bearing capacity and durability, cement concrete pavement has long been the preferred material for the construction of major transportation arteries. However, with the increase in service life and traffic load, pavement is affected by various factors during long-term service, leading to structural damage such as deformation and cracking. This damage not only affects the strength, stability, and durability of the pavement but may also cause serious traffic safety hazards and increase the difficulty of road maintenance and repair. Therefore, timely and accurate monitoring of changes in its mechanical response is crucial to ensuring the long-term safety and functionality of the pavement structure.

[0003] Currently, monitoring of cement concrete pavements mainly relies on traditional equipment such as strain gauges and fiber optic sensors. These devices are typically complex to install, difficult to maintain, and unsuitable for long-term monitoring. Furthermore, existing monitoring equipment is often incompatible with the cement concrete matrix material, resulting in differences in interfacial adhesion. This not only easily leads to sensor detachment from the concrete matrix but can also cause signal distortion and drift. Consequently, existing monitoring systems struggle to provide stable and accurate real-time data when facing complex environmental changes.

[0004] Self-sensing cement-based materials are a type of intelligent building material that can sense its own stress and strain state in real time. Its core working principle is that under the action of external load, the internal conductive network of the material is reconstructed, which in turn causes regular changes in resistivity. By monitoring the resistivity, the stress state of the material can be inverted, providing a brand-new technical approach for road health monitoring.

[0005] However, self-sensing cement-based materials with single conductive fillers generally suffer from an irreconcilable "mechanical-electric contradiction." While the incorporation of conductive fillers can effectively reduce the resistivity of cement-based materials, it inevitably has a negative impact on their mechanical properties. Currently, research on composite conductive fillers mainly focuses on graphite and carbon fiber systems. Although graphite has excellent conductivity, its cost is more than twice that of other commonly used conductive fillers.

[0006] For example, the Chinese invention patent CN1821152A discloses a method for preparing a cement-based graphite steel fiber composite conductive material, belonging to the field of new materials technology. The aim is to provide a novel cement-based composite material for bridge decks and road surfaces that conducts electricity and generates heat for snow melting and de-icing. This novel composite material can also be used in structural components for self-monitoring of structural damage. The invention is characterized by spreading steel fibers in a mold, then injecting a highly fluid slurry prepared from cement-graphite mortar and a high-efficiency water-reducing agent in a specific ratio into the mold, followed by curing to form a novel conductive building material. The effects and benefits of this invention are that this novel cement-based composite material has low resistivity, excellent mechanical properties, and good long-term resistivity stability, with no significant change in resistivity over two years. This material is simple to produce, inexpensive, and widely applicable in bridge deck and road surface snow melting and de-icing, indoor floor heating, and structural damage self-monitoring projects, generating significant economic and social benefits.

[0007] For example, Chinese invention patent CN104628294A discloses a graphene oxide-based composite material for cement-based materials. This graphene oxide-based composite material is made from the following materials in parts by weight: 20-25 parts of modified graphene oxide nanosheet dispersion, 120-140 parts of macromonomer aqueous solution, 30-40 parts of small monomer mixture aqueous solution, and 30-40 parts of initiator aqueous solution. This invention also discloses a method for preparing a graphene oxide-based composite material for cement-based materials. The graphene oxide-based composite material prepared by this invention, when incorporated into cement materials, has the characteristics of high water reduction rate, good initial fluidity, and significant later strengthening and toughening effects. At the same time, using this composite material can significantly reduce the amount of graphene oxide incorporated. The method and product of this invention have the characteristics of unique preparation process and significant economic and social benefits.

[0008] Therefore, combining the aforementioned patents and existing technologies, when the graphite content reaches 10 wt.%, although the resistivity of cement-based materials decreases significantly, their compressive strength decreases by 77%. Carbon fiber systems, on the other hand, are prone to fiber entanglement and agglomeration, making it difficult to achieve uniform dispersion in monofilament form. Specifically, single conductive filler systems face the dual bottlenecks of imbalanced electromechanical properties and excessively high economic costs: at low dosages, the conductive filler exhibits poor dispersion and struggles to form a continuous conductive network; at high dosages, although the resistivity decreases significantly, excessive filler introduces numerous interface defects, disrupting the dense structure of cement hydration products and failing to meet the basic load-bearing requirements of engineering structures, severely limiting the feasibility of large-scale engineering applications.

[0009] Therefore, overcoming the dual limitations of poor performance of single conductive fillers and excessively high cost of composite filler systems, and designing and preparing self-sensing cement-based materials with high mechanical strength, excellent conductivity, and economically controllable preparation costs has become a core scientific problem in this research. Against this backdrop, utilizing self-sensing cement-based materials as sensors to achieve long-term stress and strain monitoring of cement concrete pavements is a core requirement for overcoming the bottlenecks in pavement monitoring technology. Furthermore, it is a crucial technical issue that urgently needs to be addressed to improve the scientific nature of pavement maintenance decisions, increase maintenance efficiency, and ensure the long-term operational safety of roads. Summary of the Invention

[0010] The purpose of this invention is to propose a response monitoring method for embedded self-sensing cement-based sensors on cement concrete pavements, which effectively solves the technical problems of interface detachment and signal drift of traditional sensors, and is suitable for long-term health monitoring of cement concrete pavements.

[0011] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0012] A method for monitoring the response of embedded self-sensing cement-based sensors on cement concrete pavements includes the following steps:

[0013] S1. Preparation and performance testing of self-sensing cement-based sensor;

[0014] S2, Embedded sensor packaging process;

[0015] S3. Pre-calibration of sensor piezoresistive performance;

[0016] S4. Sensor installation and monitoring on road surface;

[0017] S5. Monitoring data preprocessing;

[0018] S6. Methods for identifying road surface damage;

[0019] S7 and SVM intelligent damage recognition and early warning.

[0020] As a preferred technical solution of the present invention, step S1 is specifically as follows:

[0021] S11. The fabrication of the self-sensing cement-based sensor is as follows:

[0022] S111. Determine the optimal composition:

[0023] It is a mixture of raw materials comprising the following parts by weight: 100 wt.% cement, 0.2 wt.%–0.5 wt.% multi-walled carbon nanotubes (MWCNTs), 0.2 wt.%–1.0 wt.% nano-carbon black (NCB), 0.1 wt.%–0.3 wt.% polycarboxylate superplasticizer, and 30 wt.%–45 wt.% water.

[0024] S112, Preparation of components:

[0025] Pre-dispersion treatment: Dissolve polycarboxylate superplasticizer in water and stir until completely dissolved to obtain a base solution. Add nano carbon black NCB to the obtained base solution and stir for 3 minutes. The nano carbon black NCB agglomerates are initially broken by mechanical shear force, so that the polycarboxylate superplasticizer is fully adsorbed on the surface of nano carbon black NCB to obtain a pre-dispersion liquid.

[0026] Secondary synergistic dispersion: Multi-walled carbon nanotubes (MWCNTs) are added to the above pre-dispersion liquid and stirred for 3 min; the polycarboxylate superplasticizer is adsorbed on the surface of MWCNTs through π-π stacking, and its hydrophilic branches extend outward to form a steric hindrance layer, which prevents the aggregation of MWCNTs and simultaneously achieves synergistic and stable dispersion of MWCNTs and NCB nanocarbon black, resulting in a mixed dispersion liquid;

[0027] Ultrasonic homogeneous dispersion: The above-mentioned mixed dispersion after secondary dispersion is placed in an ultrasonic disperser, the ultrasonic frequency is set to 25kHz~35kHz, and the ultrasonic treatment time is 15min~25min. The residual agglomerates of conductive fillers are further broken by ultrasonic cavitation effect, so that the two conductive fillers form a uniform and stable co-dispersion system and a conductive filler composite suspension is obtained.

[0028] Demolding and curing: The prepared conductive filler composite suspension is added to cement while being stirred at low speed for 2 minutes, and then stirred at high speed for 3 minutes to obtain a uniform cement slurry. The obtained cement slurry is poured into a preset mold for molding. After demolding, it is placed in a standard curing environment for curing until the specified age to obtain a self-sensing cement-based sensor.

[0029] S12. Performance testing of the self-sensing cement-based sensor, as detailed below:

[0030] The mechanical properties of the fabricated self-sensing cement-based sensor were tested according to the "Test Method for Strength of Cement Mortar". The resistivity of the fabricated self-sensing cement-based sensor was tested using the four-electrode method.

[0031] As a preferred technical solution of the present invention, step S2 is specifically as follows:

[0032] S21. Pretreatment and Electrode Setup:

[0033] After 28 days of curing, the self-sensing cement-based sensor was used with a four-electrode method. The wires were soldered to the self-sensing cement-based sensor with a soldering iron. The soldering points and the external parts of the wires were wrapped with insulating tape.

[0034] S22, Surface modification treatment:

[0035] The self-sensing cement-based sensor was sandblasted using a compressed air sandblasting machine. After sandblasting, the surface was cleaned with compressed air and allowed to dry.

[0036] S23. Waterproofing treatment:

[0037] The self-sensing cement-based sensor, after sandblasting, is then encapsulated in waterproof epoxy resin, as detailed below:

[0038] Peelable masking tape was used to precisely mask the electrode contact points and wire lead-out parts at both ends of the self-sensing cement-based sensor. Then, a two-component epoxy resin coating was uniformly applied to the entire surface of the specimen, and the dry film thickness of the coating was controlled to be 0.2 mm.

[0039] After coating, let it stand at room temperature until the epoxy resin coating initially sets, then remove the masking tape to ensure that the electrode contact points are completely exposed and the wires are led out normally.

[0040] After the epoxy resin coating has fully cured, the waterproof encapsulation of the self-sensing cement-based sensor is completed.

[0041] As a preferred technical solution of the present invention: In step S22, the pressure of the compressed air sandblasting machine is set to 0.6MPa, the sandblasting distance is 10cm, and the quartz sand particle size is 60 mesh to 100 mesh, so that the surface roughness Ra of the self-sensing cement-based sensor reaches 2.0μm to 3.5μm.

[0042] As a preferred technical solution of the present invention, step S3 is as follows:

[0043] S31. Elimination of polarization effect:

[0044] The self-sensing cement-based sensor was connected to the test circuit using the four-electrode method. A constant test current was provided by a DC regulated power supply. After the sensor was powered on and left to stand for 3 minutes, the initial resistance value and initial resistivity of the self-sensing cement-based sensor were recorded.

[0045] S32, Loading System:

[0046] The stress control mode was used to load the self-sensing cement-based sensor, and the real-time resistance signal, stress value and strain value of the self-sensing cement-based sensor under different load levels were collected simultaneously.

[0047] S33. Characteristic parameter calculation and calibration model establishment:

[0048] Based on the resistivity, stress, and strain data obtained in step S32, the influence of contact resistance and wire resistance is eliminated by the four-electrode method. The following core characteristic parameters are calculated to establish a quantitative piezoresistive model of resistivity change rate-stress-strain for each self-sensing cement-based sensor.

[0049] The change in resistivity is quantitatively characterized by the rate of change of resistivity (FCR):

[0050] (1);

[0051] R0 and These represent the initial resistance and initial resistivity of the self-sensing cement-based sensor after it has been powered on for a period of time, respectively; Rt and Here, represents the resistance and resistivity values ​​at time t, respectively; A is the lateral cross-sectional area of ​​the self-sensing cement-based sensor; and l is the distance between the two electrodes.

[0052] Sensitivity coefficient calculation:

[0053] (2);

[0054] In the formula, σ is the stress, and SSC is the stress sensitivity coefficient;

[0055] (3);

[0056] In the formula, ε is the strain and GF is the strain sensitivity coefficient.

[0057] As a preferred technical solution of the present invention: in step S32, the loading stress amplitude range is 0.5 to 10 MPa, the loading rate is controlled at 1 kN / s, and the number of complete loading and unloading cycles is 5.

[0058] As a preferred technical solution of the present invention, step S4 is specifically as follows:

[0059] S41. Sensor installation:

[0060] A two-stage pouring process is used to embed the calibrated self-sensing cement-based sensor into the road surface concrete. Non-metallic binding straps and external supports are used to fix the self-sensing cement-based sensor in the center of the road surface mold. The long axis of the self-sensing cement-based sensor is perpendicular to the driving direction of the road surface. The wire is led out from the side wall of the road surface mold. Then, the concrete is poured to half the height of the road surface mold. After being vibrated and compacted, it is left to stand for 30 minutes depending on the ambient temperature. When the lower layer of concrete is in the early stage of initial setting, the self-sensing cement-based sensor is placed in the center of the road surface. Concrete is poured to fill the mold and a second vibration is performed to ensure that the self-sensing cement-based sensor and the surrounding road surface concrete are formed and cured synchronously. This ensures that the interface between the self-sensing cement-based sensor and the road surface substrate is tightly bonded and the deformation is coordinated. The wire lead-out end is sealed and protected.

[0061] S42. Sensor Connection and Data Acquisition:

[0062] The self-sensing cement-based sensor buried in the road surface is connected to the data acquisition system via an external wire. A DC regulated power supply is used to provide a stable and constant current for the test circuit. The data acquisition system is controlled to collect the resistivity of the self-sensing cement-based sensor at a sampling frequency of 1Hz to 10Hz.

[0063] As a preferred technical solution of the present invention, step S5 is as follows:

[0064] S51. Establishment of road surface health benchmarks:

[0065] First, the self-sensing cement-based sensors were calibrated, numbered, and deployed at corresponding monitoring points on the cement concrete pavement. Then, the initial normal state under no vehicle load was selected as the health baseline, and the initial resistivity of each self-sensing cement-based sensor was collected. ,in, , This represents the total number of self-sensing cement-based sensors.

[0066] S52, Data Smoothing Processing:

[0067] The resistivity of the self-sensing cement-based sensor is collected once at preset time intervals, and the monitoring data is smoothed using a moving average method. The sliding window length is set as follows:

[0068] (4);

[0069] In the formula, The smoothed resistivity characteristic value after k-period moving average processing is obtained when the i-th self-sensing cement-based sensor performs the j-th monitoring. Let be the original measured resistivity value of the i-th self-sensing cement-based sensor during the j-th monitoring; k is the sliding window length, where the k nearest consecutive data are averaged each time; j is the monitoring number, where k and j are both positive integers and satisfy the constraint k≤j, meaning that at least k data must be accumulated before the smoothing value can be calculated.

[0070] S53. Based on the above-mentioned full-cycle monitoring data after smoothing and noise reduction, construct a structural state monitoring data sample matrix based on self-sensing cement-based sensors:

[0071] (5);

[0072] In the formula, A smoothed resistivity sample matrix for the entire monitoring period. Let be the smooth resistivity characteristic value of the i-th self-sensing cement-based sensor during the j-th monitoring period; i = 1, 2, ..., m, where m is the self-sensing cement-based sensor; j = 1, 2, ..., n, where n is the total number of monitoring times.

[0073] As a preferred technical solution of the present invention, step S6 is specifically as follows:

[0074] Based on the development pattern of the resistivity data collected in step S5, four types of working states of the road surface are defined:

[0075] Health status: Resilient stage, corresponding to SVM label 0;

[0076] Minor damage state: microcrack initiation stage, corresponding to SVM label 1;

[0077] Moderate damage state: stable crack propagation stage, corresponding to SVM label 2;

[0078] Severe damage or destruction state: the unstable propagation stage of cracks, corresponding to SVM label 3.

[0079] As a preferred technical solution of the present invention, step S7 is as follows:

[0080] S71. Feature Extraction and Label Definition:

[0081] The smoothed resistivity sequence obtained for each monitoring cycle of the i-th self-sensing cement-based sensor That is, the first digit of matrix Z Line, extract the following time-domain and statistical features to construct a feature vector. :

[0082] Mean:

[0083] (6);

[0084] Standard deviation:

[0085] (7);

[0086] Maximum fluctuation range:

[0087] (8);

[0088] First-order difference absolute sum:

[0089] (9);

[0090] Mean rate of change of resistivity:

[0091] (10);

[0092] in, Let be the initial health reference resistivity of the i-th self-sensing cement-based sensor;

[0093] Z-score normalization is applied to the initial feature vector f iThe input feature vector F is obtained by processing it to eliminate the influence of dimensions. i ;

[0094] S72. Historical Sample Construction and Labeling:

[0095] Historical monitoring data of the road surface at different service stages were collected, and combined with the actual surface inspection, non-destructive testing, or mechanical loading test results, the actual physical damage state of the monitoring location was determined and classified into four categories, denoted as F for each standardized feature sample. i Label :

[0096] 0 – Health status;

[0097] 1 – Minor injury;

[0098] 2 – Moderate damage;

[0099] 3 – Severe damage or destruction;

[0100] This allows us to construct a historical sample set for SVM model training. ;

[0101] Construction of S73 and SVM models:

[0102] A one-to-one strategy was adopted to construct a multi-class SVM model. Since the SVM model is essentially a binary classification algorithm, a binary classifier was trained between any two classes of samples in the four types of road damage states, and a total of 6 binary sub-models were constructed.

[0103] In SVM model construction, slack variables are introduced. The penalty parameter C is used to allow a small number of samples to deviate from the classification hyperplane. At the same time, the radial basis kernel function is selected to map the extracted 5-dimensional low-dimensional feature vector to the high-dimensional feature space to process the nonlinear features of the road damage data, so that the originally mixed healthy and different damage state samples can be effectively separated by the hyperplane in the high-dimensional space. The radial basis kernel function includes the width parameter γ, i.e. the model sensitivity, which is used to control the fitting accuracy and generalization ability of the SVM model.

[0104] S74. Model Training and Parameter Optimization:

[0105] Arrange the historical sample set constructed in step S72 in chronological order The data from the first 70% of the monitoring period was divided into the training set, and the data from the last 30% was divided into the test set.

[0106] The training set is input into the multi-class SVM model constructed in step S73. Time-series five-fold cross-validation combined with grid search is used to optimize the penalty parameter C and the kernel parameter γ. The parameter search range is:

[0107] (11);

[0108] Choose the parameter combination that yields the highest average classification accuracy in cross-validation. The model was retrained on the entire training set to establish the final road surface damage recognition model.

[0109] After training, the generalization ability of the SVM model is evaluated using 30% of the test set, and the following metrics are calculated:

[0110] The misclassification distribution of adjacent damage states is analyzed using a confusion matrix.

[0111] Calculate the overall accuracy to evaluate global recognition performance.

[0112] Calculate the accuracy rate to characterize the system's false alarm control capability.

[0113] The focus is on calculating the recall rate under severe damage conditions to characterize the system's false negative control capability.

[0114] Calculate the F1 score for comprehensive performance evaluation;

[0115] Construct a multidimensional joint evaluation threshold. When the SVM model meets the joint evaluation threshold, deploy the SVM model to the online monitoring system. If it does not meet the threshold, increase the number of historical samples or adjust the feature extraction strategy and retrain.

[0116] S75. Real-time damage identification and early warning:

[0117] During the online monitoring phase, the system acquires resistivity signals from the self-sensing cement-based sensor in real time. It calculates the smoothed resistivity sequence for the current monitoring period using the moving average method in step S5, and extracts the feature vector for the current period of the i-th self-sensing cement-based sensor using the method in step S71, performing the same standardization process to obtain the test feature vector F. test,i , i=1,2,...,m;

[0118] F test,i Take the trained SVM model as an independent input and output the current damage category prediction result for the i-th monitoring location:

[0119] In the formula, Let i be the damage category at the i-th monitoring location;

[0120] The system dynamically updates the pavement health status at the corresponding monitoring location based on the damage category output by the SVM model, and executes a graded early warning strategy:

[0121] When the identification result is 1, the system marks the location as a potential degradation zone, automatically increases the recording frequency of background data, and performs routine tracking.

[0122] When the identification result evolves into 2, the system generates a moderate warning, prompting the maintenance department to pay closer attention and arrange preventative inspections.

[0123] A time-sliding window mechanism is introduced to prevent false alarms: The system triggers a severe warning only when the identification result of a certain monitoring location is consistently 3 for N consecutive monitoring cycles. After the severe warning is triggered, the system automatically pushes the specific damage location coordinates, evolution trend and early warning information to the monitoring terminal, prompting the need for substantive engineering maintenance. After the on-site maintenance is completed, the system resets the status of the monitoring node and enters a new round of periodic monitoring.

[0124] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0125] 1. Good compatibility: The self-sensing cement-based sensor described in this invention is made of the same cement-based material as the road surface substrate, which has natural mechanical compatibility. This avoids stress concentration, interface separation and signal drift caused by material incompatibility, and ensures the stability of long-term monitoring.

[0126] 2. Balancing Mechanical and Electrical Properties: This invention optimizes the ratio of low-dosage multi-walled carbon nanotubes (MWCNTs) and nano-carbon black (NCB), significantly improving electrical properties while maintaining good mechanical properties. After 28 days of curing, the material's flexural strength and compressive strength reached a maximum of 11.2 MPa and 69.5 MPa, respectively, while the resistivity significantly decreased to 1148.6 Ω·cm, completely resolving the common "mechanical-electrical contradiction" in self-sensing cement-based materials.

[0127] 3. This invention constructs a complete monitoring system from sensor pre-calibration, on-site installation and real-time data acquisition, to data preprocessing, SVM intelligent damage identification, and graded early warning. It can not only monitor the mechanical response of the road surface under traffic load in real time, but also accurately identify the entire cycle of damage evolution of the road structure from initial damage to failure, providing reliable technical support for the whole life cycle health management and preventive maintenance of cement concrete pavement.

[0128] 4. Economic benefits: This invention utilizes the synergistic effect of nano-carbon black NCB and multi-walled carbon nanotubes (MWCNTs) to achieve dual optimization of cost reduction and performance improvement. Compared with traditional high-cost sensors (such as fiber optic sensors) and graphite composite systems, it has better economic benefits and is suitable for the deployment of road infrastructure. Attached Figure Description

[0129] Figure 1 This is a flowchart illustrating the overall structure of the present invention;

[0130] Figure 2 This is a flowchart illustrating the fabrication process of the self-sensing cement-based sensor of the present invention.

[0131] Figure 3 The results of the mechanical comparison between single-doped multi-walled carbon nanotubes (MWCNTs) and composite systems in the self-sensing cement-based materials of this invention are as follows. Figure I ;

[0132] Figure 4 The results of the mechanical comparison between single-doped multi-walled carbon nanotubes (MWCNTs) and composite systems in the self-sensing cement-based materials of this invention are as follows. Figure II ;

[0133] Figure 5 The results of the mechanical comparison between the self-sensing cement-based material with single-doped nano-carbon black NCB and the composite system are as follows: Figure I ;

[0134] Figure 6 The results of the mechanical comparison between the self-sensing cement-based material with single-doped nano-carbon black NCB and the composite system are as follows: Figure II ;

[0135] Figure 7 The results of the electrical comparison between single-doped multi-walled carbon nanotubes (MWCNTs) and composite systems in the self-sensing cement-based material of this invention are as follows. Figure I ;

[0136] Figure 8 The results of the electrical comparison between the single-doped nano-carbon black NCB and the composite system of the self-sensing cement-based material of this invention are as follows. Figure II ;

[0137] Figure 9 This invention illustrates the influence of the longitudinal spacing of the sensors on the relative strain difference between the road surface and the sensors.

[0138] Figure 10 This is a partial schematic diagram of the pavement mechanical response monitoring system of the present invention.

[0139] Figure 11 This is a schematic diagram (global) of the pavement mechanical response monitoring system of the present invention.

[0140] Figure 12 This is a schematic diagram of monitoring data for the entire process of road surface compressive stress loading and unloading;

[0141] Figure 13 This is a feature map for judging signal damage in the platform segment based on resistivity data collected by the sensor.

[0142] Figure 14 This is a damage discrimination feature map based on the electrical signal surge of resistivity data collected by the sensor;

[0143] Figure 15 This is a damage discrimination feature map based on the electrical signal drop in resistivity data collected by the sensor;

[0144] Figure 16 This is a flowchart of SVM intelligent damage identification and hierarchical early warning. Detailed Implementation

[0145] The present invention will be further illustrated below with reference to the accompanying drawings and specific embodiments. It should be understood that the following specific embodiments are for illustrative purposes only and are not intended to limit the scope of the invention.

[0146] like Figure 1 As shown, the present invention proposes a method for monitoring the response of an embedded self-sensing cement-based sensor on a cement concrete pavement, comprising the following steps:

[0147] S1. Preparation and performance testing of self-sensing cement-based sensor;

[0148] S2, Embedded sensor packaging process;

[0149] S3. Pre-calibration of sensor piezoresistive performance;

[0150] S4. Sensor installation and monitoring on road surface;

[0151] S5. Monitoring data preprocessing;

[0152] S6. Methods for identifying road surface damage;

[0153] S7 and SVM intelligent damage recognition and early warning.

[0154] The steps of this invention will be described in detail below.

[0155] S1. The fabrication and performance testing of the self-sensing cement-based sensor are detailed below:

[0156] S11. The fabrication of the self-sensing cement-based sensor is as follows:

[0157] S111. Determine the optimal composition:

[0158] It is a mixture of raw materials comprising the following parts by weight: 100 wt.% cement, 0.2 wt.%–0.5 wt.% multi-walled carbon nanotubes (MWCNTs), 0.2 wt.%–1.0 wt.% nano-carbon black (NCB), 0.1 wt.%–0.3 wt.% polycarboxylate superplasticizer, and 30 wt.%–45 wt.% water.

[0159] S112, Preparation of components:

[0160] Pre-dispersion treatment: Dissolve polycarboxylate superplasticizer in water and stir until completely dissolved to obtain a base solution. Add nano carbon black NCB to the obtained base solution and stir for 3 minutes. The nano carbon black NCB agglomerates are initially broken by mechanical shear force, so that the polycarboxylate superplasticizer is fully adsorbed on the surface of nano carbon black NCB to obtain a pre-dispersion liquid.

[0161] Secondary synergistic dispersion: Multi-walled carbon nanotubes (MWCNTs) are added to the above pre-dispersion liquid and stirred for 3 min; the polycarboxylate superplasticizer is adsorbed on the surface of MWCNTs through π-π stacking, and its hydrophilic branches extend outward to form a steric hindrance layer, which prevents the aggregation of MWCNTs and simultaneously achieves synergistic and stable dispersion of MWCNTs and NCB nanocarbon black, resulting in a mixed dispersion liquid;

[0162] Ultrasonic homogeneous dispersion: The above-mentioned mixed dispersion after secondary dispersion is placed in an ultrasonic disperser, the ultrasonic frequency is set to 25kHz~35kHz, and the ultrasonic treatment time is 15min~25min. The residual agglomerates of conductive fillers are further broken by ultrasonic cavitation effect, so that the two conductive fillers form a uniform and stable co-dispersion system and a conductive filler composite suspension is obtained.

[0163] Demolding and curing: The prepared conductive filler composite suspension is added to cement while being stirred at low speed for 2 minutes, and then stirred at high speed for 3 minutes to obtain a uniform cement slurry. The obtained cement slurry is poured into a preset mold for molding. After demolding, it is placed in a standard curing environment for curing until the specified age to obtain a self-sensing cement-based sensor.

[0164] S12. Performance testing of the self-sensing cement-based sensor, as detailed below:

[0165] Mechanical properties of the self-sensing cement-based sensor were tested according to the "Test Method for Strength of Cement Mortar". Resistivity of the self-sensing cement-based sensor was tested using the four-electrode method.

[0166] Step S1 requires extensive experimental studies on comparative examples (single-doped multi-walled carbon nanotubes (MWCNTs) and single-doped nano-carbon black (NCB)) and exemplary examples (multi-doped MWCNTs and NCB) to determine the optimal formulation and prepare the self-sensing cement-based sensor. The preparation method is as follows: Figure 2 As shown, based on the performance test results of the comparative and embodiment examples, the specific steps are detailed below:

[0167] Comparative Example 1 (Single-doped multi-walled carbon nanotubes (MWCNTs):

[0168] The comparative example is a cement-based material specimen with single-component multi-walled carbon nanotubes (MWCNTs). The composition ratio is as follows, based on 100 wt.% of cement: 100 wt.% cement, 0.2 wt.% multi-walled carbon nanotubes (MWCNTs), 0.1 wt.% polycarboxylate superplasticizer, and 40 wt.% water.

[0169] The preparation method is as follows:

[0170] (1) Dissolve all of the polycarboxylate superplasticizer in water and stir until completely dissolved. Add multi-walled carbon nanotubes (MWCNTs) and stir mechanically for 3 minutes to obtain a pre-dispersion.

[0171] (2) Place the pre-dispersed liquid in an ultrasonic disperser, set the ultrasonic frequency to 30kHz and the power density to 150W / L, and ultrasonically treat for 20min to obtain a multi-walled carbon nanotube (MWCNT) dispersion.

[0172] (3) Add all of the multi-walled carbon nanotube (MWCNT) dispersion to the cement, stir at low speed for 2 minutes, then stir at high speed for 3 minutes to obtain a uniform cement paste.

[0173] (4) Pour the cement paste into a 40mm×40mm×160mm triple mold, vibrate to compact it, and then mold, demold and cure according to general test conditions.

[0174] Comparative Example 2 (single-doped nano carbon black NCB):

[0175] The comparative example is a cement-based material specimen with single-component nano-carbon black NCB. The composition ratio is as follows, based on 100 wt.% of cement: 100 wt.% cement, 0.2 wt.% nano-carbon black NCB, 0.1 wt.% polycarboxylate superplasticizer, and 40 wt.% water.

[0176] The preparation method is as follows:

[0177] (1) Dissolve all of the polycarboxylate superplasticizer in water and stir until completely dissolved. Add nano carbon black NCB and stir mechanically for 3 minutes to obtain a pre-dispersion.

[0178] (2) Place the pre-dispersed liquid in an ultrasonic disperser, set the ultrasonic frequency to 30kHz and the power density to 150W / L, and ultrasonically treat for 20min to obtain a nano carbon black NCB dispersion.

[0179] (3) Add all of the nano carbon black NCB dispersion to the cement, stir at low speed for 2 minutes, then stir at high speed for 3 minutes to obtain a uniform cement paste.

[0180] (4) Pour the cement paste into a 40mm×40mm×160mm triple mold, vibrate to compact it, and then mold, demold and cure according to general test conditions.

[0181] Example 1:

[0182] This embodiment is a self-sensing cement-based material specimen prepared by a binary compound of multi-walled carbon nanotubes (MWCNTs) and nano-carbon black (NCB). Based on 100 wt.% of cement, the component ratio is: cement 100 wt.%, MWCNTs 0.2 wt.%, NCB 0.2 wt.%, polycarboxylate superplasticizer 0.1 wt.%, and water 40 wt.%. The preparation method strictly follows the staged synergistic dispersion process described in this invention, with the specific steps as follows:

[0183] (1) Pre-dispersion treatment: Dissolve all of the polycarboxylate superplasticizer in water and stir until completely dissolved. Add nano carbon black NCB and mechanically stir for 3 minutes. The nano carbon black NCB agglomerates are initially broken by mechanical shear force, so that the superplasticizer is fully adsorbed on the surface of nano carbon black NCB to obtain a pre-dispersion liquid.

[0184] (2) Secondary synergistic dispersion: Multi-walled carbon nanotubes (MWCNTs) are added to the pre-dispersed liquid and mechanically stirred for 3 min to allow the water-reducing agent to be adsorbed onto the surface of the MWCNTs, forming a steric hindrance effect to prevent the filler from agglomerating and to obtain a mixed dispersion system.

[0185] (3) Ultrasonic homogeneous dispersion: The mixed dispersion system is placed in an ultrasonic disperser, the ultrasonic frequency is set to 30kHz and the power density is 150W / L, and the ultrasonic treatment is carried out for 20min. The residual agglomerates are broken by the ultrasonic cavitation effect to obtain a uniform and stable conductive filler composite suspension.

[0186] (4) Mixing, molding and curing: Add all the conductive filler composite suspension to the cement, stir at low speed for 2 minutes, then stir at high speed for 3 minutes to obtain uniform cement paste; pour the paste into a 40mm×40mm×160mm triple mold, vibrate to compact, and then mold, demold and cure according to general test conditions.

[0187] Example 2:

[0188] The composition ratio of this embodiment is: 100 wt.% cement, 0.2 wt.% multi-walled carbon nanotubes (MWCNTs), 0.5 wt.% nano-carbon black (NCB), 0.1 wt.% polycarboxylate superplasticizer, and 40 wt.% water. The preparation method is completely consistent with that of Example 1.

[0189] Example 3:

[0190] The composition ratio of this embodiment is: 100 wt.% cement, 0.2 wt.% multi-walled carbon nanotubes (MWCNTs), 0.8 wt.% nano-carbon black (NCB), 0.1 wt.% polycarboxylate superplasticizer, and 40 wt.% water. The preparation method is completely consistent with that of Example 1.

[0191] Example 4:

[0192] The composition ratio of this embodiment is: 100 wt.% cement, 0.2 wt.% multi-walled carbon nanotubes (MWCNTs), 1.0 wt.% nano-carbon black (NCB), 0.1 wt.% polycarboxylate superplasticizer, and 40 wt.% water. The preparation method is completely consistent with that of Example 1.

[0193] Example 5:

[0194] The composition ratio of this embodiment is: 100 wt.% cement, 0.2 wt.% multi-walled carbon nanotubes (MWCNTs), 0.2 wt.% nano-carbon black (NCB), 0.1 wt.% polycarboxylate superplasticizer, and 40 wt.% water. The preparation method is completely consistent with that of Example 1.

[0195] Example 6:

[0196] The composition ratio of this embodiment is: 100 wt.% cement, 0.3 wt.% multi-walled carbon nanotubes (MWCNTs), 0.2 wt.% nano-carbon black (NCB), 0.1 wt.% polycarboxylate superplasticizer, and 40 wt.% water. The preparation method is completely consistent with that of Example 1.

[0197] Example 7:

[0198] The composition ratio of this embodiment is: 100 wt.% cement, 0.4 wt.% multi-walled carbon nanotubes (MWCNTs), 0.2 wt.% nano-carbon black (NCB), 0.1 wt.% polycarboxylate superplasticizer, and 40 wt.% water. The preparation method is completely consistent with that of Example 1.

[0199] Example 8:

[0200] The composition ratio of this embodiment is: 100 wt.% cement, 0.5 wt.% multi-walled carbon nanotubes (MWCNTs), 0.2 wt.% nano-carbon black (NCB), 0.1 wt.% polycarboxylate superplasticizer, and 40 wt.% water. The preparation method is completely consistent with that of Example 1.

[0201] All the above embodiments and comparative examples are based on cement mass (100wt.%), the water-reducing agent is polycarboxylate water-reducing agent, the dosage is fixed at 0.1wt.%, the water-cement ratio is 0.4, the specimen size is 40×40×160mm, the curing and testing conditions are uniform, and the comparability and reliability of the test data are ensured.

[0202] Performance testing:

[0203] This invention conducts mechanical property tests according to the "Test Method for Strength of Cement Mortar" (GB / T 17671-2021). Some mechanical property test results are as follows: Figure 3-6 As shown in Table 1, Figure 3 and Figure 4The mechanical property test results are for Comparative Example 1 and some of the mixed-component systems; Figure 5 and Figure 6 Table 1 shows the mechanical property test results of some specimens of the present invention, which are compared with those of Example 2 and some mixed-component systems.

[0204]

[0205] Table 1

[0206] Analysis of mechanical performance test results:

[0207] Mechanical property improvement: The flexural strength and compressive strength of the specimens in each embodiment of the present invention at a standard curing age of 28 days were significantly higher than those of Comparative Example 1, which was doped with multi-walled carbon nanotubes (MWCNTs), and Comparative Example 2, which was doped with nano-carbon black (NCB), and no mechanical property degradation was observed after 28 days. This indicates that the MWCNTs and NCB co-doping system can effectively improve the problems of easy agglomeration, easy formation of internal matrix defects, and limited mechanical reinforcement effect of single conductive fillers. At 28 days, Example 2, which had the best mechanical properties, achieved a flexural strength of 11.2 MPa, which was 62.3% higher than that of Comparative Example 1 (6.9 MPa) and 75.0% higher than that of Comparative Example 2 (6.4 MPa); the compressive strength reached 69.5 MPa, which was 47.6% higher than that of Comparative Example 1 (47.1 MPa) and 85.8% higher than that of Comparative Example 2 (37.4 MPa). Even taking Example 8, which has a relatively low 28-day compressive strength, as an example, its 28-day flexural strength still reaches 8.6 MPa and its compressive strength reaches 55.3 MPa, which are 24.6% and 17.4% higher than Comparative Example 1, and 34.4% and 47.9% higher than Comparative Example 2, respectively. This shows that the composite system of the present invention has clear mechanical performance advantages under different ratio conditions.

[0208] Synergistic Enhancement Mechanism Verification: Compared to comparative examples of single-doped multi-walled carbon nanotubes (MWCNTs) or single-doped nano-carbon black (NCB), the significant improvement in mechanical properties of the embodiments of this invention is due to two main factors. Firstly, MWCNTs can form a microscale bridging network between cement hydration products, acting as a bridge between cracks and inhibiting microcrack propagation during stress, thereby improving the flexural strength of the material. Secondly, NCB can fill the pores in the cement matrix and the gaps between MWCNTs, improving the particle packing structure, increasing matrix density, and reducing interfacial defects caused by carbon nanotube agglomeration. The combined effect of these two factors makes the internal structure of the cement-based material more compact and restricts crack propagation paths, ultimately achieving a simultaneous increase in flexural and compressive strength.

[0209] The resistivity of the self-sensing cement-based sensor of this invention is measured using the four-electrode method. Some electrical performance test results are as follows: Figures 7-8 As shown in Table 2, Figure 7 The results show the electrical performance test results of Comparative Example 1 and some of the mixed-doped systems; Figure 8 Table 2 shows the electrical performance test results of Comparative Example 2 and some of the composite doped systems; Table 2 shows the electrical performance test results of some specimens of this invention.

[0210]

[0211] Table 2

[0212] Analysis of electrical performance test results:

[0213] Electrical performance test results: The resistivity of Comparative Example 1, which is doped with 0.2 wt.% multi-walled carbon nanotubes (MWCNTs), is 5623.1 Ω·cm, while the resistivity of Comparative Example 2, which is doped with 0.2 wt.% nano-carbon black (NCB), is 20614.3 Ω·cm. This indicates that although the single conductive filler system has a certain conductive modification effect, it is still difficult to form a stable and continuous conductive network. In particular, the resistivity is relatively high when doped with NCB, and it has not yet shown a significant percolation conductivity effect. In contrast, the resistivity of the self-sensing cement-based sensor in each embodiment of the present invention is consistently in the range of 1148.6–1613.5 Ω·cm, which is about 71.3%–79.6% lower than that of Comparative Example 1 and about 92.2%–94.4% lower than that of Comparative Example 2, indicating a significant improvement in conductivity. Among them, Example 8 had the lowest resistivity, at 1148.6 Ω·cm, which was 79.6% lower than Comparative Example 1 and 94.4% lower than Comparative Example 2. This indicates that the multi-walled carbon nanotubes (MWCNTs) and nano-carbon black (NCB) can effectively construct a synergistic conductive network, significantly reducing the resistivity of cement-based materials. This lays a good electrical performance foundation for the materials to achieve stable electromechanical response, deformation sensing, and damage monitoring.

[0214] Validation of the synergistic effect mechanism: Comparative Examples 1 and 2 used single-doped multi-walled carbon nanotubes (MWCNTs) and single-doped nano-carbon black (NCB) systems, respectively. Their resistivities were 5623.1 Ω·cm and 20614.3 Ω·cm, respectively, remaining at relatively high levels. This indicates that a single conductive filler is unlikely to form a sufficiently continuous conductive network in the cement matrix under low dosage conditions. In contrast, Examples 1-8 all used a combined MWCNT and NCB system, with resistivity steadily decreasing to 1148.6–1613.5 Ω·cm, significantly lower than the two single-doped comparative examples, indicating a significant synergistic conductive effect between the two conductive fillers. The synergistic effect is mainly manifested in the following ways: one-dimensional multi-walled carbon nanotubes (MWCNTs) construct a long-range conductive framework in the cement matrix, while zero-dimensional nano-carbon black (NCB) is distributed between the carbon nanotubes and in the pores of the cement matrix, filling voids, connecting breakpoints, and increasing conductive contact points. Simultaneously, the pre-dispersion, secondary synergistic dispersion, and ultrasonic homogenization processes effectively reduce the agglomeration of conductive fillers, making the "point-line" composite conductive network more uniform, continuous, and stable. Therefore, the composite system of this invention significantly reduces the resistivity of cement-based materials, achieving superior conductivity compared to single-doped systems, providing crucial support for self-sensing cement-based sensors to obtain stable electromechanical responses and long-term monitoring capabilities.

[0215] In summary, this invention addresses the shortcomings of existing technologies where single-doped multi-walled carbon nanotubes (MWCNTs) and nano-carbon black (NCB) tend to agglomerate, have limited modification effects, and struggle to balance mechanical load-bearing capacity and conductive self-sensing properties. It achieves efficient and stable dispersion and synergistic effects of both NCB and MWCNTs in a cement matrix. The resulting self-sensing cement-based sensor possesses excellent mechanical load-bearing capacity and stable conductive self-sensing properties. Furthermore, the preparation method is simple, highly adaptable, requires no complex production equipment, and can be scaled up for widespread application. This self-sensing cement-based sensor is widely applicable to long-term health monitoring scenarios for cement concrete pavements, demonstrating significant engineering application value and market potential.

[0216] Step S2: Encapsulate the sensor prepared in step S1 to improve its waterproof performance and compatibility with cement concrete pavement. The specific steps are as follows:

[0217] Step S21, Pretreatment and Electrode Setup:

[0218] After 28 days of curing, the self-sensing cement-based sensor was made into a 40mm×40mm×160mm specimen. The four-electrode method was used, and the wires were soldered to the specimen with a soldering iron. The solder joints and the external parts of the wires were wrapped with insulating tape.

[0219] Step S22, Surface modification treatment:

[0220] The specimens were sandblasted using a compressed air sandblasting machine at a pressure of 0.6 MPa and a blasting distance of 10 cm. The quartz sand particle size was 60-100 mesh, resulting in a surface roughness Ra of 2.0 μm-3.5 μm. After sandblasting, the surface was cleaned with compressed air and allowed to dry.

[0221] Step S23, Waterproofing treatment:

[0222] After sandblasting, the specimens were encapsulated with epoxy resin for waterproofing to isolate the sensor's internal conductive network from the corrosive media such as rainwater and de-icing salt in the road service environment, ensuring signal stability during long-term monitoring. Peelable masking tape was used to precisely mask the electrode contact points and wire leads at both ends of the specimen. Then, a two-component epoxy resin coating was uniformly applied to the entire surface of the specimen, with the dry film thickness controlled at 0.2 mm. After coating, the specimens were left to stand at room temperature until the epoxy resin coating initially set. The masking tape was then removed to ensure complete exposure of the electrode contact points and normal wire lead-out. Once the epoxy resin coating was fully cured, the waterproof encapsulation of the sensor was completed.

[0223] Step S3: Pre-calibration of sensor piezoresistive performance:

[0224] Monotonic loading and graded cyclic loading calibration tests were conducted on the packaged sensors to establish a quantitative mapping relationship between resistivity change rate, stress, and strain specific to each sensor. The steps are as follows:

[0225] Step S31, Elimination of polarization effect:

[0226] The sensor is connected to the test circuit using the four-electrode method. A constant test current is provided by a DC regulated power supply. After the sensor is powered on and left to stand for 3 minutes, the initial resistance value and initial resistivity are recorded to eliminate the interference of the electrode polarization effect of the material on the test results.

[0227] Step S32, Loading System:

[0228] The sensor was loaded using a stress control mode, with the loading stress amplitude ranging from 0.5 to 10 MPa (referencing the tire ground contact stress of 0.7 MPa in the "Highway Engineering Technical Standard" JTG B01-2014). The loading rate was controlled at 1 kN / s, and the number of complete loading and unloading cycles was 5. Real-time resistance signals, stress values, and strain values ​​of the sensor under different load levels were collected simultaneously.

[0229] Step S33: Feature parameter calculation and calibration model establishment:

[0230] Based on the collected resistivity, stress, and strain data, the influence of contact resistance and wire resistance is eliminated by the four-electrode method. The following core characteristic parameters are calculated to establish a quantitative piezoresistive model of resistivity change rate-stress-strain for each sensor.

[0231] The change in resistivity is quantitatively characterized by the rate of change of resistivity (FCR):

[0232] (1);

[0233] R0 and These represent the initial resistance and initial resistivity of the sensor specimen after it has been powered on for a period of time, respectively, to avoid the influence of polarization effects; Rt and Here, t represents the resistance and resistivity values ​​at time t, respectively; A is the transverse cross-sectional area of ​​the sensor specimen; and l is the distance between the two electrodes.

[0234] Sensitivity coefficient calculation:

[0235] (2);

[0236] In the formula, σ is the stress. This is the stress sensitivity coefficient.

[0237] (3);

[0238] In the formula, ε is the strain and GF is the strain sensitivity coefficient.

[0239] Step S4: Sensor installation and monitoring on the road surface.

[0240] The sensors calibrated in step S3 are then installed on the road surface to be monitored, with the longitudinal spacing of the sensors determined according to... Figure 9 The relative difference curves of the road surface and the sensor strain are comprehensively determined (considering the thickness of the road surface layer, the sensor only studies the longitudinal spacing), and must not be less than 35mm; the sensor is connected to the data acquisition system through an external wire to monitor the road surface.

[0241] Figure 10 This is a partial schematic diagram of the pavement mechanical response monitoring system of the present invention. Figure 11 This is a schematic diagram (global) of the pavement mechanical response monitoring system of the present invention. Figure 12 This is a schematic diagram illustrating the monitoring data throughout the entire process of force loading and unloading. The steps are as follows:

[0242] Step S41, Sensor installation:

[0243] The calibrated sensor was embedded in the road concrete using a two-stage pouring process. The sensor was fixed to the center of the road mold using non-metallic binding straps and external supports. The long axis of the sensor was perpendicular to the road traffic direction, and the wires were led out from the side wall of the mold.

[0244] The first concrete pour is poured to half the height of the mold. After compaction with a vibrating device, it is left to stand for 30 minutes depending on the ambient temperature. When the lower layer of concrete is in the early stage of initial setting, the sensor is precisely placed in the center of the road surface. Concrete is poured to fill the mold, and a second vibration is performed to ensure that the sensor and the surrounding road surface concrete are formed and cured synchronously. This ensures that the interface between the sensor and the road surface substrate is tightly bonded and the deformation is coordinated. The lead-out end of the wire is sealed and protected.

[0245] Step S42, Sensor Connection and Data Acquisition:

[0246] The sensor buried in the road surface is connected to the data acquisition system by an external wire. A DC regulated power supply is used to provide a stable and constant current for the test circuit. The data acquisition system is controlled to collect the resistivity of the sensor at a sampling frequency of 1Hz to 10Hz.

[0247] Step S5: Monitoring Data Preprocessing: The resistivity data collected in step S4 is preprocessed to eliminate noise interference and provide reliable input for subsequent damage identification. The specific steps are as follows:

[0248] Step S51, Establishing Road Surface Health Benchmarks:

[0249] Sensors were calibrated and numbered (1, 2, 3....m). The sensors were deployed at corresponding monitoring points on the cement concrete pavement; the initial normal state under no vehicle load was selected as the health baseline, and the initial resistivity of each sensor was collected. ).

[0250] Step S52, Data Smoothing Processing:

[0251] For engineering structures in service, the load conditions they experience are very complex. Long-term monitoring of the structural health involves collecting resistivity data at regular intervals. To avoid interference from data fluctuations and to make the long-term monitoring trend more apparent, a moving average method is used to smooth the monitoring data, and a sliding window length (i.e., the number of periodic monitoring cycles, k) is set.

[0252] (4);

[0253] In the formula, The smoothed resistivity characteristic value of the i-th sensor after k-period moving average processing during the j-th monitoring; is the original measured resistivity value of the i-th sensor during the j-th monitoring; k is the sliding window length, which takes the k nearest consecutive data points for averaging each time; j is the monitoring number, and k and j are both positive integers, satisfying the constraint k≤j, that is, at least k data points must be accumulated before the smoothing value can be calculated.

[0254] Step S53: Based on the above-mentioned full-cycle monitoring data after smoothing and noise reduction, construct a structural state monitoring data sample matrix based on self-sensing cement-based sensors.

[0255] (5);

[0256] In the formula, A smoothed resistivity sample matrix for the entire monitoring period. Let be the smooth resistivity characteristic value of the i-th sensor in the j-th monitoring period; (i=1,2,…,m, where m is the total number of sensors; j=1,2,…,n, where n is the total number of monitoring times).

[0257] Step S6, Damage Assessment Method:

[0258] The development pattern based on resistivity data collected by sensors, such as Figure 13-15 As shown, four working states of the road surface are defined:

[0259] Healthy state (elastic stage, corresponding to SVM label 0): In the initial stage of load application, the applied load amplitude is small, no significant deformation or cracks occur inside the pavement concrete, the triaxial stress state of the sensor is stable, and the conductive network structure is intact and undamaged; at this time, the resistivity of the sensor decreases linearly with the increase of load, and the resistivity change rate FCR is significantly linearly correlated with the stress σ, which is the healthy working state of the pavement structure.

[0260] Minor damage (microcrack initiation stage, corresponding to SVM label 1): As the load amplitude or the number of load cycles increases, microcracks begin to form inside the pavement. The local stress concentration caused by the cracks leads to a slight increase in the tunneling distance between conductive particles inside the sensor, but the conductive network does not break. At this time, the rate of decrease in sensor resistivity slows down slightly, the slope of the FCR-σ curve decreases slightly, and the standard deviation of the monitoring data increases. First-order difference absolute value sum It shows a slow upward trend, indicating minor damage to the road surface structure.

[0261] Moderate damage (stable crack propagation stage, corresponding to SVM label 2): ​​As pavement damage further develops, the initial internal cracks gradually connect and propagate stably, the crack width and length continue to increase, and the conductive network inside the sensor experiences local breakage; at this time, the rate of decrease in sensor resistivity slows down significantly, the nonlinear characteristics of the FCR-σ curve become prominent, the stress sensitivity coefficient SSC shows a significant decay, and the maximum fluctuation amplitude and the average resistivity change rate of the monitoring data increase significantly, which is the moderate damage state of the pavement structure.

[0262] Severe damage or failure (crack instability propagation stage, corresponding to SVM label 3): When the stress level on the pavement exceeds 80% of its ultimate strength, or reaches the ultimate bearing strength, internal cracks in the concrete enter the instability propagation stage, and the cracks rapidly penetrate to form macroscopic cracks. When the FCR in the monitoring data exhibits one of the following three characteristics, it is determined that the pavement structure has suffered severe damage or the sensor has failed: 1) The FCR response curve shows a stable "plateau segment," meaning the FCR no longer changes regularly with load variations; 2) Irreversible abrupt increase in resistivity; 3) Irreversible abrupt drop in resistivity.

[0263] Step S7: Intelligent Damage Identification and Graded Early Warning Based on Support Vector Machine (SVM): For long-term monitoring data, a monitoring data sample matrix is ​​constructed in step S5. Building upon this foundation, the Support Vector Machine (SVM) algorithm is further introduced to achieve automatic and high-precision identification of the damage state of cement concrete pavement structures and output four states: healthy, slightly damaged, moderately damaged, and severely damaged. An early warning is triggered when moderate or severe damage is detected. Based on the principle of structural risk minimization, SVM exhibits excellent performance in small-sample, nonlinear classification problems, such as... Figure 16 As shown.

[0264] The specific steps are as follows:

[0265] Step S71, Feature Extraction and Label Definition:

[0266] The smoothed resistivity sequence obtained for each monitoring period of the i-th sensor , (i.e., the first digit of matrix Z) (Line), extract the following time-domain and statistical features to construct a feature vector. (In this invention, f is used to refer to the input feature vector for road surface damage identification):

[0267] Mean:

[0268] (6);

[0269] Standard deviation:

[0270] (7);

[0271] Maximum fluctuation range:

[0272] (8);

[0273] First-order difference absolute sum:

[0274] (9);

[0275] Mean rate of change of resistivity:

[0276] (10);

[0277] in, Let be the initial health reference resistivity of the i-th self-sensing cement-based sensor.

[0278] Since the dimensions and numerical ranges of each feature differ, Z-score standardization is used to normalize the initial feature vector f. i The input feature vector F is obtained by processing it to eliminate the influence of dimensions. i .

[0279] Step S72, Historical Sample Construction and Labeling:

[0280] Historical monitoring data of the pavement at different service stages were collected. Combined with the results of actual pavement surface inspection, non-destructive testing, or mechanical loading experiments, the actual physical damage state at the monitoring location was determined and classified into four categories, denoted as F for each standardized feature sample. i Label (In this invention, 'y' is used to uniformly refer to the classification result label of road surface health status.)

[0281] 0 – Healthy state (Resilient stage)

[0282] 1 – Minor damage (initiation of microcracks)

[0283] 2 – Moderate damage (stable crack propagation)

[0284] 3 – Severe damage or disruption (plateau segment, sudden increase / decrease in signal)

[0285] This leads to the construction of a historical sample set for model training.

[0286] Step S73, SVM model construction:

[0287] A one-to-one (OVO) strategy is used to construct a multi-class SVM model. Since SVM is essentially a binary classification algorithm, this step trains a binary classifier between any two classes of samples from the four types of road damage states, thus constructing a total of 6 binary sub-models.

[0288] In model construction, slack variables are introduced. A penalty parameter C is used to allow a small number of samples to deviate from the classification hyperplane, effectively solving the problem of linear inseparability of different damage state samples due to noise in actual road engineering monitoring data. Simultaneously, considering the high probability of overlapping feature samples of different damage states in the low-dimensional space, a radial basis function (RBF) is selected to map the extracted 5-dimensional low-dimensional feature vector to a high-dimensional feature space, processing the nonlinear characteristics of road damage data and enabling the originally overlapping healthy and different damage state samples to be effectively separated by the hyperplane in the high-dimensional space. The RBF kernel function includes a width parameter γ (i.e., model sensitivity) to control the model's fitting accuracy and generalization ability.

[0289] Step S74, Model Training and Parameter Optimization:

[0290] Considering the irreversible time-series nature of pavement damage evolution, and to prevent future data leakage into the training phase, the historical sample set constructed in step S7.2 is arranged in chronological order. The data from the first 70% of the monitoring period was divided into the training set, and the data from the last 30% was divided into the test set.

[0291] The training set is input into the multi-class SVM model constructed in step S7.3. Time series five-fold cross-validation combined with grid search is used to optimize the penalty parameter C and the kernel parameter γ. The parameter search range is:

[0292] (11);

[0293] This parameter range is an optimal range set for the characteristics of small sample engineering data such as road surface monitoring, which can effectively improve the optimization efficiency.

[0294] Choose the parameter combination that yields the highest average classification accuracy in cross-validation. The model was retrained on the entire training set to establish the final road surface damage recognition model.

[0295] After training, the model's generalization ability is evaluated using 30% of the test set. The following metrics are calculated: the misclassification distribution of adjacent damage states (such as healthy and minor damage) is analyzed using the confusion matrix; the overall accuracy is calculated to evaluate the global recognition performance; the precision is calculated to characterize the system's false alarm control capability (the higher the precision, the fewer false alarms); the recall rate for severe damage states is calculated to characterize the system's false negative control capability (the higher the recall rate, the fewer false negatives, and the higher the engineering safety); and the F1 score is calculated for comprehensive performance evaluation.

[0296] To ensure the safety of engineering applications, a multi-dimensional joint evaluation threshold is constructed (i.e., the overall accuracy must reach a basic threshold, and the recall rate for severely damaged states must reach a higher safety threshold). When the model meets this joint threshold condition, it is deployed to the online monitoring system; if it does not, the number of historical samples is increased or the feature extraction strategy is adjusted for retraining.

[0297] Step S75, Real-time Damage Identification and Early Warning:

[0298] During the online monitoring phase, the system acquires sensor resistivity signals in real time, calculates the smoothed resistivity sequence within the current monitoring period using the moving average method (Formula 4) in step S5, extracts the feature vector of the i-th sensor for the current period using the method in step S7.1, and performs the same standardization processing to obtain the test feature vector F. test,i (i=1,2...,m).

[0299] F test,i Take the trained SVM model as an independent input and output the current damage category prediction result for the i-th monitoring location:

[0300] In the formula, Let represent the damage category at the i-th monitoring location.

[0301] The system dynamically updates the pavement health status at the corresponding monitoring location based on the damage category output by the model, and executes a graded early warning strategy:

[0302] When the identification result is 1 (minor damage), the system marks the location as a potential degradation zone, automatically increases the recording frequency of background data, and performs routine tracking.

[0303] When the identification result evolves to 2 (moderate damage), the system generates a moderate warning, prompting the maintenance department to pay closer attention and arrange preventive inspections.

[0304] To avoid false alarms caused by occasional heavy vehicles or environmental noise, the system introduces a time-sliding window mechanism to prevent false alarms. A severe warning is triggered only if the identification result of a certain monitoring location is consistently output as 3 (severe damage) for N consecutive monitoring cycles (e.g., N=2 or 3).

[0305] Upon triggering a severe warning, the system automatically pushes the specific damage location coordinates, evolution trend, and early warning information to the monitoring terminal, prompting substantive engineering maintenance. After on-site maintenance is completed, the system resets the status of the monitoring node and begins a new round of periodic monitoring.

[0306] It should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention in any other way. Any modifications or equivalent changes made based on the technical essence of the present invention shall still fall within the scope of protection claimed by the present invention.

Claims

1. A method for monitoring the response of an embedded self-sensing cement-based sensor on a cement concrete pavement, characterized in that, Includes the following steps: S1. Preparation and performance testing of self-sensing cement-based sensor; S2, Embedded sensor packaging process; S3. Pre-calibration of sensor piezoresistive performance; S4. Sensor installation and monitoring on road surface; S5. Monitoring data preprocessing; S6. Methods for identifying road surface damage; S7 and SVM intelligent damage recognition and early warning.

2. The method for monitoring the response of an embedded self-sensing cement-based sensor on a cement concrete pavement according to claim 1, characterized in that, Step S1 is as follows: S11. The fabrication of the self-sensing cement-based sensor is as follows: S111. Determine the optimal composition: It is a mixture of raw materials comprising the following parts by weight: 100 wt.% cement, 0.2 wt.%–0.5 wt.% multi-walled carbon nanotubes (MWCNTs), 0.2 wt.%–1.0 wt.% nano-carbon black (NCB), 0.1 wt.%–0.3 wt.% polycarboxylate superplasticizer, and 30 wt.%–45 wt.% water. S112, Preparation of components: Pre-dispersion treatment: Dissolve polycarboxylate superplasticizer in water and stir until completely dissolved to obtain a base solution. Add nano carbon black NCB to the obtained base solution and stir for 3 minutes. The nano carbon black NCB agglomerates are initially broken by mechanical shear force, so that the polycarboxylate superplasticizer is fully adsorbed on the surface of nano carbon black NCB to obtain a pre-dispersion liquid. Secondary synergistic dispersion: Multi-walled carbon nanotubes (MWCNTs) are added to the above pre-dispersion liquid and stirred for 3 min; the polycarboxylate superplasticizer is adsorbed on the surface of MWCNTs through π-π stacking, and its hydrophilic branches extend outward to form a steric hindrance layer, which prevents the aggregation of MWCNTs and simultaneously achieves synergistic and stable dispersion of MWCNTs and NCB nanocarbon black, resulting in a mixed dispersion liquid; Ultrasonic homogeneous dispersion: The above-mentioned mixed dispersion after secondary dispersion is placed in an ultrasonic disperser, the ultrasonic frequency is set to 25kHz~35kHz, and the ultrasonic treatment time is 15min~25min. The residual agglomerates of conductive fillers are further broken by ultrasonic cavitation effect, so that the two conductive fillers form a uniform and stable co-dispersion system and a conductive filler composite suspension is obtained. Demolding and curing: The prepared conductive filler composite suspension is added to cement while being stirred at low speed for 2 minutes, and then stirred at high speed for 3 minutes to obtain a uniform cement slurry. The obtained cement slurry is poured into a preset mold for molding. After demolding, it is placed in a standard curing environment for curing until the specified age to obtain a self-sensing cement-based sensor. S12. Performance testing of the self-sensing cement-based sensor, as detailed below: Mechanical properties of the self-sensing cement-based sensor were tested according to the "Test Method for Strength of Cement Mortar". Resistivity of the self-sensing cement-based sensor was tested using the four-electrode method.

3. The method for monitoring the response of an embedded self-sensing cement-based sensor on a cement concrete pavement according to claim 1, characterized in that, Step S2 is as follows: S21. Pretreatment and Electrode Setup: After 28 days of curing, the self-sensing cement-based sensor was used with a four-electrode method. The wires were soldered to the self-sensing cement-based sensor with a soldering iron. The soldering points and the external parts of the wires were wrapped with insulating tape. S22, Surface modification treatment: The self-sensing cement-based sensor was sandblasted using a compressed air sandblasting machine. After sandblasting, the surface was cleaned with compressed air and allowed to dry. S23. Waterproofing treatment: The self-sensing cement-based sensor, after sandblasting, is then encapsulated in waterproof epoxy resin, as detailed below: Peelable masking tape was used to precisely mask the electrode contact points and wire lead-out parts at both ends of the self-sensing cement-based sensor. Then, a two-component epoxy resin coating was uniformly applied to the entire surface of the specimen, and the dry film thickness of the coating was controlled to be 0.2 mm. After coating, let it stand at room temperature until the epoxy resin coating initially sets, then remove the masking tape to ensure that the electrode contact points are completely exposed and the wires are led out normally. After the epoxy resin coating has fully cured, the waterproof encapsulation of the self-sensing cement-based sensor is completed.

4. The method for monitoring the response of an embedded self-sensing cement-based sensor on a cement concrete pavement according to claim 3, characterized in that, In step S22, the pressure of the compressed air sandblasting machine is set to 0.6MPa, the sandblasting distance is 10cm, and the quartz sand particle size is 60-100 mesh, so that the surface roughness Ra of the self-sensing cement-based sensor reaches 2.0μm-3.5μm.

5. The method for monitoring the response of an embedded self-sensing cement-based sensor on a cement concrete pavement according to claim 1, characterized in that, Step S3 is as follows: S31. Elimination of polarization effect: The self-sensing cement-based sensor was connected to the test circuit using the four-electrode method. A constant test current was provided by a DC regulated power supply. After the sensor was powered on and left to stand for 3 minutes, the initial resistance value and initial resistivity of the self-sensing cement-based sensor were recorded. S32, Loading System: The stress control mode was used to load the self-sensing cement-based sensor, and the real-time resistance signal, stress value and strain value of the self-sensing cement-based sensor under different load levels were collected simultaneously. S33. Characteristic parameter calculation and calibration model establishment: Based on the resistivity, stress, and strain data obtained in step S32, the influence of contact resistance and wire resistance is eliminated by the four-electrode method. The following core characteristic parameters are calculated to establish a quantitative piezoresistive model of resistivity change rate-stress-strain for each self-sensing cement-based sensor. The change in resistivity is quantitatively characterized by the rate of change of resistivity (FCR): （1）； R0 and These represent the initial resistance and initial resistivity of the self-sensing cement-based sensor after it has been powered on for a period of time, respectively; Rt and Here, represents the resistance and resistivity values ​​at time t, respectively; A is the lateral cross-sectional area of ​​the self-sensing cement-based sensor; and l is the distance between the two electrodes. Sensitivity coefficient calculation: （2）； In the formula, σ is the stress. This is the stress sensitivity coefficient; （3）； In the formula, ε is the strain and GF is the strain sensitivity coefficient.

6. The method for monitoring the response of an embedded self-sensing cement-based sensor on a cement concrete pavement according to claim 1, characterized in that, In step S32, the loading stress amplitude range is 0.5 to 10 MPa, the loading rate is controlled at 1 kN / s, and the number of complete loading and unloading cycles is 5.

7. The method for monitoring the response of an embedded self-sensing cement-based sensor on a cement concrete pavement according to claim 1, characterized in that, Step S4 is as follows: S41. Sensor installation: A two-stage pouring process is used to embed the calibrated self-sensing cement-based sensor into the road surface concrete. Non-metallic binding straps and external supports are used to fix the self-sensing cement-based sensor in the center of the road surface mold. The long axis of the self-sensing cement-based sensor is perpendicular to the driving direction of the road surface. The wire is led out from the side wall of the road surface mold. Then, the concrete is poured to half the height of the road surface mold. After being vibrated and compacted, it is left to stand for 30 minutes depending on the ambient temperature. When the lower layer of concrete is in the early stage of initial setting, the self-sensing cement-based sensor is placed in the center of the road surface. Concrete is poured to fill the mold and a second vibration is performed to ensure that the self-sensing cement-based sensor and the surrounding road surface concrete are formed and cured synchronously. This ensures that the interface between the self-sensing cement-based sensor and the road surface substrate is tightly bonded and the deformation is coordinated. The wire lead-out end is sealed and protected. S42. Sensor Connection and Data Acquisition: The self-sensing cement-based sensor buried in the road surface is connected to the data acquisition system via an external wire. A DC regulated power supply is used to provide a stable and constant current for the test circuit. The data acquisition system is controlled to collect the resistivity of the self-sensing cement-based sensor at a sampling frequency of 1Hz to 10Hz.

8. The method for monitoring the response of an embedded self-sensing cement-based sensor on a cement concrete pavement according to claim 1, characterized in that, Step S5 is as follows: S51. Establishment of road surface health benchmarks: First, the self-sensing cement-based sensors were calibrated, numbered, and deployed at corresponding monitoring points on the cement concrete pavement. Then, the initial normal state under no vehicle load was selected as the health baseline, and the initial resistivity of each self-sensing cement-based sensor was collected. ,in, , This represents the total number of self-sensing cement-based sensors. S52, Data Smoothing Processing: The resistivity of the self-sensing cement-based sensor is collected once at preset time intervals, and the monitoring data is smoothed using a moving average method. The sliding window length is set as follows: （4）； In the formula, The smoothed resistivity characteristic value after k-period moving average processing is obtained when the i-th self-sensing cement-based sensor performs the j-th monitoring. Let be the original measured resistivity value of the i-th self-sensing cement-based sensor during the j-th monitoring; k is the sliding window length, where the k nearest consecutive data are averaged each time; j is the monitoring number, where k and j are both positive integers and satisfy the constraint k≤j, meaning that at least k data must be accumulated before the smoothing value can be calculated. S53. Based on the above-mentioned full-cycle monitoring data after smoothing and noise reduction, construct a structural state monitoring data sample matrix based on self-sensing cement-based sensors: （5）； In the formula, A smoothed resistivity sample matrix for the entire monitoring period. Let be the smooth resistivity characteristic value of the i-th self-sensing cement-based sensor during the j-th monitoring period; i = 1, 2, ..., m, where m is the self-sensing cement-based sensor; j = 1, 2, ..., n, where n is the total number of monitoring times.

9. The method for monitoring the response of an embedded self-sensing cement-based sensor on a cement concrete pavement according to claim 1, characterized in that, Step S6 is as follows: Based on the development pattern of the resistivity data collected in step S5, four types of working states of the road surface are defined: Health status: Resilient stage, corresponding to SVM label 0; Minor damage state: microcrack initiation stage, corresponding to SVM label 1; Moderate damage state: stable crack propagation stage, corresponding to SVM label 2; Severe damage or destruction state: the unstable propagation stage of cracks, corresponding to SVM label 3.

10. A method for monitoring the response of an embedded self-sensing cement-based sensor on a cement concrete pavement according to claim 1, characterized in that, Step S7 is as follows: S71. Feature Extraction and Label Definition: The smoothed resistivity sequence obtained for each monitoring cycle of the i-th self-sensing cement-based sensor That is, the first digit of matrix Z Line, extract the following time-domain and statistical features to construct a feature vector. : Mean: （6）； Standard deviation: （7）； Maximum fluctuation range: （8）； First-order difference absolute sum: （9）； Mean rate of change of resistivity: （10）； in, Let be the initial health reference resistivity of the i-th self-sensing cement-based sensor; Z-score normalization is applied to the initial feature vector f i The input feature vector F is obtained by processing it to eliminate the influence of dimensions. i ; S72. Historical Sample Construction and Labeling: Historical monitoring data of the road surface at different service stages were collected, and combined with the actual surface inspection, non-destructive testing, or mechanical loading test results, the actual physical damage state of the monitoring location was determined and classified into four categories, denoted as F for each standardized feature sample. i Label : 0 – Health status; 1 – Minor injury; 2 – Moderate damage; 3 – Severe damage or destruction; This allows us to construct a historical sample set for SVM model training. ; Construction of S73 and SVM models: A one-to-one strategy was adopted to construct a multi-class SVM model. Since the SVM model is essentially a binary classification algorithm, a binary classifier was trained between any two classes of samples in the four types of road damage states, and a total of 6 binary sub-models were constructed. In SVM model construction, slack variables are introduced. The penalty parameter C is used to allow a small number of samples to deviate from the classification hyperplane. At the same time, the radial basis kernel function is selected to map the extracted 5-dimensional low-dimensional feature vector to the high-dimensional feature space to process the nonlinear features of the road damage data, so that the originally mixed healthy and different damage state samples can be effectively separated by the hyperplane in the high-dimensional space. The radial basis kernel function includes the width parameter γ, i.e. the model sensitivity, which is used to control the fitting accuracy and generalization ability of the SVM model. S74. Model Training and Parameter Optimization: Arrange the historical sample set constructed in step S72 in chronological order The data from the first 70% of the monitoring period was divided into the training set, and the data from the last 30% was divided into the test set. The training set is input into the multi-class SVM model constructed in step S73. Time-series five-fold cross-validation combined with grid search is used to optimize the penalty parameter C and the kernel parameter γ. The parameter search range is: （11）； Choose the parameter combination that yields the highest average classification accuracy in cross-validation. The model was retrained on the entire training set to establish the final road surface damage recognition model. After training, the generalization ability of the SVM model is evaluated using 30% of the test set, and the following metrics are calculated: The misclassification distribution of adjacent damage states is analyzed using a confusion matrix. Calculate the overall accuracy to evaluate global recognition performance. Calculate the accuracy rate to characterize the system's false alarm control capability. The focus is on calculating the recall rate under severe damage conditions to characterize the system's false negative control capability. Calculate the F1 score for comprehensive performance evaluation; Construct a multidimensional joint evaluation threshold. When the SVM model meets the joint evaluation threshold, deploy the SVM model to the online monitoring system. If it does not meet the threshold, increase the number of historical samples or adjust the feature extraction strategy and retrain. S75. Real-time damage identification and early warning: During the online monitoring phase, the system acquires resistivity signals from the self-sensing cement-based sensor in real time. It calculates the smoothed resistivity sequence for the current monitoring period using the moving average method in step S5, and extracts the feature vector for the current period of the i-th self-sensing cement-based sensor using the method in step S71, performing the same standardization process to obtain the test feature vector F. test,i , i=1,2,...,m; F test,i Take the trained SVM model as an independent input and output the current damage category prediction result for the i-th monitoring location: In the formula, Let i be the damage category at the i-th monitoring location; The system dynamically updates the pavement health status at the corresponding monitoring location based on the damage category output by the SVM model, and executes a graded early warning strategy: When the identification result is 1, the system marks the location as a potential degradation zone, automatically increases the recording frequency of background data, and performs routine tracking. When the identification result evolves into 2, the system generates a moderate warning, prompting the maintenance department to pay closer attention and arrange preventative inspections. A time-sliding window mechanism is introduced to prevent false alarms: The system triggers a severe warning only when the identification result of a certain monitoring location is consistently 3 for N consecutive monitoring cycles. After the severe warning is triggered, the system automatically pushes the specific damage location coordinates, evolution trend and early warning information to the monitoring terminal, prompting the need for substantive engineering maintenance. After the on-site maintenance is completed, the system resets the status of the monitoring node and enters a new round of periodic monitoring.

Images