Fabricated pavement plate changing digital intelligent control and optimized construction method

Abstract

The invention discloses a prefabricated pavement slab changing digital intelligent control and optimization construction method, which realizes high-precision control and quality monitoring in a pavement slab construction process by introducing innovative technologies such as a self-adaptive partition interpolation method, reinforcement learning, a deep neural network and a time sequence convolutional network; compared with a traditional method, intelligent processing of all links such as construction preparation, prefabrication quality detection, hoisting and leveling is optimized, and the construction efficiency and quality are improved; through data-driven automatic control, human intervention is reduced, the construction risk is reduced, and the pavement precision and stability of the pavement slab are ensured; in addition, health monitoring and prediction are performed by using a hypergraph algorithm and an ARIMA model, so that the monitoring precision and the response speed in the operation and maintenance stage are effectively improved.

Description

Technical Field

[0001] This application relates to the field of intelligent construction technology, and particularly to a digital intelligent control and optimized construction method for replacing slabs of prefabricated pavement. Background Art

[0002] With the rapid development of urbanization and transportation, the road traffic volume is increasing day by day, posing higher requirements for pavement performance and service life. During the long-term service process of traditional asphalt and cement concrete pavements, diseases such as cracks, potholes, and ruts are likely to occur, which not only affect driving comfort and safety but also increase road maintenance costs and traffic congestion. Traditional pavement repair methods usually require long-term road closure construction, causing serious interference to traffic. At the same time, the pollution such as noise and dust generated during the construction process also has an adverse impact on the surrounding environment.

[0003] As a new type of pavement structure, prefabricated pavement has the advantages of fast construction speed, recyclability, and good environmental protection performance, showing great potential in meeting the increasing traffic demand and sustainable development. Its core lies in prefabricating the pavement into block units. When maintenance is needed, only the damaged pavement slabs need to be quickly replaced to restore the pavement function, greatly shortening the construction period and reducing traffic interference.

[0004] At present, the existing prefabricated pavement technology still faces some challenges in practical applications: for example, how to ensure the leveling accuracy during the rapid replacement process to guarantee pavement flatness; how to improve the durability of prefabricated slabs, extend their service life, and reduce the replacement frequency; and how to achieve intelligent control and optimization of the construction process to improve construction efficiency and quality.

[0005] Especially in the field of highways, due to factors such as large traffic volume, heavy load, and complex environment, higher performance requirements are imposed on prefabricated pavements. Although the existing prefabricated technology for airport pavements is mature, its prefabricated slab size is large and its self-weight is heavy, which is not suitable for highway scenarios. Therefore, developing a rapid replacement technology for small and lightweight prefabricated slabs suitable for highways and realizing intelligent control and optimized construction in combination with Internet of Things technology is of great significance for improving highway maintenance levels and ensuring smooth road safety. Based on this, researching an intelligent control and optimized construction method for replacing slabs of prefabricated pavement based on the Internet of Things has important theoretical value and practical significance for promoting the popularization and application of prefabricated pavement technology in the highway field. Summary of the Invention

[0006] In order to solve the above technical problems, the embodiments of this application provide a digital intelligent control and optimized construction method for replacing slabs of prefabricated pavement to promote the wide application of prefabricated pavement technology in the highway field.

[0007] The embodiment of this application provides a digital intelligent control and optimized construction method for assembled pavement slab replacement, including the following steps:

[0008] Step S1) Intelligent construction preparation: Obtain the full - range scanning data of the construction site, generate a high - precision digital terrain model through 3D modeling technology, introduce the adaptive zoning interpolation method, dynamically adjust the interpolation accuracy, optimize the construction design to ensure that the construction plan adapts to complex terrains, and obtain the layout design of the pavement slabs;

[0009] Step S2) Intelligent quality inspection of precast pavement slabs: Construct a convolutional neural network model for defect recognition, and generate an adversarial network to classify defect images;

[0010] Step S3) Intelligent hoisting of pavement slabs: Combine reinforcement learning and particle swarm optimization algorithms for hoisting path planning, and use the Kalman filter algorithm to control the hoisting accuracy in real - time to ensure that the hoisting equipment avoids obstacles;

[0011] Step S4) Intelligent quality inspection after pavement slab paving: Construct an LSTM model to conduct real - time monitoring and time - series analysis of the flatness and contact quality during the paving process to ensure the continuous stability of the paving quality;

[0012] Step S5) Intelligent leveling of pavement slabs: Automatically identify the grouting hole positions through a laser scanner or machine vision technology, combine with the previous data analysis of elevation deviation, control the grouting equipment to conduct automatic grouting, cooperate with vibration and pressure sensors to monitor the grouting process in real - time, ensure the accuracy of slurry ratio and filling, and optimize the leveling effect;

[0013] Step S6) Intelligent gap filling and sealing detection: Identify the geometric parameters of the gaps through a machine vision camera, use an improved extreme learning machine algorithm to predict the filling requirements, and use modified asphalt mortar and early - strength cement mortar for gap filling to ensure the uniformity and compressive strength of the gaps;

[0014] Step S7) Intelligent post - operation maintenance: Collect the strain, displacement, and vibration data of the pavement slabs, use the hypergraph algorithm and ARIMA model for health monitoring and prediction, and formulate a maintenance plan;

[0015] Step S8) Global data management and optimization: Integrate and manage all data during the construction process through the Internet of Things platform, optimize the construction plan and process in real - time, ensure the construction quality and efficiency, and dynamically adjust the construction strategy to achieve the best construction effect.

[0016] Furthermore, in the step S1), when introducing the adaptive zoning interpolation method in the process of constructing the digital terrain model, the interpolation accuracy is dynamically adjusted according to the terrain complexity of different regions to generate the digital terrain model: , where, is the interpolation result at position x, where x represents the coordinates on the digital terrain model. represents the weight function corresponding to the i-th known data point at position x. is the value of the known data point i, and n is the number of known data points in the space.

[0017] Furthermore, step S1) specifically includes:

[0018] The layout design is optimized through the constructed digital terrain model and on-site parameters, and using a reinforcement learning algorithm model: aiming at the flatness, material utilization rate, and time cost of the panel layout, an optimal scheme is designed, and continuous evaluation and feedback are carried out through the policy iteration algorithm: , where is the optimal policy, is the immediate reward for each step, is the discount factor, is the expected operation, is to select the policy that maximizes the expected cumulative reward , t is the time step, are the state and action at time step t respectively;

[0019] Based on the completed layout design, the base layer flatness is detected: using a flatness analysis algorithm based on wavelet transform. , where is the original data of the signal in the time domain; is the complex conjugate of the mother wavelet; is the mother wavelet function; is the translation factor, and s is the scale factor.

[0020] Furthermore, step S2) specifically includes:

[0021] Using a machine vision system and combining with an improved convolutional neural network for defect detection: , where is the activation value of the output layer, representing the detection result of cracks or damages; is the activation function; is the convolutional kernel weight; is the feature of the input image; is the bias term; R is the number of feature dimensions;

[0022] Using a deep neural network for defect recognition, and generating an adversarial network to perform fusion analysis on multi-source data to generate a higher-quality defect image classification model. , where represents the sample e sampled from the real data distribution. Denote the input \(z\) sampled from a random noise distribution, and \(E\) represents the expectation operation, that is, the average loss calculation of the samples. Denote the loss of the real samples, that is, the logarithmic loss of the classification probability of the discriminator for the real data. Denote the loss of the generated samples, that is, the logarithmic loss of the classification probability of the discriminator for the generated data; \(G\) are the parameters of the generator, and the generator generates data close to the real data by optimizing these parameters; \(D\) are the parameters of the discriminator, and the discriminator accurately distinguishes between real and generated data by optimizing these parameters.

[0023] For internal quality inspection, non-destructive testing techniques are adopted and combined with the temporal convolutional network in deep learning for data modeling and analysis, dynamically tracking the change trend of the pavement slab during the production process. Each convolutional kernel will obtain the weighted features at each time step in the time series through the convolutional operation with the input features. After processing these features through a multi-layer convolutional network, finally output a result for evaluating the current internal quality state of the pavement slab: , where is the predicted value at time step \(t\); is the convolutional kernel weight; TCN is the temporal convolutional operation, extracting the data features at this moment; is the input feature at time step \(c - 1\); \(N\) is the length of the time series;

[0024] Generate the quality score of the pavement slab through a multi-dimensional weighted evaluation model for all the detection data, and generate an electronic label with complete quality information: , where is the comprehensive quality score of the pavement slab; 、 are the weights of each quality index; is the score of the \(p\)-th quality index; is the score of the \(q\)-th depth feature; \(m\) is the number of quality indexes; \(v\) is the number of depth features.

[0025] Furthermore, the hoisting path planning in step S3) specifically includes:

[0026] S31: Through a multi-objective optimization method combining reinforcement learning and particle swarm optimization algorithm, considering the movement trajectory of the hoisting equipment, the size, weight of the pavement slab, and the distribution of on-site obstacles, optimize the hoisting path to ensure that the equipment avoids obstacles and reduces the energy consumption and time of the equipment movement: , where represents the minimum distance between the hoisting equipment and the obstacle in the \(u\)-th hoisting trajectory; 、 are the lateral and longitudinal offsets of the \(u\)-th equipment path respectively; The optimization objective function adjusts the weight coefficient To balance different optimization objectives and achieve the optimal adjustment of the path; T is the number of hoisting trajectory segments;

[0027] S32: Establish a virtual 3D model synchronized with the construction site using digital twin technology, and use the Monte Carlo method to conduct risk assessment and uncertainty analysis of the hoisting process. The Monte Carlo method evaluates risks through random simulation of multiple scenarios: , where, is the estimated probability of potential risk; is the result of the m-th simulation; is the defined failure area, indicates that it is 1 when belongs to the failure area, otherwise it is 0; where the failure area represents the area exceeding any one of the maximum allowable tilt angle, maximum allowable displacement error, and maximum force on the hoisting rope; O represents the total number of simulations;

[0028] S33: For hoisting precision control, adopt the Kalman filter algorithm combined with RTK or UWB positioning system and high-precision sensors to fuse sensor data in real time to accurately estimate the position and attitude of the pavement slab: , where, is the estimated value at time k; is the Kalman gain; is the current observed value, is the observation matrix, and the initial observation matrix = ; , indicating that the initial pavement slab is located at a certain position in the site, with the xy axes of its three-dimensional coordinates being 0 and the initial height being 3 meters; in addition also includes setting the initial attitude of the pavement slab to , indicating that the pavement slab has no tilt initially, and the initial Kalman gain ;

[0029] During the hoisting process, the real-time monitoring system combines an anomaly detection algorithm to dynamically identify potential hoisting risks. When the sensor data corresponding to the hoisting rope force and the pavement slab attitude exceeds the safety threshold, the hoisting risk is evaluated through a hybrid model of support vector machine and anomaly detection algorithm, , where, is the prediction result of anomaly detection; is the current sensor data; is the mean value of the data; is the preset safety threshold; when the difference between the current sensor data and the mean value exceeds the threshold , it indicates that there is an anomaly and an early warning is immediately issued.

[0030] Furthermore, the specific content of step S4) is as follows:

[0031] After the pavement slab paving is completed, a mobile robot equipped with a laser scanner or a structured light sensor is used to quickly scan the pavement slab and intelligently detect key indicators; the key indicators at least include flatness, elevation difference, and the width of the gap between slabs.

[0032] Through the long short-term memory network, perform temporal analysis on the scanned data, capture the dynamic change trend of the pavement slab flatness over time, and predict and optimize the future paving quality:

[0033]

[0034] Among them, is the forget gate, which determines the proportion of information to be discarded currently;

[0035] is the input gate, which determines the proportion of information to be updated currently;

[0036] is the output gate, which determines the proportion of information read from the cell state at the current moment;

[0037] is the cell state, which updates the cell state according to the outputs of the forget gate and the input gate;

[0038] is the hidden state, the network output at the current moment, which is used to be passed to the LSTM cell at the next moment;

[0039] represents the hidden state at the previous moment and the input data at the current moment;

[0040] is the forget gate bias vector, which is used to adjust the output of the forget gate;

[0041] is the input gate bias vector, which determines the information to be updated into the cell state at the current moment;

[0042] is the output gate bias vector, which determines which information in the current cell state will be passed to the hidden state at the next moment;

[0043] is the cell state bias vector, which is used to update the current cell state;

[0044] , , , The weight matrices corresponding to the update of the forget gate, input gate, output gate, and cell state;

[0045] In terms of contact quality detection, obtain the acoustic signals collected by knocking on the road slab, use the decision tree algorithm to analyze the propagation characteristics of sound waves, and intelligently identify the contact state between the road slab and the base layer;

[0046] The decision tree algorithm realizes classification by continuously dividing the feature space and constructs a tree-shaped decision model , where is the sample set of the current node, is the number of categories, is the probability that the sample belongs to the th category;

[0047] Based on the acoustic signals, automatically identify the hollowing or loose adhesion phenomenon between the road slab and the base layer by training the decision tree model. Then, combine laser scanning, acoustic detection and image recognition technologies, and use the neural network algorithm to comprehensively analyze multi-source data. The optimization goal is to minimize the error, and the parameters are updated through the backpropagation algorithm to comprehensively evaluate the paving quality of the road slab and generate a detection report: , where is the loss function; is the learning rate; are the network parameters (weights and biases).

[0048] Furthermore, the specific steps of step S5) are as follows:

[0049] During the leveling process of the road slab, first automatically identify the reserved grouting hole positions through a laser scanner or image recognition technology, and analyze the elevation deviation of each area in combination with the previous detection data;

[0050] Based on this, the intelligent grouting equipment automatically performs the proportioning, stirring and transportation of the grout to ensure the uniformity and accuracy of grouting;

[0051] Vibration sensors and pressure sensors continuously monitor the grout diffusion and filling conditions, and dynamically optimize the grouting model according to the monitoring data to ensure the high efficiency and stability of the grouting process;

[0052]

[0053] where is the real-time grouting height; is the target filling height; is the real-time diffusion radius; is the maximum allowable diffusion radius; 、 are the change amounts of flow rate and pressure respectively; 、 、 、 are the weight coefficients.

[0054] Further, the step S6) is specifically as follows:

[0055] Automatically identify and measure the width and depth of the gap through machine vision technology, predict the filling requirements using an improved extreme learning machine algorithm, and automatically calculate the most suitable filling material ratio and dosage; the extreme learning machine algorithm accelerates the training process by randomly generating hidden layer nodes and optimizing the output layer weights:

[0056] ,

[0057] wherein, is the hidden layer output matrix; , , …, are input samples; is the output of the rd node of the hidden layer, is the number of samples; the output of each node is determined by the activation function ; is the output layer weight, is 's transpose matrix, is the identity matrix, is the regularization parameter, is the expected output, i.e., the filling requirement;

[0058] During the filling process, the system uses fuzzy adaptive control technology to dynamically adjust the grouting flow rate, pressure, and slurry ratio; the fuzzy controller makes adjustments based on real-time sensor data: , wherein, is the grouting control output, is the fuzzy membership function, is the fuzzy control rule gain coefficient, is the change amount of the control input, and fuzzy adaptive control ensures precise control of the grouting process by dynamically adjusting the rules and gain coefficients; is the number of fuzzy rules (set to 5);

[0059] After the filling is completed, the filling effect of the gap is monitored in real time through an optical rheometer and laser measurement technology to ensure filling uniformity and quality; and image recognition technology is used to check for defects such as air bubbles and cracks.

[0060] Further, the step S7) is specifically as follows:

[0061] During the later operation and maintenance process of the road slab, the intelligent sensor network is responsible for collecting health monitoring data of strain, displacement, and vibration, and transmitting it to the cloud platform in real time through a wireless network;

[0062] Based on the hypergraph algorithm and multi-modal information fusion technology, conduct in-depth analysis of sensor data, , where, represents the total relationship between nodes; represents the node set; represents the edge set, representing the relationship between nodes; represents the hyperedge set;

[0063] Adopt the autoregressive integrated moving average model for prediction, accurately capture the changing trend of the road slab state through autoregression and moving average, predict potential structural problems, and issue early warnings in a timely manner: , where, is the model parameter at the current moment, is the gradient of the loss function at the current moment, is the learning rate.

[0064] Advantages of the present invention:

[0065] By introducing innovative technologies such as adaptive partition interpolation method, reinforcement learning, deep neural network (DNN), and temporal convolutional network (TCN), high-precision control and quality monitoring of the road slab construction process are achieved. Compared with traditional methods, the present invention optimizes the intelligent processing of each link such as construction preparation, precast quality inspection, hoisting, and leveling, improves construction efficiency and quality. Through data-driven automatic control, human intervention is reduced, construction risks are lowered, and the accuracy and stability of road slab paving are ensured. In addition, health monitoring and prediction are carried out using the hypergraph algorithm and ARIMA model, effectively improving the monitoring accuracy and response speed in the operation and maintenance stage. The present invention has significant technological innovation and promotes the wide application of prefabricated pavement technology in the highway field. Description of the Drawings

[0066] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0067] Figure 1 is the flowchart of the present application. Detailed Embodiments

[0068] In order to make the application purpose, features, and advantages of this application more obvious and understandable, the following will describe the technical solutions in the embodiments of this application clearly and completely in conjunction with the accompanying drawings in the embodiments of this application. Obviously, the embodiments described below are only a part of the embodiments of this application, rather than all the embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of this application.

[0069] The following further clarifies the present invention in conjunction with the accompanying drawings and specific embodiments.

[0070] In the description of this application, it should be understood that the orientation or positional relationships indicated by the terms "upper", "lower", "top", "bottom", "inner", "outer", etc. are based on the orientation or positional relationships shown in the accompanying drawings, and are only for the convenience of describing this application and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation to this application.

[0071] The following describes this application in conjunction with specific embodiments:

[0072] This embodiment proposes an intelligent control and optimized construction method for prefabricated pavement slab replacement based on the Internet of Things. The specific operation steps are as follows:

[0073] Step S1: Intelligent construction preparation:

[0074] Use a DJI Matrice 300 RTK drone equipped with a Velodyne VLP-16 lidar to perform on-site terrain scanning, generate a high-precision digital terrain model (DTM), and introduce an adaptive zoning interpolation method to dynamically adjust the interpolation accuracy according to different terrain complexities, optimize the construction design, and ensure that the construction plan adapts to complex terrains and improves the design accuracy.

[0075] As a specific embodiment,

[0076] First, use a drone equipped with a lidar or an oblique photography camera to perform a comprehensive and rapid scan of the construction site, and generate a high-precision digital terrain model (DTM) through three-dimensional modeling technology.

[0077] For the model generation process, traditional methods often rely on interpolation algorithms with fixed parameters, such as the Kriging method. However, they often face problems of low accuracy and poor adaptability when dealing with complex terrains, especially in areas with large terrain variations, and cannot effectively improve the interpolation accuracy. In contrast, the present invention innovatively introduces an adaptive partition interpolation method. This algorithm dynamically adjusts the interpolation accuracy according to the terrain complexity of different regions, generates a more accurate and adaptable digital terrain model, significantly overcomes the limitations of traditional methods, and improves the overall accuracy and computational efficiency: , where is the interpolation result at position x, represents the weight function corresponding to the i-th known data point at position x, is the value of the known data point i, and n is the number of known data points in the space.

[0078] Based on this model and on-site parameters, a reinforcement learning algorithm is used to optimize the layout design of the pavement slabs. The reinforcement learning model aims at the flatness, material utilization rate, and time cost of the slab layout, designs an optimal solution, and continuously evaluates and feeds back through the policy iteration algorithm: , where is the optimal policy, is the immediate reward at each step, is the discount factor, is the expected operation, is to select the policy that maximizes the expected cumulative reward , t is the time step, are the state and action at time step t, respectively;

[0079] During the subgrade flatness detection process, a flatness analysis algorithm based on wavelet transform is adopted,

[0080]

[0081] where is the original data of the signal in the time domain; is the complex conjugate of the mother wavelet; is the mother wavelet function; is the translation factor, and s is the scale factor.

[0082] This algorithm has higher accuracy and multi-scale characteristics compared with traditional methods. Traditional methods can only analyze signals of a single frequency and are difficult to comprehensively capture the multi-dimensional characteristics of flatness. In contrast, the algorithm based on wavelet transform can accurately extract different frequency components through multi-scale analysis and comprehensively reflect the local and overall flatness of the subgrade.

[0083] Step S2: Intelligent quality inspection of precast pavement slabs:

[0084] Integrate an automated production line with the Keyence CV-X series machine vision system and the LMI Technologies Gocator 2490 laser measurement system to conduct real-time inspection of the road slabs and ensure production quality. Adopt a convolutional neural network (CNN) model under the TensorFlow framework for defect recognition, and combine it with a generative adversarial network (GAN) to classify defect images, further improving the detection accuracy and ensuring that each road slab meets the design quality requirements. Use C50 high-strength concrete during production, with a compressive strength ≥ 50 MPa, to ensure the durability and structural safety of the slabs.

[0085] As a specific embodiment,

[0086] During the prefabrication of road slabs, the intelligent system integrates machine vision and laser measurement technologies through an automated production line to monitor and automatically adjust various key parameters in the production process in real time, ensuring the stability of prefabrication quality. Compared with the traditional manual monitoring and detection using a single measurement tool, which cannot achieve real-time feedback and automatic adjustment, the present invention introduces a machine vision system and combines it with an improved convolutional neural network (CNN) for defect detection, overcoming the problems of low accuracy and many missed detections in traditional manual visual inspection and simple image processing methods, and significantly improving the accuracy and efficiency of defect recognition: , where is the activation value of the output layer, representing the detection result of cracks or damages; is the activation function; is the convolutional kernel weight; is the feature of the input image; is the bias term; R is the number of feature dimensions.

[0087] In the quality inspection after production, use the generative adversarial network (GAN) in the deep neural network (DNN) to conduct fusion analysis on multi-source data and generate a higher-quality defect image classification model:

[0088] , where, represents the sample e sampled from the real data distribution, represents the input z sampled from the random noise distribution, E represents the expectation operation, that is, the average loss calculation of the samples, represents the loss of the real sample, that is, the logarithmic loss of the classification probability of the discriminator for the real data, represents the loss of the generated sample, that is, the logarithmic loss of the classification probability of the discriminator for the generated data; G is the parameter of the generator, and the generator optimizes these parameters to generate data close to the real; D is the parameter of the discriminator, and the discriminator optimizes these parameters to accurately distinguish between real and generated data;

[0089] Internal quality inspection uses non-destructive testing techniques such as ultrasonic or ray testing, and combines the Temporal Convolutional Network (TCN) in deep learning for data modeling and analysis to dynamically track the changing trends of the road slabs during the production process, thereby accurately identifying potential defects. Each convolutional kernel will obtain weighted features for each time step in the time series through convolution operations with the input features. After being processed by multiple layers of convolutional networks, these features finally output a result for evaluating the current internal quality status of the road slab (such as voids, density, etc.): , where is the predicted value at time step t; is the convolutional kernel weight; TCN is the temporal convolution operation to extract the data features at that moment; is the input feature at time step c - 1; N is the length of the time series;

[0090] Finally, all inspection data will be integrated into the intelligent quality rating system, and a quality score for the road slab will be generated through a multi-dimensional weighted evaluation model, generating an electronic label with complete quality information to achieve traceability and control of quality problems, further strengthening the quality management in the subsequent construction stage.

[0091] Generate a quality score for the road slab through a multi-dimensional weighted evaluation model for all inspection data, generating an electronic label with complete quality information: , where is the comprehensive quality score of the road slab; , are the weights of each quality index; is the score of the p-th quality index (such as crack width, surface flatness, etc.); is the score of the q-th depth feature (such as internal density, void defect, etc.); m is the number of quality indexes; v is the number of depth features. By monitoring the quality score, re-production is carried out for the parts that do not reach the preset cut-off score.

[0092] Step S3: Intelligent lifting of road slabs:

[0093] Use a KUKA KR Quantec robot to cooperate with reinforcement learning and the Particle Swarm Optimization Algorithm (PSO) for lifting path planning to ensure that the lifting equipment can avoid obstacles and reduce energy consumption. Combine Siemens Simatic S7-1500 PLC and the CalmanFilter algorithm to real-time control the lifting accuracy through digital twin technology to ensure the efficiency and precision of the lifting operation process.

[0094] During hoisting, use 8.8-grade high-strength steel screws (screw diameter: 20 mm) to ensure the stability and safety during hoisting. The screws are made of high-strength alloy steel materials to withstand large loads and maintain fatigue resistance. The hoisting process combines digital twin technology and Kalman filtering algorithm to achieve precise control of hoisting accuracy, avoid equipment collisions, and minimize energy consumption to the greatest extent.

[0095] As a specific embodiment,

[0096] S31: For hoisting path planning, first, a multi-objective optimization method combining reinforcement learning (RL) and particle swarm optimization (PSO) is used. Considering the movement trajectory of the hoisting equipment, the size, weight of the road slab, and the distribution of on-site obstacles, the hoisting path is optimized to ensure that the equipment avoids obstacles and reduces the energy consumption and time of equipment movement:

[0097] , where represents the minimum distance between the hoisting equipment and the obstacle in the u-th hoisting trajectory; and are the lateral and longitudinal offsets of the u-th equipment path respectively; the optimization objective function adjusts the weight coefficient to balance different optimization objectives and achieve the optimal adjustment of the path; T is the number of segments of the hoisting trajectory.

[0098] S32: Based on this, a virtual 3D model synchronized with the construction site is established using digital twin technology, and the Monte Carlo method is used for risk assessment and uncertainty analysis of the hoisting process. The Monte Carlo method evaluates risks through random simulation of multiple scenarios: , where is the estimated probability of potential risks; is the result of the m-th simulation; is the defined failure area, represents 1 when belongs to the failure area, otherwise 0; where the failure area represents the area exceeding any one of the maximum allowable tilt angle, maximum allowable displacement error, and maximum rope tension; O represents the total number of simulations.

[0099] In terms of hoisting accuracy control, the Kalman filtering algorithm is combined with RTK or UWB positioning system and high-precision sensors to fuse sensor data in real time to accurately estimate the position and attitude of the road slab: , where is the estimated value at time k; is the Kalman gain; is the current observation value, is the observation matrix, and the initial observation matrix = ; , indicating that the initial slab is located at a certain position in the site, with the xy axes of its three-dimensional coordinates being 0 and the initial height being 3 meters; in addition also includes setting the initial attitude of the slab to , indicating that initially the slab has no inclination and the initial Kalman gain .

[0100] During the hoisting process, the real-time monitoring system combined with the anomaly detection algorithm can identify potential hoisting risks. When the sensor data such as the force on the hoisting rope and the attitude of the slab exceed the safety threshold (in this embodiment, the safety threshold is set to 80% of the maximum load of the hoisting equipment), the system evaluates the hoisting risk through a hybrid model of support vector machine and anomaly detection algorithm.

[0101] , where is the prediction result of anomaly detection; is the current sensor data; is the mean value of the data; is the preset safety threshold; when the difference between the current sensor data and the mean value exceeds the threshold , it indicates that there is an anomaly and an early warning is immediately issued.

[0102] Step S4: Quality inspection after intelligent paving of the slab:

[0103] Use the Clearpath Robotics Jackal mobile robot equipped with the FARO Focus S70 laser scanner to monitor the flatness and contact quality of the slab in real time. Use the long short-term memory network (LSTM) under the TensorFlow framework to perform time series analysis on the scanned data to predict the change of paving quality and ensure the continuous stability of quality during the paving operation. Compared with the traditional static quality inspection method, the LSTM network can effectively process time series data, realize the prediction and optimization of paving quality, so as to identify potential problems in advance and provide solutions. In addition, the system combines the decision tree analysis method of acoustic signals to identify the contact state between the slab and the base layer, automatically detect the phenomenon of hollowing or loose adhesion, and further improve the accuracy and intelligence level of paving quality inspection.

[0104] Precast slab specifications: The standard precast slab size is 3750mm×3000mm×80mm, using C50 high-strength concrete with a compressive strength ≥50MPa. The laser positioning system ensures that the level error is controlled within ±3mm.

[0105] As a specific embodiment,

[0106] After the pavement of the road slab is completed, a laser scanner or a structured light sensor carried by a mobile robot is used to quickly scan the road slab and intelligently detect key indicators such as flatness, height difference, and the width of the gap between slabs.

[0107] Next, through the long short-term memory network (LSTM), temporal analysis is performed on the scanned data. The LSTM network can effectively capture the dynamic change trend of the flatness of the road slab over time, so as to predict and optimize the future paving quality.

[0108] , where is the forget gate (determining the proportion of information to be discarded currently), is the input gate (determining the proportion of information to be updated currently), is the output gate (determining the proportion of information read from the cell state at the current moment), is the cell state (updating the cell state according to the outputs of the forget gate and the input gate), is the hidden state (the output of the network at the current moment, used to be passed to the LSTM cell at the next moment), represents the hidden state at the previous moment and the input data at the current moment, is the forget gate bias vector, used to adjust the output of the forget gate;

[0109] is the input gate bias vector, determining the information to be updated into the cell state at the current moment;

[0110] is the output gate bias vector, determining which information in the current cell state will be passed to the hidden state at the next moment;

[0111] is the cell state bias vector, used to update the current cell state;

[0112] , , , The weight matrices corresponding to the updates of the forget gate, input gate, output gate, and cell state.

[0113] In terms of contact quality detection, the system analyzes the acoustic signals collected by knocking on the road slab using the decision tree algorithm to intelligently identify the contact state between the road slab and the base layer. The basic principle of the decision tree algorithm is to achieve classification by continuously dividing the feature space and constructing a tree-shaped decision model: , where is the sample set of the current node, is the number of categories, is that the sample belongs to the Probability of the class.

[0114] By training a decision tree model, the system can automatically identify the phenomenon of hollowing or loose adhesion between the slab and the base course based on acoustic signals. Subsequently, a combination of laser scanning, acoustic detection, and image recognition technologies is adopted, and the neural network (NN) algorithm is used to comprehensively analyze multi-source data. The optimization objective is to minimize the error, and parameter updates are performed through the backpropagation algorithm. The paving quality of the slab is comprehensively evaluated, and a detailed inspection report is generated to provide key data support for subsequent intelligent grouting and leveling work: , where is the loss function; is the learning rate; are the network parameters (weights and biases).

[0115] Step S5: Intelligent slab leveling:

[0116] Automatically identify the reserved grouting holes through a FARO Focus S70 laser scanner or a Basler acA2000-50gm machine vision camera, and analyze the elevation deviation of each area in combination with the previous detection data.

[0117] Use a Graco Fusion Air Purge grouting device for automatic grouting to ensure accurate proportioning, mixing, and transportation of the grout. Vibration sensors (such as PCB Piezotronics vibration sensors) and pressure sensors (such as Honeywell pressure sensors) monitor the grouting diffusion and filling process in real time, and the system dynamically adjusts the grouting parameters according to the sensor feedback data to optimize the grouting process.

[0118] Select modified cement slurry as the grouting material. The ratio is P.O4.25 cement: water: medium sand = 1: 2.5: 0.4, the compressive strength is ≥30MPa, and the final setting time is about 2 hours. The grouting flow rate is controlled at 150 - 250L / h, and the pressure is controlled at 0.6 - 0.9MPa to ensure that the elevation deviation area is fully filled without excessive diffusion. After completion, use laser scanning technology to monitor the leveling effect to ensure a uniform layer, no bubbles and cracks, and ensure the construction quality.

[0119] As a specific embodiment, during the slab leveling process, first automatically identify the reserved grouting holes through a laser scanner or image recognition technology, and analyze the elevation deviation of each area in combination with the previous detection data. Based on this, the intelligent grouting equipment automatically performs proportioning, mixing, and transportation of the grout to ensure the uniformity and accuracy of grouting. Vibration sensors and pressure sensors monitor the grouting diffusion and filling conditions in real time, and the system dynamically optimizes the grouting model according to the monitoring data to ensure an efficient and stable grouting process.

[0120] , where is the real-time grouting height; is the target filling height; is the real-time diffusion radius; is the maximum allowable diffusion radius; and are the change amounts of flow rate and pressure respectively; and and and are the weight coefficients.

[0121] Step S6: Intelligent gap filling and sealing detection:

[0122] Identify the gap geometric parameters through a Basler acA2000-50gm machine vision camera, and use an improved Extreme Learning Machine (ELM) algorithm to predict the filling requirements to ensure accurate matching. The filling materials are selected as modified asphalt mortar (compressive strength ≥ 25 MPa, ductility ≥ 10%) and early-strength cement mortar (initial setting time ≤ 2 hours, compressive strength ≥ 35 MPa). The modified asphalt mortar is used in the upper 1 / 2 area, and the early-strength cement mortar is used in the lower 1 / 2 area. The slurry flow rate is controlled at 200 - 300 L / h, and the pressure is 0.5 - 0.8 MPa to ensure uniformity and quality. After filling, a stress absorption layer (thickness 10 mm) is laid, and a protective layer is laid through an automatic laying machine, with a construction speed of 20 square meters per hour, and the layer thickness uniformity is controlled within ±1 mm.

[0123] As a specific embodiment:

[0124] During the process of treating the slab gap, the system automatically identifies and measures the width and depth of the gap through machine vision technology, providing accurate data support for subsequent intelligent filling. Based on these geometric parameters, the system uses an improved Extreme Learning Machine (ELM) algorithm to predict the filling requirements and automatically calculates the most suitable filling material ratio and dosage. The ELM algorithm accelerates the training process by randomly generating hidden layer nodes and optimizing the output layer weights:

[0125] ,

[0126] where is the hidden layer output matrix; and and …, are the input samples; is the output of the th node in the hidden layer, is the number of samples; the output of each node is determined by the activation function to determine; is the output layer weight, is the transpose matrix of, is the identity matrix, is the regularization parameter, is the expected output, i.e., the filling requirement. This optimization formula obtains the optimal weight by minimizing the error, thus accelerating the training process and reducing the calculation time.

[0127] During the filling process, the system uses fuzzy adaptive control (FAC) technology to dynamically adjust the grouting flow rate, pressure, and slurry ratio. The fuzzy controller makes adjustments based on real-time sensor data (such as pressure and flow rate):

[0128] , where is the grouting control output, is the fuzzy membership function, is the fuzzy control rule gain coefficient, is the change amount of the control input, is the number of fuzzy rules. Fuzzy adaptive control ensures precise control of the grouting process by dynamically adjusting the rules and gain coefficients.

[0129] After the filling is completed, the system uses an optical rheometer and laser measurement technology to monitor the filling effect of the gap in real time to ensure filling uniformity and quality. Finally, image recognition technology is used to check for defects such as bubbles and cracks to ensure the final compliance of the filling quality.

[0130] Step S7: Intelligent post-operation and maintenance:

[0131] In the post-operation and maintenance stage, by collecting health data such as strain, displacement, and vibration of the road slab, and conducting in-depth analysis and prediction in combination with the hypergraph algorithm and ARIMA model, long-term monitoring of the road surface health condition is achieved. Based on the prediction results, the system can formulate an optimized maintenance plan, thereby effectively extending the service life of the road slab. Compared with the traditional graph neural network (GNN), the present invention adopts the hypergraph algorithm, which can handle the high-order relationships between nodes, exceeding the limitation of only second-order relationships, and can effectively model the interaction between different types of sensors and the influence of environmental factors on the health state of the road slab. In addition, the system combines the autoregressive integrated moving average (ARIMA) model for health state prediction, uses historical monitoring data for incremental update, optimizes the model structure, and ensures accurate prediction of the health state of the road slab.

[0132] As a specific embodiment,

[0133] During the later operation and maintenance of the road slab, the intelligent sensor network is responsible for collecting health monitoring data such as strain, displacement, and vibration, and transmitting it to the cloud platform in real time through a wireless network. Based on the hypergraph algorithm and multi-modal information fusion technology, the system deeply analyzes the sensor data. Compared with the traditional graph neural network (GNN), the hypergraph algorithm can not only handle the second-order relationships between nodes, but also deal with complex higher-order associations, such as the interaction between different types of sensors and the impact of environmental factors on the health of the road slab, improving the accuracy of health status assessment.

[0134] , where represents the node set (sensors, structural units, etc.); represents the edge set (representing the relationships between nodes); represents the hyperedge set (higher-order structural relationships, such as the time dependence of sensor data and environmental factors).

[0135] At the same time, the system uses the autoregressive integrated moving average (ARIMA) model for prediction. This model combines historical monitoring data and accurately captures the changing trend of the road slab state through autoregression and moving average, so as to more accurately predict potential structural problems and issue early warnings in a timely manner.

[0136]

[0137] where are the model parameters at the current time, is the gradient of the loss function at the current time, is the learning rate. This algorithm optimizes the model through continuous incremental updates, thus ensuring the sensitivity of the model to new data.

[0138] Step S8: Global data management and optimization:

[0139] Integrate and manage all construction process data through the Siemens MindSphere Internet of Things platform, optimize the construction plan and process in real time, and ensure construction quality and efficiency. Through data analysis and feedback mechanisms, dynamically adjust construction strategies to ensure that each link achieves the best results.

[0140] As a specific embodiment:

[0141] During the entire intelligent construction process, the data collected in all links will be integrated and managed to form a complete construction process data chain. The system monitors the construction progress and quality in real time, compares and analyzes them with the plan, and dynamically optimizes the construction plan and process parameters through data feedback to ensure construction quality and efficiency. At the same time, the key information of the construction process is visually displayed in the form of 3D models and charts, and managers can conduct retrospective analysis of historical data at any time to optimize future construction management processes.

[0142] The intelligent control and optimized construction method for prefabricated pavement slab replacement based on the Internet of Things proposed in the present invention realizes high-precision control and quality monitoring of the pavement slab construction process by introducing innovative technologies such as adaptive zonal interpolation method, reinforcement learning, deep neural network (DNN), and temporal convolutional network (TCN). Compared with traditional methods, the present invention optimizes the intelligent processing of each link such as construction preparation, prefabrication quality inspection, hoisting, and leveling, improving construction efficiency and quality. Through data-driven automatic control, it reduces human intervention, lowers construction risks, and ensures the accuracy and stability of pavement slab paving. In addition, the use of hypergraph algorithms and ARIMA models for health monitoring and prediction effectively improves the monitoring accuracy and response speed in the operation and maintenance stage. The present invention has significant technological innovation and promotes the wide application of prefabricated pavement technology in the highway field.

[0143] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts among the various embodiments can be referred to each other.

[0144] The preferred embodiments of the present invention have been described in detail above. However, the present invention is not limited to the specific details in the above embodiments. Within the scope of the technical concept of the present invention, various equivalent transformations (such as quantity, shape, position, etc.) can be made to the technical solution of the present invention, and these equivalent transformations all fall within the protection scope of the present invention.

Claims

1. A digital intelligent control and optimization construction method for prefabricated pavement plate replacement, characterized in that: The steps include: Step S1) Intelligent construction preparation: Obtain all-round scanning data of the construction site, generate a high-precision digital terrain model through 3D modeling technology, introduce adaptive partition interpolation method, dynamically adjust interpolation accuracy, optimize construction design to ensure that the construction plan adapts to complex terrain, and obtain the layout design of road panels; Step S2) Intelligent road panel prefabrication quality inspection: Construct a convolutional neural network model for defect recognition and generate an adversarial network to classify defect images; Step S3) Intelligent road panel hoisting: Combine reinforcement learning and particle swarm optimization algorithms to plan the hoisting path, and use the Kalman filter algorithm to control the hoisting accuracy in real time to ensure that the hoisting equipment avoids obstacles; Step S4) Intelligent road slab post-paving quality inspection: Build an LSTM model to conduct real-time monitoring and time series analysis of the flatness and contact quality during the paving process to ensure continuous and stable paving quality; Step S5) Intelligent road panel leveling: Automatically identify the grouting hole position through laser scanner or machine vision technology, analyze the elevation deviation in combination with previous data, control the grouting equipment to perform automatic grouting, and use vibration and pressure sensors to monitor the grouting process in real time to ensure accurate slurry ratio and filling, and optimize the leveling effect; Step S6) Intelligent gap filling and sealing detection: The gap geometric parameters are identified by a machine vision camera, the filling requirements are predicted using an improved extreme learning machine algorithm, and modified asphalt mortar and early strength cement mortar are used to fill the gaps to ensure the uniformity and compressive strength of the gaps; Step S7) Intelligent post-operation and maintenance: by collecting the strain, displacement and vibration data of the road panel, using the hypergraph algorithm and ARIMA model to perform health monitoring and prediction, and formulate maintenance plans; Step S8) Global data management and optimization: All data in the construction process are integrated and managed through the IoT platform, construction plans and processes are optimized in real time, construction quality and efficiency are ensured, and construction strategies are dynamically adjusted to achieve the best construction results.

2. The method for digital intelligent control and optimized construction of prefabricated pavement plate replacement according to claim 1 is characterized in that: In the step S1), when the adaptive partition interpolation method is introduced in the process of constructing the digital terrain model, the interpolation accuracy is dynamically adjusted according to the complexity of the terrain in different regions to generate the digital terrain model: ,in, is the interpolation result at position x, where x represents the coordinate on the digital terrain model. represents the weight function corresponding to the i-th known data point at position x, is the value of the known data point i, and n is the number of known data points in space.

3. The method for digital intelligent control and optimization construction of prefabricated pavement plate replacement according to claim 1 is characterized in that: The step S1) specifically includes: The layout design is optimized by using the constructed digital terrain model and on-site parameters and the reinforcement learning algorithm model: the optimal solution is designed with the flatness, material utilization and time cost of the panel layout as the goal, and the feedback is continuously evaluated through the strategy iteration algorithm: ,in, is the optimal strategy, is the immediate reward for each step, is the discount factor, For the expected operation, To select the strategy that maximizes the expected cumulative reward , t is the time step, are the state and action at time step t respectively; Based on the completed layout design, the flatness of the base layer is detected: the flatness analysis algorithm based on wavelet transform is used. ,in, is the original data of the signal in the time domain, is the complex conjugate of the mother wavelet, is the mother wavelet function, is the translation factor and s is the scale factor.

4. The method for digital intelligent control and optimization construction of prefabricated pavement plate replacement according to claim 1 is characterized in that: The step S2) specifically includes: Defect detection using machine vision system combined with improved convolutional neural network: ,in is the activation value of the output layer, indicating the detection result of cracks or damage; is the activation function; is the convolution kernel weight; is the feature of the input image; is the bias term; R is the number of feature dimensions; Use deep neural networks to identify defects and generate adversarial networks to fuse and analyze multi-source data to generate higher-quality defect image classification models ,in, represents the sample e sampled from the true data distribution, represents the input z sampled from a random noise distribution, E represents the expected operation, that is, the average loss calculation of the sample, Represents the loss of the real sample, that is, the logarithmic loss of the classification probability of the discriminator for the real data, represents the loss of generated samples, that is, the logarithmic loss of the classification probability of the discriminator for the generated data; G is the parameter of the generator, and the generator generates data close to the real data by optimizing these parameters; D is the parameter of the discriminator, and the discriminator accurately distinguishes between real and generated data by optimizing these parameters; For internal quality inspection, non-destructive testing technology is used in combination with the time series convolution network in deep learning for data modeling and analysis, dynamically tracking the changing trend of the panel during the production process. Each convolution kernel The weighted features of each time step in the time series are obtained through convolution operation with the input features. After these features are processed by a multi-layer convolutional network, a result of evaluating the current internal quality status of the road panel is finally output: ,in, is the predicted value at time step t; is the convolution kernel weight; TCN is the temporal convolution operation, which extracts the data features at that moment; is the input feature at time step c-1; N is the length of the time series; All inspection data are used to generate quality scores for panels through a multi-dimensional weighted evaluation model, and electronic labels with complete quality information are generated: ,in, Score the overall quality of the road panels; , is the weight of each quality indicator; The score of the pth quality indicator; is the score of the qth deep feature; m is the number of quality indicators; v is the number of deep features.

5. The method for digital intelligent control and optimization construction of prefabricated pavement plate replacement according to claim 1 is characterized in that: The hoisting path planning in step S3) specifically includes: S31: Through a multi-objective optimization method combining reinforcement learning and particle swarm optimization algorithm, the movement trajectory of the lifting equipment, the size and weight of the road panel and the distribution of obstacles on site are considered to optimize the lifting path, ensure that the equipment avoids obstacles and reduce the energy consumption and time of equipment movement: ,in, It represents the minimum distance between the lifting equipment and the obstacle in the u-th lifting trajectory; , are the lateral and longitudinal offsets of the u-th device path respectively; the optimization objective function is achieved by adjusting the weight coefficient To balance different optimization objectives and achieve optimal adjustment of the path; T is the number of lifting trajectory segments; S32: Digital twin technology is used to build a virtual 3D model synchronized with the construction site, and the Monte Carlo method is used to conduct risk assessment and uncertainty analysis of the lifting process. The Monte Carlo method assesses risks through random simulation of multiple scenarios: ,in, is the estimated probability of the potential risk; is the result of the mth simulation; is the defined failure region, Indicates when It is 1 when it belongs to the failure area, otherwise it is 0; the failure area refers to the area that exceeds any one of the maximum allowable inclination angle, the maximum allowable displacement error and the maximum force of the suspension rope; O represents the total number of simulations; S33: For hoisting precision control, the Kalman filter algorithm is combined with the RTK or UWB positioning system and high-precision sensors to fuse sensor data in real time to accurately estimate the position and attitude of the road panel: ,in, is the estimated value at time k; is the Kalman gain; is the current observation value, is the observation matrix, the initial observation matrix = ; , indicating that the initial road panel is located at a certain position on the site, the xy axis in its three-dimensional coordinates is 0, and the initial height is 3 meters; in addition The initial attitude of the road panel is also set as , indicating that the panel is not tilted at the initial time, and the initial Kalman gain ; During the hoisting process, the real-time monitoring system combines with the anomaly detection algorithm to dynamically identify potential hoisting risks. When the sensor data corresponding to the rope force and the road panel posture exceeds the safety threshold, the hoisting risk is evaluated through the hybrid model of support vector machine and anomaly detection algorithm. ,in, is the prediction result of anomaly detection; is the current sensor data; is the mean of the data; is the preset safety threshold; the difference between the current sensor data and the mean exceeds the threshold When it is abnormal, it indicates that there is an abnormality and an early warning is issued immediately.

6. The method for digital intelligent control and optimized construction of prefabricated pavement plate replacement according to claim 1 is characterized in that: The step S4) is specifically as follows: After the pavement slab is paved, a laser scanner or structured light sensor carried by a mobile robot is used to quickly scan the pavement slab and intelligently detect key indicators; the key indicators include at least flatness, height difference, and the width of the gap between slabs; Through long short-term memory network, the scanning data is analyzed in time series to capture the dynamic change trend of the pavement surface flatness over time, and the future pavement quality is predicted and optimized: ; in, It is the forget gate, which determines the proportion of information that needs to be discarded at present; It is the input gate, which determines the proportion of information that needs to be updated at present; It is the output gate, which determines the proportion of information read from the cell state at the current moment; is the cell state, which is updated according to the output of the forget gate and the input gate; is the hidden state, the network output at the current moment, which is used to pass to the LSTM unit at the next moment; Represents the hidden state of the previous moment and the input data of the current moment; is the forget gate bias vector, which is used to adjust the output of the forget gate; is the input gate bias vector, which determines the information that should be updated to the unit state at the current moment; The output gate bias vector determines which information in the current cell state will be passed to the hidden state at the next moment; is the cell state bias vector, used to update the current cell state; , , , The weight matrix corresponding to the forget gate, input gate, output gate and unit state update; In terms of contact quality detection, the acoustic signal collected by tapping the road panel is obtained, and the propagation characteristics of the sound wave are analyzed using the decision tree algorithm to intelligently identify the contact status between the road panel and the base layer; The decision tree algorithm achieves classification by continuously dividing the feature space and constructing a tree-like decision model. ,in, is the sample set of the current node, is the number of categories, The sample belongs to Probability of class; The decision tree model is trained to automatically identify hollowing or virtual adhesion between the pavement slab and the base layer based on acoustic signals. Laser scanning, acoustic detection and image recognition technology are combined, and a neural network algorithm is used to conduct a comprehensive analysis of multi-source data. The optimization goal is to minimize the error, and the parameters are updated through the back propagation algorithm to comprehensively evaluate the pavement quality of the pavement slab and generate a test report: ,in, is the loss function; is the learning rate; is the network parameter.

7. The method for digital intelligent control and optimized construction of prefabricated pavement plate replacement according to claim 1 is characterized in that: The step S5) is specifically as follows: During the leveling process of the road slab, the reserved grouting holes are first automatically identified through laser scanners or image recognition technology, and the elevation deviations of various areas are analyzed in combination with the previous test data; On this basis, the intelligent grouting equipment automatically mixes, stirs and transports the slurry to ensure the uniformity and accuracy of grouting; Vibration sensors and pressure sensors monitor the grouting diffusion and filling conditions in real time, and dynamically optimize the grouting model based on the monitoring data to ensure that the grouting process is efficient and stable; ; in, is the real-time grouting height; Fill the height for the target; is the real-time diffusion radius; is the maximum allowable diffusion radius; , are the changes in flow rate and pressure respectively; , , , is the weight coefficient.

8. The method for digital intelligent control and optimized construction of prefabricated pavement plate replacement according to claim 1 is characterized in that: The step S6) is specifically as follows: The width and depth of the gap are automatically identified and measured through machine vision technology, and the filling demand is predicted using an improved extreme learning machine algorithm to automatically calculate the most suitable filling material ratio and amount; the extreme learning machine algorithm accelerates the training process by randomly generating hidden layer nodes and optimizing the output layer weights: , ; in, is the hidden layer output matrix; , , …, is the input sample; For the hidden layer The output of the node, is the number of samples; the output of each node is determined by the activation function to decide; is the output layer weight, yes The transposed matrix of is the identity matrix, is the regularization parameter, is the desired output, i.e., filling demand; During the filling process, the system uses fuzzy adaptive control technology to dynamically adjust the grouting flow, pressure and slurry ratio; the fuzzy controller makes adjustments based on real-time sensor data: ,in, Output for grouting control, is the fuzzy membership function, is the gain coefficient of the fuzzy control rule, To control the change in input, is the number of fuzzy rules; After the filling is completed, the filling effect of the gap is monitored in real time by optical rheometer and laser measurement technology to ensure the filling uniformity and quality; and image recognition technology is used to check whether there are bubbles and cracks.

9. The method for digital intelligent control and optimized construction of prefabricated pavement plate replacement according to claim 1 is characterized in that: The step S7) is specifically as follows: During the later operation and maintenance of the road panels, the intelligent sensor network is responsible for collecting health monitoring data of strain, displacement, and vibration, and transmitting them to the cloud platform in real time through the wireless network; Based on hypergraph algorithm and multimodal information fusion technology, sensor data is deeply analyzed. ,in, Represents the total relationship between nodes; Represents a node set; represents the edge set, which represents the relationship between nodes; represents a hyperedge set; The autoregressive integrated moving average model is used for prediction. The autoregressive and moving average methods are used to accurately capture the trend of panel status changes, predict potential structural problems, and issue early warnings in a timely manner: ,in, is the model parameter at the current moment, is the gradient of the loss function at the current moment, is the learning rate.

Images