Municipal road construction quality tracking monitoring management platform

Abstract

The invention relates to the technical field of municipal construction quality tracking, in particular to a municipal road construction quality tracking monitoring management platform. The system comprises a road surface data monitoring unit, and the road surface data monitoring unit monitors road surface basic data based on a sensor monitoring module; the construction monitoring unit is used for monitoring the asphalt penetration, the operation track of the construction machinery and the vibration frequency of the construction machinery; performing multi-source data fusion on the monitored data based on a multi-source fusion algorithm; constructing a comprehensive risk scoring model based on the fused comprehensive features; and the road risk early warning unit analyzes the road construction quality and performs risk early warning through a multi-modal road analysis model based on the road surface basic data, the asphalt penetration and the mechanical parameters monitored by the construction monitoring unit in real time. A crack-construction-environment dynamic correlation model is constructed, the dynamic interaction strength of crack characteristics and construction parameters is calculated through a cross attention mechanism, and a high-risk area is accurately positioned.

Description

Technical Field

[0001] The present invention relates to the technical field of municipal construction quality tracking, and specifically, to a quality tracking and monitoring management platform for municipal road construction. Background Art

[0002] The quality tracking and monitoring management platform for municipal road construction is an information system integrating various monitoring technologies and data analysis methods. It uses sensors, monitoring devices, etc. to collect various data during the municipal road construction process in real time, and through means such as data processing, analysis, and evaluation, realizes the tracking, monitoring, and management of road construction quality.

[0003] During the construction of newly built, rebuilt, or expanded municipal roads, the construction quality is monitored in real time through the quality tracking and monitoring management platform for municipal road construction. However, the existing monitoring of municipal road construction quality relies on manual sampling inspections, lacks full-process automated tracking technology, is prone to construction parameter deviations due to human errors, and ignores the multi-factor coupling effect, resulting in insufficient correlation between construction parameters and road quality, and a large deviation between the risk assessment results and the actual situation, leading to poor risk positioning. In view of this, a quality tracking and monitoring management platform for municipal road construction is designed. Summary of the Invention

[0004] The purpose of the present invention is to provide a quality tracking and monitoring management platform for municipal road construction to solve the problems of ignoring the multi-factor coupling effect, insufficient correlation between construction parameters and road quality, large deviation between risk assessment results and the actual situation, and poor risk positioning proposed in the above background art.

[0005] To achieve the above purpose, the present invention aims to provide a quality tracking and monitoring management platform for municipal road construction, including:

[0006] A road surface data monitoring unit, which monitors the road surface basic data based on a sensor monitoring module and transmits the monitored road surface basic data to an edge node calculation module through a LoRa gateway, and the edge node calculation module preprocesses the road surface basic data;

[0007] A construction monitoring unit, which is used to monitor the asphalt penetration, the working track of construction machinery, and the vibration frequency of construction machinery;

[0008] A cloud data processing unit, which performs multi-source data fusion on the data monitored by the road surface data monitoring unit and the construction monitoring unit based on a multi-source fusion algorithm;

[0009] A risk assessment unit, which introduces the water accumulation depth through a cross-attention mechanism based on the crack propagation probability, the final characteristics, and the settlement risk value and traffic flow Perform feature fusion, construct a comprehensive risk scoring model based on the fused comprehensive features, and quantitatively analyze the influencing factors of pavement basic data by the comprehensive risk scoring model;

[0010] A road risk warning unit, which analyzes the road construction quality and issues risk warnings through a multi-modal road analysis model based on the pavement basic data, asphalt penetration, and mechanical parameters real-time monitored by the construction monitoring unit.

[0011] As a further improvement of this technical solution, the pavement basic data includes road settlement data, pavement crack data, and pavement compaction data;

[0012] Among them, the road settlement data includes at least the road settlement amount;

[0013] The pavement crack data includes at least the crack length, crack width, and crack depth.

[0014] As a further improvement of this technical solution, the construction monitoring unit includes a raw material quality monitoring module and a mechanical detection module;

[0015] Among them, the raw material quality monitoring module is used to detect the asphalt penetration;

[0016] The mechanical detection module is used to real-time monitor the operation trajectory of construction machinery and the vibration frequency of construction machinery.

[0017] As a further improvement of this technical solution, the cloud data processing unit includes a crack feature fusion module, a construction parameter fusion module, and a multi-source fusion module;

[0018] Among them, the crack feature fusion module is used to perform multi-source data fusion on the pavement crack data monitored by the pavement data monitoring unit;

[0019] The construction parameter fusion module is used to perform multi-source data fusion on the asphalt penetration, operation trajectory of construction machinery, and vibration frequency of construction machinery monitored by the construction monitoring unit;

[0020] The multi-source fusion module dynamically fuses the crack tensor with the construction parameter tensor , and introduces the water accumulation depth and traffic flow , to generate comprehensive risk features .

[0021] As a further improvement of this technical solution, the crack feature fusion module is used to perform multi-source data fusion on the pavement crack data monitored by the pavement data monitoring unit. The specific steps involved are:

[0022] Generate a three-dimensional point for each sampling point , since the road crack is in an irregular shape, the arc length parameter is used to represent the crack path, and the crack path equation is ;

[0023] In the formula, represents the position of the th sampling point, the position along the crack extension direction, reflecting the longitudinal distribution of the crack; represents the width of the th sampling point, the lateral width perpendicular to the crack length, reflecting the lateral expansion of the crack; represents the depth of the th sampling point, the longitudinal depth perpendicular to the road surface, reflecting the penetration degree of the crack; where, , and ; represents the sampling interval;

[0024] Then the three-dimensional point cloud coordinates of the road crack are:

[0025] ;

[0026] In the formula, , , , represents the total length of the crack; represents the arc length position of the th sampling point, the cumulative length from the starting point to this point along the crack extension path; represents at the arc length position , the axis coordinate of the crack path in the road surface plane; represents at the arc length position , the axis coordinate of the crack path in the road surface plane; represents the arc length sampling interval; represents the three-dimensional coordinates of the th sampling point in the crack three-dimensional point cloud;

[0027] All sampling points form a three-dimensional point cloud set:

[0028] , ;

[0029] In the formula, represents the last sampling point in the crack three-dimensional point cloud set; represents the total number of sampling points on the crack path; represents the three-dimensional point cloud set formed by all sampling points;

[0030] Convert the 3D point cloud set into a 128-dimensional crack feature vector based on the PointNet++ model , realizing the feature fusion of pavement crack data.

[0031] As a further improvement of this technical solution, the construction parameter fusion module is used to perform multi-source data fusion on the asphalt penetration, construction machinery operation trajectory, and construction machinery vibration frequency monitored by the construction monitoring unit. The specific steps involved are as follows:

[0032] Define the mechanical trajectory node feature vector and the neighborhood ;

[0033] Based on the learnable weight matrix map the mechanical trajectory node feature vector to a higher-order space ;

[0034] Calculate the unnormalized attention score through the spatio-temporal attention mechanism , and at the same time introduce the time decay function and the space decay function to quantify the dynamic influence weight of the vibration parameters on the compaction quality of specific trajectory points:

[0035] ;

[0036] In the formula, represents vector concatenation; represents the activation function; represents the time decay function; represents the space decay function; represents the time interval when the construction machinery reaches the trajectory node and the trajectory node ; represents the lateral coordinate distance interval between the construction machinery reaching the trajectory node and the trajectory node ; represents the feature vector of the trajectory node after being mapped by the learnable weight matrix to the higher-order space representation, where and the trajectory node belongs to the neighborhood of the trajectory node ; represents the unnormalized attention coefficient ;

[0037] Among them:

[0038] ​

[0039]

[0040] In the formula, represents the attenuation coefficient; represents the spatial sensitivity coefficient;

[0041] Generate a normalized attention coefficient based on the Softmax function ;

[0042] Where:

[0043]

[0044] In the formula, represents; represents;

[0045] Fill the attention coefficient into the sparse matrix to generate a spatio-temporal correlation matrix . The spatio-temporal correlation matrix is used as the edge weight input of the graph neural network (GNN) to support the dynamic traceability of quality defects, and the water accumulation depth and traffic flow are introduced as modulation parameters to obtain an optimized spatio-temporal correlation matrix :

[0046]

[0047] In the formula, represents the Sigmoid function, and the output range is , representing the amplification / suppression effect of the environment on the correlation strength; represents a learnable parameter matrix, and .

[0048] As a further improvement of this technical solution, the multi-source fusion module dynamically fuses the crack tensor with the construction parameter tensor based on the cross-attention mechanism, and introduces the water accumulation depth and traffic flow to generate a comprehensive risk feature . The specific steps involved are:

[0049] Expand the crack feature vector into a three-dimensional tensor to match the spatio-temporal dimensions of the construction parameter matrix, and obtain the crack tensor ;

[0050] Perform feature encoding on the optimized spatio-temporal correlation matrix to obtain the construction parameter tensor ;

[0051] Dynamically fuse the crack tensor based on the cross-attention mechanism with the construction parameter tensor , and obtain the construction parameter features after attention weighting:

[0052]

[0053] Where: ; ; ;

[0054] In the formula, is the query vector, representing the query demand for three-dimensional crack features; is the key vector, representing the index features of construction parameters; is the value vector, representing the actual influence value of construction parameters; represents the sensitivity of three-dimensional crack features to construction parameters; represents the matchable features of construction parameters; represents the actual influence strength of construction parameters; represents the dimension of the key vector; represents the construction parameter features after attention weighting; represents the correlation strength between three-dimensional crack features and construction parameters;

[0055] Based on the residual connection, add the crack tensor to the construction parameter features after attention weighting , and perform layer normalization to generate the comprehensive risk feature :

[0056]

[0057] In the formula, represents the comprehensive risk feature; represents layer normalization;

[0058] Fuse the water accumulation depth and the traffic flow with the comprehensive risk feature :

[0059]

[0060] In the formula, represents the initial feature vector of node , and ; represents the water accumulation depth encoder; represents the traffic flow encoder; represents the bias term; represents the weight matrix; Denotes the activation function.

[0061] As a further improvement of this technical solution, the risk assessment unit includes a crack risk prediction module, a settlement prediction module, a multi-order risk transfer module, and a comprehensive road analysis module;

[0062] Among them, the crack risk prediction module constructs a crack propagation risk probability model based on the crack feature vector and the comprehensive risk feature :

[0063]

[0064] In the formula, Denotes the Sigmoid function; Is the weight matrix; Is the bias term; Denotes vector concatenation; Is the crack propagation probability, ;

[0065] The settlement prediction module obtains the settlement prediction value based on the LSTM-ARIMA hybrid model, and constructs a settlement risk prediction model through the settlement prediction value and the influence of construction parameters to dynamically correct the settlement risk. Then the settlement risk prediction model is:

[0066]

[0067] In the formula, Denotes the settlement prediction value at time ; Denotes the weight coefficient of the predicted settlement amount, Denotes the weight coefficient of the influence of construction parameters, And Are both obtained by fitting historical data; Denotes the settlement amount The partial derivative with respect to temperature ; Denotes the settlement amount The partial derivative with respect to the vibration frequency ; Denotes the settlement risk value; Denotes the node Belongs to the spatio-temporal neighborhood set of the node ;

[0068] The multi-order risk transfer module propagates crack and settlement risks based on a graph neural network, and introduces the water depth And traffic flow To construct a multi-risk transfer model:

[0069] ;

[0070] In the formula, represents the node at the layer's hidden state; represents the node update weight matrix; represents the learnable parameter; represents the residual mapping matrix, which maps the environmental factor and to the node feature space; represents the layer normalization operation, which is used to stabilize the training process and prevent gradient explosion / vanishing; represents the node at the layer-aggregated neighborhood message; represents the node at the layer's hidden state; represents the number of GNN propagation layers;

[0071] Where:

[0072]

[0073] In the formula, represents element-wise multiplication; represents the node at the layer's hidden state; represents the environmental gating vector;

[0074] And ;

[0075] In the formula, represents the dimension of the environmental gating vector , that is, the length of the gating vector.

[0076] As a further improvement of this technical solution, the comprehensive road analysis module comprehensively conducts the final road risk prediction and assessment based on the multi-modal road analysis model. The specific steps involved are:

[0077] Normalize the crack propagation probability , the final feature after layers of propagation, and the dynamically corrected settlement risk value ;

[0078] Fuse the crack propagation probability , the final feature , and the settlement risk value after normalization through the cross-attention mechanism:

[0079]

[0080] Among them, represents the environmental factor encoding, , among which, is the waterlogging depth and represents the traffic flow, represents the long-term effect of waterlogging and traffic flow; represents the fused comprehensive feature;

[0081] Combine with to form a 128-dimensional vector;

[0082] Map through the fully connected layer to 128 dimensions;

[0083] Output the comprehensive feature by the activation function ; ;

[0084] Based on the fused comprehensive feature , generate a comprehensive risk scoring model:

[0085] ;

[0086] Introduce the coupling term of crack risk and construction parameters into the comprehensive risk scoring model:

[0087] ;

[0088] In the formula, represents the parameters of the fusion layer; represents the scoring mapping parameters; represents the bias term of the fusion layer; represents the bias term of the scoring mapping; represents the coupling strength coefficient; represents the sensitivity of settlement to vibration frequency; represents the contribution of environmental factors to long-term risk; represents the dynamic interaction strength between crack risk and construction parameters; represents the comprehensive risk score after introducing the coupling term of crack risk and construction parameters, where 0 means no risk and 1 means extremely high risk.

[0089] As a further improvement of this technical solution, the road risk warning unit includes a real-time scoring module and a risk warning module;

[0090] Among them, the real-time scoring module generates a comprehensive risk score based on the real-time monitored road surface basic data, asphalt penetration, and mechanical parameters through a comprehensive risk scoring model, and the risk warning module triggers risk warnings at different levels using a segmented threshold algorithm.

[0091] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0092] 1. In this quality tracking and monitoring management platform for municipal road construction, road settlement, crack morphology, construction parameters, and environmental factors are integrated to construct a dynamic correlation model of cracks-construction-environment, avoiding misjudgment of complex problems (such as the problem of "water accumulation + excessive vibration" accelerating subgrade damage). Through the cross-attention mechanism, the dynamic interaction intensity between crack characteristics and construction parameters is calculated to accurately locate high-risk areas.

[0093] 2. In this quality tracking and monitoring management platform for municipal road construction, the settlement trend is predicted through the LSTM-ARIMA hybrid model, and the real-time influence weight of mechanical operations on cracks is quantified by combining the spatio-temporal attention mechanism to achieve dynamic risk scoring. Considering the long-term hidden impact of water accumulation depth and traffic flow on construction quality, the dynamic management of the whole life cycle of construction quality is realized. Description of the Drawings

[0094] Figure 1 It is the overall process block diagram of the present invention.

[0095] The meanings of each label in the figure are as follows:

[0096] 1. Road surface data monitoring unit; 11. Sensor monitoring module; 12. Edge node calculation module;

[0097] 2. Construction monitoring unit; 21. Raw material quality monitoring module; 22. Mechanical detection module;

[0098] 3. Cloud data processing unit; 31. Crack feature fusion module; 32. Construction parameter fusion module; 33. Multi-source fusion module;

[0099] 4. Risk assessment unit; 41. Crack risk prediction module; 42. Settlement prediction module; 43. Multi-stage risk transfer module; 44. Comprehensive road analysis module;

[0100] 5. Road risk warning unit; 51. Real-time scoring module; 52. Risk warning module. Detailed Embodiment

[0101] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0102] Please refer to Figure 1 As shown, a quality tracking and monitoring management platform for municipal road construction is provided, including a road surface data monitoring unit 1. The road surface data monitoring unit 1 monitors the road surface basic data based on the sensor monitoring module 11, and transmits the monitored road surface basic data to the edge node calculation module 12 through the LoRa gateway, and the edge node calculation module 12 preprocesses the road surface basic data;

[0103] The road surface basic data includes road settlement data, road surface crack data, and road surface compaction data;

[0104] Among them, the road settlement data at least includes the road settlement amount; the road settlement data is monitored by arranging a static level (accuracy ±0.1mm);

[0105] The road surface crack data at least includes the crack length, crack width, and crack depth.

[0106] In this embodiment, the preprocessing methods for the road surface basic data (road settlement data, road surface crack data, and compaction) are as follows:

[0107] The sliding window adaptive filtering algorithm is used to preprocess the road settlement data:

[0108] Median filtering with a dynamic window size is performed on the road settlement time series data to eliminate impulse noise;

[0109] The window size is automatically adjusted through the variance threshold to balance noise suppression and signal fidelity;

[0110] The road surface crack data includes three-dimensional crack characteristics (crack length, crack width, and crack depth);

[0111] The wavelet packet multi-scale decomposition algorithm is used to preprocess the three-dimensional crack data:

[0112] Perform 5-layer db4 wavelet decomposition on the crack detection signal (vibration / image eigenvalue), and use the SURE threshold method to retain the effective crack characteristics greater than 0.5mm;

[0113] The edge node computing module 12 preliminarily processes and analyzes the data locally, extracts valuable information, and only uploads the processed data and key information to the cloud data processing unit 3. This can effectively reduce the data transmission volume, improve the data transmission efficiency, and at the same time relieve the computing and storage pressure on the cloud, enabling the cloud to perform complex operations such as multi-source data fusion more efficiently;

[0114] In this embodiment, the municipal road construction quality tracking and monitoring management platform further includes a construction monitoring unit 2, and the construction monitoring unit 2 is used to monitor the asphalt penetration, the operation trajectory of construction machinery, and the vibration frequency of construction machinery;

[0115] The construction monitoring unit 2 includes a raw material quality monitoring module 21 and a machinery detection module 22;

[0116] Among them, the raw material quality monitoring module 21 is used to detect the asphalt penetration;

[0117] The machinery detection module 22 is used to monitor the operation trajectory of construction machinery and the vibration frequency of construction machinery in real time.

[0118] In this embodiment, the municipal road construction quality tracking and monitoring management platform further includes a cloud data processing unit 3, and the cloud data processing unit 3 performs multi-source data fusion on the data monitored by the road surface data monitoring unit 1 and the construction monitoring unit 2 based on a multi-source fusion algorithm;

[0119] Specifically, the cloud data processing unit 3 includes a crack feature fusion module 31, a construction parameter fusion module 32, and a multi-source fusion module 33;

[0120] Among them, the crack feature fusion module 31 is used to perform multi-source data fusion on the road surface crack data monitored by the road surface data monitoring unit 1;

[0121] The construction parameter fusion module 32 is used to perform multi-source data fusion on the asphalt penetration, the operation trajectory of construction machinery, and the vibration frequency of construction machinery monitored by the construction monitoring unit 2;

[0122] The multi-source fusion module 33 dynamically fuses the crack tensor with the construction parameter tensor , and since the synergistic effect of the water accumulation depth and the crack depth can accelerate the settlement risk, the water accumulation depth and the traffic flow are introduced to generate a comprehensive risk feature .

[0123] In this embodiment, the crack feature fusion module 31 is used to perform multi-source data fusion on the road surface crack data monitored by the road surface data monitoring unit 1. The specific steps involved are:

[0124] Generate a three-dimensional point for each sampling point , since road cracks are irregular in shape, the arc length parameter is used to represent the crack path, that is, the cumulative length from the crack starting point to the current position. Then the crack path equation is , and from the crack path equation represents the two-dimensional coordinates of the crack in the road surface plane;

[0125] In the formula, represents the position of the th sampling point, the position along the crack extension direction, reflecting the longitudinal distribution of the crack; represents the width of the th sampling point, the lateral width perpendicular to the crack length, reflecting the lateral expansion of the crack; represents the depth of the th sampling point, the longitudinal depth perpendicular to the road surface, reflecting the penetration degree of the crack; among them, , and ; represents the sampling interval;

[0126] Then the three-dimensional point cloud coordinates of the road crack are:

[0127] ;

[0128] In the formula, , , , represents the total length of the crack; represents the arc length position of the th sampling point, the cumulative length from the starting point to this point along the crack extension path; represents at the arc length position , the axis coordinate of the crack path in the road surface plane; represents at the arc length position , the axis coordinate of the crack path in the road surface plane; represents the arc length sampling interval; represents the three-dimensional coordinates of the th sampling point in the crack three-dimensional point cloud;

[0129] All sampling points form a three-dimensional point cloud set:

[0130] , ;

[0131] In the formula, represents the last sampling point in the crack three-dimensional point cloud set; Indicates the total number of sampling points on the crack path; Indicates that all sampling points form a three-dimensional point cloud set;

[0132] Form a point sequence in three-dimensional space along the length, width, and depth of the crack to reflect the morphological characteristics of the crack;

[0133] Based on the PointNet++ model, convert the three-dimensional point cloud set into a 128-dimensional crack feature vector , realizing the feature fusion of pavement crack data.

[0134] In this embodiment, the PointNet++ model maps the point cloud to a 128-dimensional vector through hierarchical feature extraction. The specific steps are as follows:

[0135] Input the three-dimensional point cloud set through the input layer ; Indicates the total number of sampling points;

[0136] Based on farthest point sampling (FPS), select key points ;

[0137] With each key point as the center, aggregate the local neighborhood point set with a radius of ; ;

[0138] For each local point set Use a multi-layer perceptron (MLP) to extract local features ;

[0139] Repeat the above steps, gradually expanding the receptive field, and finally output the global feature vector .

[0140] Furthermore, the construction parameter fusion module 32 is used to perform multi-source data fusion on the asphalt penetration, construction machinery operation trajectory, and construction machinery vibration frequency monitored by the construction monitoring unit 2. The specific steps involved are as follows:

[0141] Define the mechanical trajectory node feature vector and the neighborhood ;

[0142] For the feature vector of each trajectory node , the feature vector , specifically ;

[0143] Neighborhood condition: The trajectory node belongs to the neighborhood of the trajectory node , if and only if it satisfies:

[0144] Spatial proximity: ;

[0145] Temporal proximity: ;

[0146] Wherein, represents the lateral coordinate of the trajectory point of the construction machinery to locate the position of the construction machinery in the lateral direction of the road; represents the lateral coordinate of the trajectory point of the construction machinery ; represents the longitudinal coordinate of the trajectory point of the construction machinery ; represents the time when the construction machinery reaches the trajectory point to associate with the sequential construction events and capture the dynamic changes of compaction quality; represents the time when the construction machinery reaches the trajectory point ; represents the real-time vibration frequency of the construction machinery at the trajectory point to quantify the compaction energy input and prevent over-compaction or under-compaction; represents the traveling speed of the construction machinery at the trajectory point ; represents the real-time temperature of the asphalt material to ensure that the compaction operation is completed within the effective temperature window; represents the penetration of the asphalt material to evaluate the performance of the asphalt material and guide the adaptation of vibration parameters; represents the spatial neighborhood radius to limit the spatial influence range of the vibration effect, reduce invalid calculations, and usually takes a value of 3 - 5 meters, indicating that only the mutual influence of construction events within this range is considered; represents the time window to constrain the temporal relevance of construction parameters and match the material curing process, and usually takes a value of 10 - 30 seconds, indicating that only the construction event association within this time range is considered; the temperature of the asphalt mixture decreases with time, and the compaction effect is limited by the time window;

[0147] Based on the learnable weight matrix map the mechanical trajectory node feature vector to a higher-order space ;

[0148] Calculate the unnormalized attention score through the spatio-temporal attention mechanism , and at the same time introduce the time decay function and the space decay function to quantify the dynamic influence weight of the vibration parameters on the compaction quality of a specific trajectory point:

[0149] ;

[0150] Wherein, represents vector concatenation; represents an activation function, introducing non-linearity, allowing negative gradient transmission, and preventing neuron death; represents a time decay function; represents a spatial decay function; represents the arrival of construction machinery at a trajectory node and the time interval to reach the trajectory node ; represents the arrival of construction machinery at a trajectory node and the lateral coordinate distance interval from the trajectory node ; represents the feature vector of the trajectory node obtained by mapping through a learnable weight matrix to a high-order spatial representation, where , the trajectory node belongs to the neighborhood of the trajectory node ; represents the unnormalized attention coefficient ; ;

[0151] Among them:

[0152]

[0153]

[0154] In the formula, represents the decay coefficient; represents the spatial sensitivity coefficient;

[0155] Generate a normalized attention coefficient based on the Softmax function ;

[0156] Among them:

[0157]

[0158] In the formula, represents; represents;

[0159] Fill the attention coefficient into the sparse matrix to generate a spatio-temporal correlation matrix , and the spatio-temporal correlation matrix is used as the edge weight input of the graph neural network (GNN) to support the dynamic traceability of quality defects, and introduce the water depth and traffic flow as modulation parameters to obtain the optimized spatio-temporal correlation matrix :

[0160]

[0161] In the formula, represents the Sigmoid function, and the output range , representing the amplification / suppression effect of the environment on the correlation strength; represents the learnable parameter matrix, and .

[0162] Among them:

[0163]

[0164] In the formula, represents the dynamic influence strength of the vibration effect of the trajectory node on the compaction quality at the trajectory node during the construction process.

[0165] Specifically, the multi-source fusion module 33 dynamically fuses the crack tensor and the construction parameter tensor , and introduces the water accumulation depth and the traffic flow to generate the comprehensive risk feature . The specific steps involved are:

[0166] Expand the crack feature vector into a three-dimensional tensor, match the spatio-temporal dimension of the construction parameter matrix, and obtain the crack tensor ; the crack tensor is an abstract representation of the morphological features of the crack in space;

[0167] Among them,

[0168]

[0169] Perform feature encoding on the optimized spatio-temporal correlation matrix to obtain the construction parameter tensor ;

[0170]

[0171] Among them, is a two-layer fully connected network: ;

[0172] Unify the feature dimensions of the crack and the construction parameters to provide a compatible input space for multi-modal fusion;

[0173] Based on the cross-attention mechanism, dynamically fuse the crack tensor and the construction parameter tensor to obtain the attention-weighted construction parameter features:

[0174]

[0175] Among them: ; ; ;

[0176] In the formula, is the query vector, representing the query requirements for three-dimensional fracture characteristics; is the key vector, representing the index characteristics of construction parameters; is the value vector, representing the actual influence value of construction parameters; represents the sensitivity of three-dimensional fracture characteristics (fracture depth, width and length) to construction parameters (real-time vibration frequency of construction machinery at trajectory point , travel speed of construction machinery at trajectory point and real-time temperature of asphalt material); represents the matchable characteristics of construction parameters, that is, which characteristics in construction parameters need to be associated and matched with fracture characteristics; represents the actual influence strength of construction parameters, that is, the specific contribution value of construction parameters to fracture risk; represents the key vector dimension; represents the construction parameter characteristics after attention weighting; represents the association strength between three-dimensional fracture characteristics and construction parameters;

[0177] Through the learnable parameter matrix , the fracture tensor is mapped into the query vector , and the construction parameter tensor is mapped into the key vector and the value vector ;

[0178] Calculate the association strength matrix between the query vector and the key vector , and generate attention weights through the scaling factor and the function; perform weighted summation on the value vector to obtain the construction parameter characteristics after attention weighting;

[0179] In this embodiment, for the fracture depth mutation region ( with high value), it is necessary to query whether there is a construction event with abnormal vibration frequency in this region;

[0180] The vibration frequency of the mechanical trajectory point ( with high value) is used to match the fracture propagation requirements and identify which construction characteristics are related to the fracture;

[0181] High-frequency vibration ( with a high value) represents the actual impact strength of the construction event on crack propagation;

[0182] Based on the residual connection, the crack tensor is added to the construction parameter features after attention weighting and, after layer normalization, a comprehensive risk feature is generated:

[0183]

[0184] In the formula, represents the comprehensive risk feature; represents layer normalization; the correlation between the crack morphology and the construction machinery operation is captured through the attention mechanism (such as whether the area with abnormal vibration frequency overlaps with the area with sudden change in crack depth);

[0185] The water accumulation depth and traffic flow are feature-fused with the comprehensive risk feature :

[0186]

[0187] In the formula, represents the initial feature vector of node and ; represents the water accumulation depth encoder (3-layer MLP, output dimension: 32), characterizing the influence of moisture on material properties; represents the traffic flow encoder (3-layer MLP, output dimension: 32), characterizing the fatigue effect of long-term load on the road surface; represents the bias term; represents the weight matrix; represents the activation function.

[0188] The quality tracking and monitoring management platform for the municipal road construction further includes a risk assessment unit 4. The risk assessment unit 4, based on the crack propagation probability, final features, and settlement risk values, through the cross-attention mechanism, introduces the water accumulation depth and traffic flow to perform feature fusion, constructs a comprehensive risk scoring model based on the fused comprehensive features, and quantitatively analyzes the influencing factors of the road surface basic data by the comprehensive risk scoring model;

[0189] In this embodiment, the risk assessment unit 4 includes a crack risk prediction module 41, a settlement prediction module 42, a multi-order risk transfer module 43, and a comprehensive road analysis module 44;

[0190] Among them, the crack risk prediction module 41 is based on the crack feature vector With the comprehensive risk characteristics Construct a risk probability model for crack propagation:

[0191]

[0192] In the formula, represents the Sigmoid function; is the weight matrix; is the bias term; represents vector concatenation; is the crack propagation probability, ;

[0193] The settlement prediction module 42 obtains the settlement prediction value based on the LSTM-ARIMA hybrid model, and constructs a settlement risk prediction model through the settlement prediction value and the influence of construction parameters to dynamically correct the settlement risk. Then the settlement risk prediction model is:

[0194]

[0195] In the formula, represents the time of the settlement prediction value; represents the weight coefficient of the predicted settlement amount, represents the weight coefficient of the influence of construction parameters, and are both obtained by fitting historical data; represents the settlement amount with respect to the temperature partial derivative; represents the settlement amount with respect to the vibration frequency partial derivative; represents the settlement risk value; represents the node belongs to the spatio-temporal neighborhood set of the node ;

[0196] In this embodiment, the LSTM-ARIMA hybrid model is adopted, and the groundwater level, earth pressure, and construction load are used as input parameters to obtain the settlement prediction value , specifically, the LSTM-ARIMA hybrid time series model is as follows: collect groundwater level data, geological condition data, time data, and simultaneously collect the corresponding settlement data as the target variable. Use the groundwater level data, geological condition data, time, and other relevant data as input parameters to predict the linear part of the time series by ARIMA, generating residuals (i.e., the non-linear information not captured by the linear trend). Use the ARIMA residuals as the input of LSTM to train it to capture non-linear features. Finally, combine the outputs of the two models by weighting. Use historical data (historical groundwater level, historical earth pressure, and historical construction load data) to train the LSTM-ARIMA hybrid model. By continuously adjusting the model parameters, such as the number of hidden layer neurons of LSTM, learning rate, etc., to minimize the prediction error, and use the trained model to predict the future settlement volume;

[0197] The multi-order risk transfer module 43 propagates crack and settlement risks based on the graph neural network (GNN), and introduces the waterlogging depth and traffic flow to construct a multi-risk transfer model:

[0198] ;

[0199] In the formula, represents the hidden state of node at the th layer; represents the node update weight matrix; represents the learnable parameter; represents the residual mapping matrix, which maps the environmental factor and to the node feature space; represents the layer normalization operation, which is used to stabilize the training process and prevent gradient explosion / vanishing; represents the node at the th layer aggregates the neighborhood messages; represents the node at the th layer's hidden state; represents the number of GNN propagation layers;

[0200] Among them:

[0201]

[0202] In the formula, represents the element-wise multiplication; represents the hidden state of node at the th layer; represents the environmental gating vector;

[0203] When or increases, the value of the gating vector tends to 1, retaining the risk information of more neighborhood nodes (such as cracks and abnormal vibration frequencies);

[0204] Conversely, if the environmental conditions are good (such as = 0), the gating value tends to 0, suppressing the influence of irrelevant noise on the current node;

[0205] And ;

[0206] In the formula, represents the dimension of the environmental gating vector , that is, the length of the gating vector;

[0207] The optimized multi-risk transfer model realizes the deep interaction between construction parameters and environmental factors, providing a more accurate dynamic risk assessment for road construction quality in complex scenarios.

[0208] In this embodiment, the comprehensive road analysis module 44 comprehensively conducts the final road risk prediction and assessment based on the multi-modal road analysis model. The specific steps involved are as follows:

[0209] Normalize the crack propagation probability , the final feature after passing through layers of propagation, and the dynamically corrected settlement risk value ;

[0210] After passing through layers of propagation, the calculation process of the final feature is as follows:

[0211] The initial feature vector of the initial node fuses the comprehensive risk feature and the environmental factor encoding:

[0212] Among them, ;

[0213] Propagate layer by layer through GNN:

[0214] Aggregate the neighborhood node features and filter the noise through the environmental gating mechanism ;

[0215] Fuse the current feature and the neighborhood message, and inject the environmental factor residual: ;

[0216] Perform layer normalization to stabilize the training and output the updated feature ;

[0217] After K - layer propagation, the final features of the nodes are output by the final layer ; It represents the influence of construction parameters, the association of crack risks, and the environmental modulation effect that integrate the K - hop neighborhood;

[0218] The crack propagation probability after being standardized , the final features and the settlement risk value are subjected to feature fusion through the cross - attention mechanism:

[0219]

[0220] Among them, represents the environmental factor encoding, , among which, is the waterlogging depth and represents the traffic flow, represents the long - term effect of waterlogging and traffic flow; represents the fused comprehensive features;

[0221] Combine with to form a 128 - dimensional vector;

[0222] Map through the fully - connected layer to 128 dimensions;

[0223] The activation function outputs the comprehensive features ;

[0224] Based on the fused comprehensive features , generate a comprehensive risk scoring model:

[0225] ;

[0226] Introduce the coupling term of crack risk and construction parameters into the comprehensive risk scoring model:

[0227] ;

[0228] In the formula, represents the parameters of the fusion layer; represents the scoring mapping parameters; represents the bias term of the fusion layer; represents the bias term of the scoring mapping; represents the coupling strength coefficient (fitted through historical data); represents the sensitivity of settlement to vibration frequency, which is used to quantify the promotion effect of construction vibration on crack propagation; Indicates the coupling effect of explicit quantification of crack propagation and vibration frequency. The larger the value, the more severe the crack risk caused by vibration; Indicates the contribution of environmental factors to long-term risk (e.g., accumulated water accelerating material deterioration, increased traffic flow causing fatigue damage); Indicates the dynamic interaction intensity between crack risk and construction parameters; Indicates the comprehensive risk score after introducing the coupling term of crack risk and construction parameters, where 0 represents no risk and 1 represents extremely high risk.

[0229] In the comprehensive risk scoring model that introduces the coupling term of crack risk and construction parameters, if and vehicles per hour, then the comprehensive risk scoring model explicitly increases the correlation weight between cracks and vibration frequency through the coupling strength coefficient so that the risk score significantly increases;

[0230] Where: Indicates the use of the cross-attention mechanism to model the dynamic interaction of crack risk and construction parameter characteristics, and its mathematical expression is as follows:

[0231] ;

[0232] ;

[0233] ;

[0234] .

[0235] The quality tracking and monitoring management platform for the municipal road construction also includes a road risk warning unit 5. The road risk warning unit 5 analyzes the road construction quality and issues risk warnings through a multi-modal road analysis model based on the pavement foundation data, asphalt penetration, and mechanical parameters real-time monitored by the construction monitoring unit 2.

[0236] Specifically, the road risk warning unit 5 includes a real-time scoring module 51 and a risk warning module 52;

[0237] Among them, the real-time scoring module 51 generates a comprehensive risk score based on the pavement foundation data, asphalt penetration, and mechanical parameters real-time monitored, and the risk warning module 52 triggers risk warnings at different levels using a segmented threshold algorithm;

[0238] In the embodiment, the risk warning module 52 uses a segmented threshold to trigger different-level warnings, and the specific warning steps involved are:

[0239] When the comprehensive risk score Within the time window and continuously exceeding the threshold for a cumulative time reaching , the risk warning module 52 triggers an alarm to avoid false alarms caused by instantaneous noise:

[0240]

[0241] wherein, represents the indicator function, which converts the continuous risk score into a binary signal and simplifies the time accumulation calculation; represents the length of the sliding time window, that is, the monitoring period, which is used to analyze the time range of risk persistence. The larger the window, the stronger the anti-noise ability, but the response delay increases; represents the risk continuous trigger time; represents the warning trigger time point; represents the current time point, the time stamp monitored by the system in real time;

[0242] Meanwhile, when a risk is detected, the high-risk causes are located through the attention weight matrix:

[0243] When the weight , cracks dominate the risk;

[0244] When the contribution degree is 30% of the total score, construction parameters dominate the risk;

[0245] If , the warning level is increased by one level. At this time, the environmental factor amplification effect;

[0246] In this embodiment, the specific warning output and intervention strategy involved in the risk warning module 52 are:

[0247]

[0248] wherein, 1 represents the first-level risk, which is a risk-free state at this time, that is, the normal state; 2 represents the second-level risk; 3 represents the third-level risk; 4 represents the fourth-level risk;

[0249] Among them, the trigger condition for the second-level risk is:

[0250] The local crack expansion probability or the settlement rate exceeds the limit ;

[0251] wherein, represents the settlement rate, which represents the change rate of the settlement amount per unit time and is used to quantify the sinking speed of the subgrade or pavement, reflecting the construction quality or foundation stability; represents the monitoring time span;

[0252] Response measures: Reduce the mechanical vibration frequency , and increase the compaction passes;

[0253] The triggering conditions for level 3 risk are:

[0254] Multiple cracks converge or settlement prediction ;

[0255] Response measures: Suspend construction, inject asphalt slurry to repair cracks, and adjust the ratio (penetration increases by 5%);

[0256] The triggering conditions for level 4 risk are:

[0257] The comprehensive score exceeds the limit and the environment deteriorates (e.g., and vehicles / hour);

[0258] Response measures: Close the traffic, start subgrade reinforcement (pressure grouting volume , reconstruct the construction plan;

[0259] Furthermore, by calculating the partial derivatives of the comprehensive risk score with respect to the input features, the importance of the features is quantified:

[0260]

[0261]

[0262] In the formula, is the importance of the crack feature, representing the partial derivative of the comprehensive risk score with respect to the crack propagation probability ; is the importance of the settlement feature, representing the partial derivative of the comprehensive risk score with respect to the settlement risk value ;

[0263] When , the crack probability increases, which will push up the comprehensive risk score;

[0264] When , the settlement risk decreases, which will reduce the total risk;

[0265] Based on historical data statistics, when , it is determined as a crack-dominated risk at this time; the crack propagation speed is usually 3 to 5 times the settlement rate, so a threshold of 2 is set to early warn of more urgent crack risks, avoid misjudgment due to minor fluctuations, and ensure the capture of significant differences;

[0266] In the case of crack-dominated risks, the following responses are usually adopted:

[0267] Initiate a crack repair plan (such as injecting epoxy resin);

[0268] Adjust the vibration frequency of construction machinery to prevent further cracking;

[0269] Furthermore, it is set that when it is determined as settlement-dominated risk at this time;

[0270] For settlement-dominated risks, the following responses are usually adopted:

[0271] Enhance subgrade compaction;

[0272] Monitor the groundwater level and initiate drainage measures if necessary;

[0273] Reduce the priority of crack monitoring.

[0274] Furthermore, in this implementation, the edge node calculation delay constraint of the edge node calculation module 12 and the cloud computing delay of the cloud data processing unit 3 are used to meet the requirements of real-time construction regulation.

[0275] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. The above embodiments and the descriptions in the specification are only preferred examples of the present invention and are not used to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed. The scope of protection claimed by the present invention is defined by the appended claims and their equivalents.

Claims

1. A municipal road construction quality tracking, monitoring and management platform, characterized in that: include: A road surface data monitoring unit (1), wherein the road surface data monitoring unit (1) monitors basic road surface data based on a sensor monitoring module (11) and pre-processes the basic road surface data; A construction monitoring unit (2), the construction monitoring unit (2) being used to monitor asphalt penetration, construction machinery operation trajectory and construction machinery vibration frequency; A cloud data processing unit (3), wherein the cloud data processing unit (3) performs multi-source data fusion on the data monitored by the road surface data monitoring unit (1) and the construction monitoring unit (2) based on a multi-source fusion algorithm; A risk assessment unit (4), wherein the risk assessment unit (4) introduces water accumulation depth and traffic flow to perform feature fusion based on the crack propagation probability, final characteristics and settlement risk value through a cross-attention mechanism, and constructs a comprehensive risk scoring model based on the fused comprehensive characteristics, so as to quantitatively analyze the influencing factors of the road surface basic data; A road risk warning unit (5) is configured to analyze the road construction quality and issue a risk warning through a multi-modal road analysis model based on the road surface basic data, asphalt penetration and mechanical parameters monitored in real time by the construction monitoring unit (2).

2. The municipal road construction quality tracking, monitoring and management platform according to claim 1 is characterized by: The road surface basic data includes road settlement data, road surface crack data and road surface compaction data; Wherein, the road settlement data at least includes the road settlement amount; Pavement crack data at least includes crack length, crack width and crack depth.

3. The municipal road construction quality tracking, monitoring and management platform according to claim 2 is characterized by: The construction monitoring unit (2) comprises a raw material quality monitoring module (21) and a mechanical detection module (22); Wherein, the raw material quality monitoring module (21) is used to detect the penetration of asphalt; The machinery detection module (22) is used to monitor the operation trajectory of the construction machinery and the vibration frequency of the construction machinery in real time.

4. The municipal road construction quality tracking, monitoring and management platform according to claim 3 is characterized by: The cloud data processing unit (3) comprises a crack feature fusion module (31), a construction parameter fusion module (32) and a multi-source fusion module (33); The crack feature fusion module (31) is used to perform multi-source data fusion on the pavement crack data monitored by the pavement data monitoring unit (1); The construction parameter fusion module (32) is used to fuse multi-source data of asphalt penetration, construction machinery operation trajectory and construction machinery vibration frequency monitored by the construction monitoring unit (2); The multi-source fusion module (33) dynamically fuses crack tensors based on the cross-attention mechanism with the construction parameter tensor , and introduce the depth of water accumulation and traffic flow , generating a comprehensive risk profile .

5. The municipal road construction quality tracking, monitoring and management platform according to claim 4 is characterized by: The crack feature fusion module (31) is used to perform multi-source data fusion on the pavement crack data monitored by the pavement data monitoring unit (1), and the specific steps involved are: For each sample point a 3D point is generated Since the road cracks are irregular in shape, the arc length parameter is used. represents the crack path, then the crack path equation is ; In the formula, Indicates The location of the sampling points; Indicates The width of the sampling points; Indicates The depth of each sampling point; The three-dimensional point cloud coordinates of the road crack are: ； In the formula, , , , represents the total length of the crack; Indicates The arc length position of the sampling points; Indicates the arc length position The crack path is within the pavement plane. Axis coordinates; Indicates the arc length position The crack path is within the pavement plane. Axis coordinates; The arc length is expressed using intervals; Represents the first The three-dimensional coordinates of the sampling points; All sampling points constitute a 3D point cloud set: ， ； In the formula, Indicates the last sampling point in the crack 3D point cloud set; represents the total number of sampling points on the crack path; Indicates that all sampling points constitute a three-dimensional point cloud set; Based on the PointNet++ model, a 3D point cloud set is converted into a 128-dimensional crack feature vector , to achieve feature fusion of pavement crack data.

6. The municipal road construction quality tracking, monitoring and management platform according to claim 5 is characterized in that: The construction parameter fusion module (32) is used to fuse multi-source data of asphalt penetration, construction machinery operation trajectory and construction machinery vibration frequency monitored by the construction monitoring unit (2), and the specific steps involved are: Define the mechanical trajectory node feature vector and neighborhood ; Based on the learnable weight matrix The mechanical trajectory node feature vector Mapping to a higher-order space ; Calculate the unnormalized attention score through the spatiotemporal attention mechanism , and the time decay function and space decay function are introduced to quantify the dynamic influence weight of vibration parameters on the compaction quality of specific trajectory points: ； In the formula, Represents vector concatenation; represents the activation function; represents the time decay function; represents the spatial attenuation function; Indicates that the construction machinery has reached the trajectory node and the arrival trajectory node time interval; Indicates that the construction machinery has reached the trajectory node With trajectory nodes The horizontal coordinate distance interval; Represents a trajectory node The eigenvector of The learnable weight matrix The high-order spatial representation obtained by mapping; Generate normalized attention coefficient based on Softmax function ; The attention factor Filling to a sparse matrix , generating a spatiotemporal correlation matrix , and introduce the depth of water accumulation and traffic flow As the modulation parameter, the optimized spatiotemporal correlation matrix is ​​obtained : In the formula, Represents the Sigmoid function; represents the learnable parameter matrix.

7. The municipal road construction quality tracking, monitoring and management platform according to claim 6 is characterized by: The multi-source fusion module (33) dynamically fuses crack tensors based on a cross-attention mechanism with the construction parameter tensor , and introduce the depth of water accumulation and traffic flow , generating a comprehensive risk profile The specific steps involved are: The crack feature vector is expanded into a three-dimensional tensor, matching the time and space dimensions of the construction parameter matrix to obtain the crack tensor ; The optimized spatiotemporal correlation matrix Perform feature encoding to obtain the construction parameter tensor ; Dynamically fusion of crack tensors based on cross-attention mechanism with the construction parameter tensor , we get the attention-weighted construction parameter features: in: ; ; ; In the formula, is the query vector, which represents the query requirement of three-dimensional crack characteristics; is the key vector, representing the index features of the construction parameters; is a value vector, representing the actual impact value of the construction parameters; Represents the sensitivity of three-dimensional crack characteristics to construction parameters; Matchable features representing construction parameters; Indicates the actual impact intensity of construction parameters; represents the key vector dimension; represents the construction parameter characteristics after attention weighting; Indicates the strength of association between three-dimensional crack characteristics and construction parameters; Based on the residual connection, the crack tensor Construction parameter characteristics after attention weighting Add and normalize the layers to generate a comprehensive risk profile : In the formula, Represents the comprehensive risk characteristics; Representation layer normalization; The depth of water and traffic flow and comprehensive risk characteristics Perform feature fusion: In the formula, Representation Node The initial eigenvector of Indicates the water depth encoder; represents a traffic flow encoder; represents the bias term; represents the weight matrix; Represents the activation function.

8. The municipal road construction quality tracking, monitoring and management platform according to claim 7 is characterized by: The risk assessment unit (4) comprises a crack risk prediction module (41), a settlement prediction module (42), a multi-order risk transfer module (43) and a comprehensive road analysis module (44); Among them, the fracture risk prediction module (41) is based on the fracture feature vector and comprehensive risk characteristics Constructing a crack extension risk probability model: In the formula, Represents the Sigmoid function; is the weight matrix; is the bias term; Represents vector concatenation; is the crack extension probability; The settlement prediction module (42) obtains a settlement prediction value based on the LSTM-ARIMA hybrid model, and constructs a settlement risk prediction model through the settlement prediction value and the influence of construction parameters to dynamically correct the settlement risk. The settlement risk prediction model is: In the formula, Indicates time Prediction of settlement; represents the weight coefficient of predicted settlement, The weight coefficient representing the influence of construction parameters; Indicates the amount of sedimentation Temperature The partial derivative of Indicates the amount of sedimentation Vibration frequency The partial derivative of Indicates the subsidence risk value; Representation Node Belongs to Node The space-time neighborhood set of ; The multi-level risk transfer module (43) propagates crack and subsidence risks based on graph neural networks and introduces water depth and traffic flow Constructing a multi-risk transmission model: ； In the formula, Representation Node In the Hidden state of the layer; Indicates the node update weight matrix; represents a learnable parameter; represents the residual mapping matrix; Representation layer normalization operation; Representation Node In the Layer aggregated neighborhood messages; Representation Node In the Hidden state of the layer; Indicates the number of GNN propagation layers; in: In the formula, represents element-wise multiplication; Representation Node In the Hidden state of the layer; represents the environment gating vector; and ; In the formula, represents the environment gating vector Dimension.

9. The municipal road construction quality tracking, monitoring and management platform according to claim 8 is characterized by: The comprehensive road analysis module (44) performs a final road risk prediction assessment based on the multi-modal road analysis model, and the specific steps involved are: The probability of crack extension ,go through Final features after layer propagation and dynamically corrected settlement risk value Standardize the process; The standardized crack propagation probability , Final Features and subsidence risk value Perform feature fusion through the cross attention mechanism: in, Indicates the environmental factor code, ; Represents the comprehensive features after fusion; Based on the fusion comprehensive features , generate a comprehensive risk scoring model: ； The coupling term of crack risk and construction parameters is introduced into the comprehensive risk scoring model: ； In the formula, Represents the fusion layer parameters; represents the scoring mapping parameters; Represents the fusion layer bias; represents the score mapping bias term; represents the coupling strength coefficient; Indicates the sensitivity of settlement to vibration frequency; Indicates the contribution of environmental factors to long-term risk; Indicates the dynamic interaction intensity between crack risk and construction parameters; It represents the comprehensive risk score after the coupling term of crack risk and construction parameters is introduced, with 0 representing no risk and 1 representing extremely high risk.

10. The municipal road construction quality tracking, monitoring and management platform according to claim 1 is characterized by: The road risk warning unit (5) comprises a real-time scoring module (51) and a risk warning module (52); The real-time scoring module (51) generates a comprehensive risk score through a comprehensive risk scoring model based on the real-time monitored pavement basic data, asphalt penetration and mechanical parameters, and the risk warning module (52) uses a segmented threshold algorithm to trigger risk warnings of different levels.

Images