A maintenance management system for highways

Abstract

This invention discloses a maintenance management system for highways, relating to the field of data analysis technology. The system includes: collecting highway road condition data; creating an anomaly detection autoencoder; marking abnormal road condition data; and constructing a highway pavement condition trend map. Based on the highway pavement condition trend map, a pavement degradation prediction model is established to predict the probability distribution of future highway pavement damage. The system then performs a fitting analysis of the fitness of each executable maintenance scheme based on known maintenance scheme types and the highway pavement damage probability distribution map, selecting a set of executable maintenance schemes for future highway pavement damage at a given time point. Using the influencing factors of the executable maintenance schemes as influencing variables, a dynamic programming state transition equation is established to generate the optimal highway maintenance scheme. The advantages of this invention are: reduced maintenance costs, improved emergency response speed, and reduced traffic disruption.

Description

Technical Field

[0001] This invention relates to the field of data analysis technology, specifically to a maintenance management system for highways. Background Technology

[0002] Highway maintenance and management refers to the process of regularly inspecting, maintaining, repairing, and updating highway infrastructure to ensure its safety, smooth flow, and normal use.

[0003] Existing highway maintenance management mainly adopts regular maintenance or generates corresponding maintenance plans based on the aging status of roads through analysis of overall external environmental disturbances. However, the generation schemes of routine maintenance and maintenance tasks based on external environmental impact prediction have a mismatch between road maintenance needs and maintenance, resulting in poor highway maintenance effects, high maintenance costs, and low efficiency. Summary of the Invention

[0004] To address the aforementioned technical problems, a maintenance management system for highways is provided. This technical solution solves the problem that existing highway maintenance management mainly relies on regular maintenance or generates corresponding maintenance plans based on the analysis of the aging state of the road through overall external environmental interference. However, the generation schemes of routine maintenance and maintenance tasks based on the prediction of external environmental impacts have the problem of mismatch between road maintenance needs and maintenance, resulting in poor highway maintenance effects, high maintenance costs, and low efficiency.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0006] A maintenance management system for highways includes:

[0007] The system includes modules for classifying abnormal data, predicting road surface degradation, screening maintenance plans, and dynamically planning maintenance tasks.

[0008] The abnormal data segmentation module collects road condition data of the highway based on the distributed edge sensors of each road segment, creates an abnormal detection autoencoder, marks abnormal road condition data, and constructs a highway pavement condition trend map.

[0009] The road surface degradation prediction module is electrically connected to the abnormal data segmentation module. The road surface degradation prediction module is used to establish a road surface degradation prediction model based on the highway road surface condition trend map and predict the probability distribution map of future highway road surface damage.

[0010] The maintenance scheme screening module is electrically connected to the road surface degradation prediction module. The maintenance scheme screening module is used to perform fitting analysis on the fitness of each executable maintenance scheme based on the known maintenance scheme types and the probability distribution map of highway road surface damage, and to screen out the set of executable maintenance schemes for highway road surface damage at future unit time nodes.

[0011] The maintenance task dynamic planning module is electrically connected to the maintenance scheme selection module. The maintenance task dynamic planning module is used to establish a dynamic programming state transition equation based on the set of executable maintenance schemes for highway pavement damage at future unit time nodes, taking the influencing factors of the executable maintenance scheme as influencing variables, and generating the optimal highway maintenance scheme.

[0012] Preferably, the abnormal data segmentation module specifically includes:

[0013] Standardized road unit, obtain standardized parameters of highway pavement structure, and establish a standardized parameter array for highway pavement;

[0014] The data preprocessing unit performs standardization processing on the standardized parameter array of the highway pavement and the highway road condition data.

[0015] The data vector conversion unit performs vector conversion between the standardized parameter array of highway pavement and highway road state data according to linear mapping, and constructs a time-series vector of highway pavement state data.

[0016] The anomaly labeling encoder unit, based on the Autoencoder, takes the time-series vector of highway road surface state data as input and the abnormal road state data in the highway road surface state data under the labeling unit time as output to obtain the trend data of abnormal highway road surface state.

[0017] The trend visualization unit generates a highway pavement condition trend map based on abnormal road surface condition trend data.

[0018] Preferably, the road surface degradation prediction module specifically includes:

[0019] The road segment division unit is divided and marked according to the abnormal road surface condition trend data of each road segment in the highway road surface condition trend map, so as to obtain the abnormal road surface condition trend characteristic data of each highway segment.

[0020] The road damage prediction unit, based on the ST-GCN model, randomly initializes the graph convolutional layer weights according to a Gaussian distribution and sets the temporal convolutional layer with hyperparameters. It takes the abnormal road surface condition trend data of each highway segment as input, calculates the graph convolutional output for forward and backward propagation and parameter updates, and obtains the road surface degradation prediction model for each segment. The optimal parameters for each road surface degradation prediction model are then determined. According to the central aggregation of the global model, the optimal parameters of the road surface degradation prediction model for each road segment are used to calculate the federated average, update the global model parameters, and end the training by minimizing the model's error function to generate a prediction of the probability of road surface damage for each road segment of the highway in the future.

[0021] The damage visualization unit generates a distribution map of the predicted future road surface damage probability based on the predicted probability of road surface damage for each section of the highway.

[0022] Specifically, the road surface degradation prediction model is as follows:

[0023]

[0024] In the formula, Let i be the probability of road surface damage on i sections of a highway. This is the weight matrix of the fully connected layer. The feature matrix is ​​the result of spatiotemporal feature fusion. For bias terms of fully connected layers;

[0025] Specifically, the aggregation according to the global model central aggregation refers to:

[0026]

[0027] In the formula, These are global model parameters. Let N be the local parameters of the road surface degradation prediction model for the i-th road segment of the expressway, and N be the total number of road segments.

[0028] Preferably, the maintenance scheme selection module specifically includes:

[0029] The maintenance requirement unit, based on the probability distribution map of highway pavement damage, determines the maintenance requirement of highway pavement damage status at a unit time node.

[0030] The maintenance screening unit, based on the fitness function of a hybrid genetic algorithm, uses the comprehensive maintenance impact factor of the executable maintenance scheme corresponding to the maintenance demand of highway pavement damage status at a unit time node as the interference condition. It calculates the fitness index between the maintenance demand of highway pavement damage status at a unit time node and the known maintenance scheme types, and uses time as the association index to establish a set of executable maintenance schemes for highway pavement damage at a unit time. The interference conditions include: execution cost, execution efficiency, and execution risk.

[0031] Specifically, the fitness function of the hybrid genetic algorithm is as follows:

[0032]

[0033] In the formula, This serves as a fitness index between the maintenance requirements for highway pavement damage and the Sth known maintenance scheme type. The cost of the Sth known maintenance scheme type, For the efficiency of the Sth known maintenance scheme type, For the risk of the Sth known maintenance scheme type, , , All are weighting coefficients.

[0034] Preferably, the dynamic planning module for maintenance tasks specifically includes:

[0035] The influencing factor unit is a set of executable maintenance plans for highway pavement damage at future unit time nodes. It marks the influencing factors of each executable maintenance plan on the target road segment as influencing variables. The influencing variables include: maintenance revenue and traffic impact.

[0036] The system correction unit determines the correction system status of each section of the highway based on real-time comprehensive highway data at each unit time node; the correction system status includes: section damage level and traffic flow.

[0037] The optimal maintenance task unit, based on the impact factors of each executable maintenance plan on the target road segment as influencing variables and the corrected system state of each road segment of the highway, establishes a dynamic programming state transition equation to generate the optimal highway maintenance plan for highway pavement damage at a unit event node.

[0038] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0039] This invention proposes a maintenance management scheme for highways. It utilizes distributed edge sensors to collect real-time, multi-dimensional highway status data, employs an autoencoder to segment abnormal data and construct a road surface status trend map, predicts the probability distribution of future road surface damage based on a spatiotemporal graph convolutional network, selects the optimal maintenance scheme using an improved hybrid genetic algorithm, and optimizes maintenance task scheduling through dynamic programming, comprehensively considering cost, efficiency, and traffic impact to generate a globally optimal maintenance scheme. The advantages of this invention are: reduced maintenance costs, improved emergency response speed, and reduced traffic disruption. Attached Figure Description

[0040] Figure 1 This is a framework diagram of a maintenance management system applied to highways;

[0041] Figure 2 Internal structure diagram of the module for abnormal data;

[0042] Figure 3 This is a diagram showing the internal structure of the road surface degradation prediction module.

[0043] Figure 4 Internal structure diagram of the maintenance scheme screening module;

[0044] Figure 5 To maintain the internal structure diagram of the task dynamic planning module. Detailed Implementation

[0045] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0046] Reference Figure 1 As shown, a maintenance management system for highways includes:

[0047] The system includes modules for classifying abnormal data, predicting road surface degradation, screening maintenance plans, and dynamically planning maintenance tasks.

[0048] The abnormal data segmentation module collects road condition data of the highway based on the distributed edge sensors of each road segment, creates an abnormal detection autoencoder, marks abnormal road condition data, and constructs a highway pavement condition trend map.

[0049] The road surface degradation prediction module is electrically connected to the abnormal data segmentation module. The road surface degradation prediction module is used to establish a road surface degradation prediction model based on the highway road surface condition trend map and predict the probability distribution map of future highway road surface damage.

[0050] The maintenance scheme screening module is electrically connected to the road surface degradation prediction module. The maintenance scheme screening module is used to perform fitting analysis on the fitness of each executable maintenance scheme based on the known maintenance scheme types and the probability distribution map of highway road surface damage, and to screen out the set of executable maintenance schemes for highway road surface damage at future unit time nodes.

[0051] The maintenance task dynamic planning module is electrically connected to the maintenance scheme selection module. The maintenance task dynamic planning module is used to establish a dynamic programming state transition equation based on the set of executable maintenance schemes for highway pavement damage at future unit time nodes, taking the influencing factors of the executable maintenance scheme as influencing variables, and generating the optimal highway maintenance scheme.

[0052] This solution utilizes distributed edge sensors to collect real-time, multi-dimensional status data of highways, employs autoencoders to segment abnormal data and construct a road surface status trend map, predicts the probability distribution of future road surface damage based on a spatiotemporal graph convolutional network, combines an improved hybrid genetic algorithm to select the maintenance scheme with optimal fitness, and optimizes maintenance task scheduling through dynamic programming, comprehensively considering cost, efficiency, and traffic impact to generate a globally optimal maintenance scheme. The advantages of this invention are: reduced maintenance costs, improved emergency response speed, and reduced traffic disruption.

[0053] Reference Figure 2 As shown, the abnormal data segmentation module specifically includes:

[0054] Standardized road unit, obtain standardized parameters of highway pavement structure, and establish a standardized parameter array for highway pavement;

[0055] The data preprocessing unit performs standardization processing on the standardized parameter array of the highway pavement and the highway road condition data.

[0056] The data vector conversion unit performs vector conversion between the standardized parameter array of highway pavement and highway road state data according to linear mapping, and constructs a time-series vector of highway pavement state data.

[0057] The anomaly labeling encoder unit, based on the Autoencoder, takes the time-series vector of highway road surface state data as input and the abnormal road state data in the highway road surface state data under the labeling unit time as output to obtain the trend data of abnormal highway road surface state.

[0058] The trend visualization unit generates a highway pavement condition trend map based on abnormal road surface condition trend data.

[0059] Understandably, for the structure of an autoencoder, the number of hidden layers, the number of neurons, and the activation function can be designed as a deep autoencoder with multiple hidden layers, in a pyramid structure with decreasing or increasing numbers of neurons in the hidden layers. The activation function can be ReLU or Sigmoid. However, the partitioning of the training dataset (training set, validation set, test set), the choice of loss function (such as mean squared error, cross-entropy loss, etc.), the choice of optimizer (such as Adam, SGD, etc.), and the choice of training epochs, batch size, and hyperparameters depend on the training results and are determined by the experience of the implementation technicians, and will not be discussed in detail here.

[0060] Reference Figure 3 As shown, the pavement degradation prediction module specifically includes:

[0061] The road segment division unit is divided and marked according to the abnormal road surface condition trend data of each road segment in the highway road surface condition trend map, so as to obtain the abnormal road surface condition trend characteristic data of each highway segment.

[0062] The road damage prediction unit, based on the ST-GCN model, randomly initializes the graph convolutional layer weights according to a Gaussian distribution and sets the temporal convolutional layer with hyperparameters. It takes the abnormal road surface condition trend data of each highway segment as input, calculates the graph convolutional output for forward and backward propagation and parameter updates, and obtains the road surface degradation prediction model for each segment. The optimal parameters for each road surface degradation prediction model are then determined. According to the central aggregation of the global model, the optimal parameters of the road surface degradation prediction model for each road segment are used to calculate the federated average, update the global model parameters, and end the training by minimizing the model's error function to generate a prediction of the probability of road surface damage for each road segment of the highway in the future.

[0063] The damage visualization unit generates a distribution map of the predicted future road surface damage probability based on the predicted probability of road surface damage for each section of the highway.

[0064] Specifically, the road surface degradation prediction model is as follows:

[0065]

[0066] In the formula, Let i be the probability of road surface damage on i sections of a highway. This is the weight matrix of the fully connected layer. The feature matrix is ​​the result of spatiotemporal feature fusion. For bias terms of fully connected layers;

[0067] Specifically, the aggregation according to the global model central aggregation refers to:

[0068]

[0069] In the formula, These are global model parameters. Let N be the local parameters of the road surface degradation prediction model for the i-th road segment of the expressway, and N be the total number of road segments.

[0070] This solution collects real-time highway pavement condition data using distributed edge sensors, utilizes an autoencoder to segment abnormal data and construct a pavement condition trend map; based on a spatiotemporal graph convolutional network (ST-GCN), the weights of the graph convolutional layers are initialized with a Gaussian distribution, and spatiotemporal features are extracted by combining temporal convolutional layers; model parameters are optimized through forward and backward propagation; a federated learning framework is adopted to upload the optimal parameters of the local models of each road segment to a central server for federated averaging, update the global model parameters, minimize the error function, and finally generate a probability distribution map of future highway pavement damage. The beneficial effect is that it can significantly improve the accuracy of pavement degradation prediction.

[0071] Reference Figure 4 As shown, the maintenance scheme selection module specifically includes:

[0072] The maintenance requirement unit, based on the probability distribution map of highway pavement damage, determines the maintenance requirement of highway pavement damage status at a unit time node.

[0073] The maintenance screening unit, based on the fitness function of a hybrid genetic algorithm, uses the comprehensive maintenance impact factor of the executable maintenance scheme corresponding to the maintenance demand of highway pavement damage status at a unit time node as the interference condition. It calculates the fitness index between the maintenance demand of highway pavement damage status at a unit time node and the known maintenance scheme types, and uses time as the association index to establish a set of executable maintenance schemes for highway pavement damage at a unit time. The interference conditions include: execution cost, execution efficiency, and execution risk.

[0074] Specifically, the fitness function of the hybrid genetic algorithm is as follows:

[0075]

[0076] In the formula, This serves as a fitness index between the maintenance requirements for highway pavement damage and the Sth known maintenance scheme type. The cost of the Sth known maintenance scheme type, For the efficiency of the Sth known maintenance scheme type, For the risk of the Sth known maintenance scheme type, , , All are weighting coefficients.

[0077] This solution determines the maintenance requirements for road surface damage at a given time node by using a probability distribution map of highway pavement damage. Combined with a hybrid genetic algorithm (IHGA), and using comprehensive maintenance influencing factors (such as cost, efficiency, and risk) as interference conditions, it calculates the fitness index of each executable maintenance scheme, selects the optimal set of maintenance schemes, and establishes a dynamic maintenance plan using a time index. After implementation, it can significantly improve the adaptability and execution efficiency of maintenance schemes and reduce traffic disruption.

[0078] Reference Figure 5 As shown, the maintenance task dynamic planning module specifically includes:

[0079] The influencing factor unit is a set of executable maintenance plans for highway pavement damage at future unit time nodes. It marks the influencing factors of each executable maintenance plan on the target road segment as influencing variables. The influencing variables include: maintenance revenue and traffic impact.

[0080] The system correction unit determines the correction system status of each section of the highway based on real-time comprehensive highway data at each unit time node; the correction system status includes: section damage level and traffic flow.

[0081] The optimal maintenance task unit, based on the impact factors of each executable maintenance plan on the target road segment as influencing variables and the corrected system state of each road segment of the highway, establishes a dynamic programming state transition equation to generate the optimal highway maintenance plan for highway pavement damage at a unit event node.

[0082] The usage process of a maintenance management system applied to highways is as follows:

[0083] Step 1: Obtain standardized parameters for highway pavement construction and establish a standardized parameter array for highway pavement.

[0084] Step 2: Standardize the standardized parameter array for highway pavement and the highway road condition data;

[0085] Step 3: Perform vector transformation on the standardized parameter array of highway pavement and highway road state data according to linear mapping to construct a time-series vector of highway pavement state data;

[0086] Step 4: Based on the Autoencoder, take the time-series vector of highway road surface condition data as input and take the abnormal road condition data in the highway road surface condition data under the label unit time as output to obtain the trend data of abnormal highway road surface condition.

[0087] Step 5: Generate a highway pavement condition trend map based on the abnormal road surface condition trend data of the highway;

[0088] Step 6: Divide and mark the abnormal road surface condition trend data of each road segment in the highway road surface condition trend map to obtain the abnormal road surface condition trend feature data of each highway segment.

[0089] Step 7: Based on the ST-GCN model, randomly initialize the graph convolutional layer weights according to a Gaussian distribution, set the time convolutional layer with hyperparameters, and use the abnormal road surface state trend feature data of each highway segment as input. Calculate the forward and backward propagation and parameter updates of the graph convolutional output to obtain the road surface degradation prediction model for each segment, and obtain the optimal parameters for the road surface degradation prediction model of each segment. According to the central aggregation of the global model, the optimal parameters of the road surface degradation prediction model for each road segment are used to calculate the federated average, update the global model parameters, and end the training by minimizing the model's error function to generate a prediction of the probability of road surface damage for each road segment of the highway in the future.

[0090] Step 8: Generate a distribution map of the predicted future road surface damage probability for each section of the highway.

[0091] Step 9: Based on the probability distribution map of highway pavement damage, determine the maintenance requirements for highway pavement damage at each unit time node.

[0092] Step 10: Based on the fitness function of the hybrid genetic algorithm, the comprehensive maintenance impact factor of the executable maintenance scheme corresponding to the maintenance demand of highway pavement damage status at a unit time node is used as the interference condition to calculate the fitness index between the maintenance demand of highway pavement damage status at a unit time node and the known maintenance scheme type, and time is used as the association index to establish a set of executable maintenance schemes for highway pavement damage at a unit time.

[0093] Step 11: Based on the set of executable maintenance plans for highway pavement damage at future unit time nodes, mark the impact factors on the target road segment for each executable maintenance plan as impact variables;

[0094] Step 12: Based on real-time comprehensive highway data at each unit time node, determine the status of the correction system for each section of the highway;

[0095] Step 13: Based on the impact factors on the target road segment for each executable maintenance scheme as influencing variables and the corrected system state of each road segment of the highway, establish a dynamic programming state transition equation to generate the optimal highway maintenance scheme for highway pavement damage at a unit event node.

[0096] The foregoing has shown and described 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 to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection claimed by the appended claims and their equivalents is defined.

Claims

1. A maintenance management system for highways, characterized in that, include: The system includes modules for classifying abnormal data, predicting road surface degradation, screening maintenance plans, and dynamically planning maintenance tasks. The abnormal data segmentation module collects road condition data of the highway based on the distributed edge sensors of each road segment, creates an abnormal detection autoencoder, marks abnormal road condition data, and constructs a highway pavement condition trend map. The road surface degradation prediction module is electrically connected to the abnormal data segmentation module. The road surface degradation prediction module is used to establish a road surface degradation prediction model based on the highway road surface condition trend map, predicting the probability distribution of future highway road surface damage. Specifically, it includes: The road segment division unit is divided and marked according to the abnormal road surface condition trend data of each road segment in the highway road surface condition trend map, so as to obtain the abnormal road surface condition trend characteristic data of each highway segment. The road damage prediction unit, based on the ST-GCN model, randomly initializes the graph convolutional layer weights according to a Gaussian distribution, sets the temporal convolutional layer with hyperparameters, and takes the abnormal road surface condition trend feature data of each highway segment as input. It then calculates the graph convolutional output, performs forward and backward propagation and parameter updates to obtain the road surface degradation prediction model for each segment, and finally obtains the optimal parameters for each segment's road surface degradation prediction model. According to the central aggregation of the global model, the optimal parameters of the road surface degradation prediction model for each road segment are used to calculate the federated average, update the global model parameters, and end the training by minimizing the model's error function to generate a prediction of the probability of road surface damage for each road segment of the highway in the future. The damage visualization unit generates a distribution map of the predicted future road surface damage probability based on the predicted probability of road surface damage for each section of the highway. Specifically, the road surface degradation prediction model is as follows: ； In the formula, Let be the probability of road surface damage in the i-th segment of the highway. This is the weight matrix of the fully connected layer. The feature matrix is ​​the result of spatiotemporal feature fusion. For bias terms of fully connected layers; Specifically, the aggregation according to the global model central aggregation refers to: ； In the formula, These are global model parameters. Here are the local parameters of the road surface degradation prediction model for the i-th segment of the expressway, and N is the total number of roads in the segment. The maintenance scheme screening module is electrically connected to the road surface degradation prediction module. The maintenance scheme screening module is used to perform fitting analysis on the fitness of each executable maintenance scheme based on the known maintenance scheme types and the probability distribution map of highway road surface damage, and to screen out the set of executable maintenance schemes for highway road surface damage at future unit time nodes. The maintenance task dynamic planning module is electrically connected to the maintenance scheme selection module. The maintenance task dynamic planning module is used to establish a dynamic programming state transition equation based on the set of executable maintenance schemes for highway pavement damage at future unit time nodes, taking the influencing factors of the executable maintenance schemes as influencing variables, and generating the optimal highway maintenance scheme. Specifically, it includes: The influencing factor unit is a set of executable maintenance plans for highway pavement damage at future unit time nodes, and each executable maintenance plan is labeled with an influencing factor on the target road segment as an influencing variable; the influencing variables include: maintenance revenue and traffic impact; The system correction unit determines the correction system status of each section of the highway based on real-time comprehensive highway data at each unit time node; the correction system status includes: section damage level and traffic flow. The optimal maintenance task unit, based on the impact factors of each executable maintenance plan on the target road segment as influencing variables, and the correction system state of each road segment of the highway, establishes a dynamic programming state transition equation to generate the optimal highway maintenance plan for highway pavement damage per unit time node.

2. The maintenance management system for highways according to claim 1, characterized in that, The abnormal data segmentation module specifically includes: Standardized road unit, obtain standardized parameters of highway pavement structure, and establish a standardized parameter array for highway pavement; The data preprocessing unit performs standardization processing on the standardized parameter array of the highway pavement and the highway road condition data. The data vector conversion unit performs vector conversion between the standardized parameter array of highway pavement and highway road state data according to linear mapping, and constructs a time-series vector of highway pavement state data. The anomaly labeling encoder unit, based on the Autoencoder, takes the time-series vector of highway road surface state data as input and the abnormal road state data in the highway road surface state data under the labeling unit time as output to obtain the trend data of abnormal highway road surface state. The trend visualization unit generates a highway pavement condition trend map based on abnormal road surface condition trend data.

3. The maintenance management system for highways according to claim 2, characterized in that, The maintenance solution selection module specifically includes: The maintenance requirement unit, based on the highway pavement damage probability distribution map, determines the highway pavement damage status maintenance requirement at a given time node. The maintenance screening unit, based on the fitness function of a hybrid genetic algorithm, uses the comprehensive impact factor of the maintenance of executable maintenance schemes corresponding to the maintenance needs of highway pavement damage status at a unit time node as interference conditions. It calculates the fitness index between the maintenance needs of highway pavement damage status at a unit time node and the known maintenance scheme types, and uses time as the association index to establish a set of executable maintenance schemes for highway pavement damage at a unit time. The interference conditions include: execution cost, execution efficiency, and execution risk.

4. The maintenance management system for highways according to claim 3, characterized in that, The fitness function of the hybrid genetic algorithm is specifically as follows: ； In the formula, This serves as a fitness index between the maintenance requirements for highway pavement damage and the Sth known maintenance scheme type. The cost of the Sth known maintenance scheme type, For the efficiency of the Sth known maintenance scheme type, For the risk of the Sth known maintenance scheme type, , , All are weighting coefficients.

Images