Road rut depth prediction method, device and system and storage medium

Abstract

The invention discloses a road rut depth prediction method, device and system and a storage medium. The method comprises the following steps: step S10, acquiring historical road rut depth data; step S20, carrying out normalization processing on the road rut depth data; step S30, training a neural network model based on genetic algorithm optimization according to the normalized road rut depth data; and S40, inputting to-be-processed road rut depth data into the trained neural network model to realize rut depth prediction. By adopting the technical scheme of the invention, the problems that the neural network is low in training efficiency and easy to fall into local optimum in the training and prediction process, the error between the prediction result and the actual result is large, and the pavement rutting depth under the influence of temperature and axle load cannot be accurately determined can be solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of road engineering, and in particular relates to a method and device, a system and a storage medium for predicting road rutting depth. Background Art

[0002] With the rapid expansion of transportation networks and the continuous increase in vehicle loads, the maintenance of highways and urban roads has become an important issue in the operation of transportation systems. Pavement rutting is a groove-like depression that appears on the road surface under the action of long-term vehicle loads, which seriously affects the performance and driving safety of the road. Therefore, accurate prediction of road rutting depth is of great significance for pavement design, maintenance and pavement material selection. However, the formation of rutting depth is a complex process, which is affected by many factors, such as vehicle axle weight, pavement material, temperature and road age, which makes it extremely difficult to predict using traditional physical models or empirical formulas.

[0003] In recent years, artificial intelligence technology, especially artificial neural networks, has been widely used in many fields, including solving road engineering problems, due to its excellent nonlinear mapping ability and self-learning ability. As a typical feedforward neural network, BP neural network continuously adjusts the weights and thresholds in the network through gradient descent method and error back propagation algorithm until the error between the network output and the expected output meets certain accuracy requirements. However, neural networks have problems of low training efficiency and easy to fall into local optimality during training and prediction, which may lead to large deviations between the prediction results and the actual situation, especially when dealing with highly complex and nonlinear road rutting depth prediction problems. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide a road rutting depth prediction method and device, system, and storage medium, which can solve the problems of low training efficiency and easy falling into local optimality in the training and prediction process of neural networks, thereby causing large errors between the predicted results and the actual results, and thus unable to accurately determine the road rutting depth under the influence of temperature and axle load.

[0005] To achieve the above object, the present invention adopts the following technical solution:

[0006] A method for predicting road rutting depth comprises the following steps:

[0007] Step S10, obtaining historical road rutting depth data;

[0008] Step S20, normalizing the road rutting depth data;

[0009] Step S30, training a neural network model optimized by a genetic algorithm according to the normalized road rutting depth data;

[0010] Step S40: input the road rutting depth data to be processed into the trained neural network model to predict the rutting depth.

[0011] Preferably, the historical road rutting depth data includes: temperature, vehicle axle load L, and actually measured historical rutting depth data.

[0012] Preferably, step S30 includes:

[0013] A neural network model is established based on the normalized road rutting depth data;

[0014] Neural network model optimized based on genetic algorithm.

[0015] The present invention also provides a road rutting depth prediction device, comprising:

[0016] An acquisition module is used to obtain historical road rutting depth data;

[0017] A normalization module is used to normalize the road rutting depth data;

[0018] A training module, used for training a neural network model optimized by a genetic algorithm according to normalized road rutting depth data;

[0019] The prediction module is used to input the road rutting depth data to be processed into the trained neural network model to predict the rutting depth.

[0020] Preferably, the historical road rutting depth data includes: temperature, vehicle axle load L, and actually measured historical rutting depth data.

[0021] Preferably, the training module includes:

[0022] A construction unit, used for establishing a neural network model according to the normalized road rutting depth data;

[0023] Optimization unit, used for neural network models optimized based on genetic algorithms.

[0024] The present invention also provides a road rutting depth prediction system, comprising: a memory and a processor, wherein the memory stores a computer program executed by the processor, and the computer program executes a road rutting depth prediction method when executed by the processor.

[0025] The present invention also provides a storage medium, on which a computer program is stored, and the computer program executes the road rutting depth prediction method when running.

[0026] The present invention is used to accurately predict the road rutting depth. Through data preprocessing techniques, such as data cleaning and normalization, the data of ambient temperature, vehicle axle load and rutting depth are processed. Using these processed data, a neural network model is constructed, in which the temperature and axle load data are used as inputs and the rutting depth is used as output. The focus is on using genetic algorithms to optimize the neural network structure, including determining the optimal hidden layer size to achieve the minimum mean square error. This optimization process ensures the high accuracy of the model in predicting the rutting depth. Subsequently, the model is used to make predictions on an independent test set, and the reliability of the prediction results is verified by various statistical and visualization techniques, such as three-dimensional scatter plots and error analysis plots. The present invention provides an effective and scientific prediction tool for road maintenance and management, which can help take timely maintenance measures and improve road safety and life. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0028] Figure 1 The present invention is a flowchart of a method for predicting road rutting depth according to an embodiment of the present invention. DETAILED DESCRIPTION

[0029] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0030] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0031] Embodiment 1:

[0032] like Figure 1 As shown, an embodiment of the present invention provides a method for predicting road rutting depth, comprising the following steps:

[0033] Step S10, obtaining historical road rutting depth data;

[0034] Step S20, normalizing the road rutting depth data;

[0035] Step S30, training a neural network model optimized by a genetic algorithm according to the normalized road rutting depth data;

[0036] Step S40: input the road rutting depth data to be processed into the trained neural network model to predict the rutting depth.

[0037] As an implementation of the embodiment of the present invention, the historical road rutting depth data includes: temperature, vehicle axle load L, and actually measured historical rutting depth data.

[0038] As an implementation of the embodiment of the present invention, step S30 includes:

[0039] A neural network model is established based on the normalized road rutting depth data;

[0040] Neural network model optimized based on genetic algorithm.

[0041] The neural network optimized by genetic algorithm for predicting road rutting depth in the present invention integrates the global search advantage of genetic algorithm and the powerful prediction ability of neural network, effectively improving the performance of road rutting depth prediction in terms of accuracy, convergence speed and generalization ability. This method finds the optimal parameter configuration for the neural network through carefully designed genetic operations, thereby realizing the rapid and accurate prediction of road rutting depth, and providing reliable data support for road maintenance and safety management. It significantly reduces the pavement maintenance cost caused by rutting problems by optimizing the network structure and parameters, while improving the safety of road use. In addition, this method also provides a powerful decision support tool for traffic planning and management, making road maintenance strategies more forward-looking and targeted.

[0042] Genetic algorithm is a search and optimization technology that simulates the biological evolution process in nature. Its core mechanism includes operations such as gene selection, crossover and mutation. It can effectively find the global optimal or near-global optimal solution in a complex search space, and is particularly suitable for dealing with optimization problems with many parameters, strong nonlinearity and large solution space. These characteristics of genetic algorithms make it an ideal choice for optimizing complex neural network parameters and structures, especially in application scenarios that require high precision and high reliability, such as road rutting depth prediction. By integrating genetic algorithms, the training process of neural networks has not only become more efficient, but also the prediction results have become more accurate and reliable, showing the great application potential of genetic algorithms in practical engineering problems.

[0043] Embodiment 2:

[0044] The embodiment of the present invention provides a method for predicting road rutting depth, comprising the following steps:

[0045] S1: Obtain historical road rutting depth data, including temperature T, vehicle axle load L and other influencing factor data as well as actual measured historical rutting depth data.

[0046] S2: Normalize the historical road rutting depth data to form normalized data sets X, Y, where X represents the factors affecting the road rutting depth and Y represents the final measured rutting depth, and divide them into training sample sets D according to a ratio of 7:3 train And the test sample set D test .

[0047] S3: Establish a neural network model based on the normalized road rutting depth data and determine the input layer N input , hidden layer N hidden and the output layer N output The number of neurons, activation function and loss function.

[0048] S4: Based on the established neural network model, initialize the relevant parameters of the genetic algorithm: population size P size , chromosome length C length , crossover probability P corss , mutation probability P mut and the number of iterations G number wait.

[0049] S5: Based on the genetic algorithm parameters initialized in S4, the iterative process of the genetic algorithm begins. In each generation, for each individual (chromosome) C in the population i Calculate its fitness value f(C i ), fitness value f(C i ) reflects the performance of the individual, that is, the prediction error of the neural network. A new generation of population is generated through genetic operations such as selection, crossover and mutation, and then the fitness of each new individual is evaluated.

[0050] S6: Determine the best individual C in the current generation best Whether the preset convergence condition is met (i.e. the prediction error reaches the preset accuracy requirement). If so, terminate the iteration of the genetic algorithm and output the current optimal individual C best As the optimal parameter configuration of the neural network; if it is not satisfied, continue to execute the genetic algorithm iteration until the optimal individual C that meets the preset convergence condition is found best or the maximum number of iterations is reached.

[0051] S7: The optimal individual C obtained in S6 best (Parameter configuration of neural network) is used as the initial parameters of the neural network to train the network.

[0052] S8: Use the trained neural network to predict the rutting depth on the test set and calculate the mean square error (MSE) of the prediction results. test , in order to evaluate the prediction performance of the model.

[0053] S9: If the mean square error MSE test If the convergence accuracy is met, the predicted road rutting depth result is output; if not, the neural network parameters are adjusted and retrained until the prediction error reaches a satisfactory accuracy or the preset maximum number of training times is reached.

[0054] As an implementation method of the embodiment of the present invention, in step S2, the collected rutting depth data is normalized as follows:

[0055]

[0056] Among them, X is the normalized value; x is any value in the data set; x max 、x min are the maximum and minimum values ​​in the data set, respectively.

[0057] As an implementation of the embodiment of the present invention, step S3 includes:

[0058] Step S3.1: Determine the number of neurons in the hidden layer by the number of neurons in the input layer and the output layer, as shown in formula (2):

[0059]

[0060] Among them, L, m, n are the number of neurons in the input layer, hidden layer, and output layer respectively; α is the adjustment parameter α∈(1, 2, 3, ..., 10);

[0061] Step S3.2: The activation function is the nonlinear Sigmoid function δ, and the loss function is the mean square error function MSE. The activation function expression is as follows: Formula (3), and the mean square error function expression is as follows: Formula (4):

[0062]

[0063] Where n is the number of neurons; is the target rutting depth; i To predict the output rutting depth, e is a natural constant, x is the input (x represents a constant in the input layer activation, and x represents the output of the input layer in the hidden layer activation), and E is the mean square error.

[0064] As an implementation of the embodiment of the present invention, step S5 includes:

[0065] S5.1: Calculate the fitness value of each individual in the population, where the fitness value calculation formula (5) is as follows:

[0066]

[0067] S5.2: Perform selection operation according to the fitness value, where the fitness value selection formula (6) is as follows:

[0068]

[0069] S5.3: The crossover operation generates new offspring individuals, where the crossover operation formula (7) (8) is as follows:

[0070]

[0071] S5.4: The mutation operation mutates the newly generated offspring individuals, where the mutation formula (9) (10) is as follows:

[0072] To encode binary:

[0073]

[0074] Encoding real numbers:

[0075] O[i]=O[i]+δ (10)

[0076] S5.5: Evaluate the fitness of the new generation population and select the best individual as the best individual of the current generation. The formula (11) for selecting the best individual is as follows:

[0077]

[0078] Where: f(C i ) represents the fitness value of the i-th individual (chromosome); n represents the number of samples; y j represents the actual measurement value of the jth sample; represents the value of the jth sample predicted by the neural network model; P i Represents individual C i The probability of being selected as a parent; P size represents the size of the population, that is, the number of individuals in the population; P 1 , P 2 Two parent individuals involved in the crossover (pairing) operation; O 1 , O 2 represents two offspring individuals generated by the crossover operation; k represents the randomly selected crossover point position; j represents an index used to traverse all elements or positions of an operation; δ represents the random perturbation value added to a gene or parameter in the mutation operation; represents the best individual in the gth generation.

[0079] In step S5, when the optimal individual in a generation meets the preset convergence condition, that is, when the prediction error reaches the preset accuracy requirement, the iterative process of the genetic algorithm is terminated, and the optimal individual at this time is used as the optimal parameter configuration of the neural network. The preset convergence condition is usually measured by the mean square error (MSE), and the mean square error formula (12) is as follows:

[0080]

[0081] Where n represents the number of samples; y j represents the actual measurement value of the jth sample; Represents the value of the jth sample predicted by the neural network model.

[0082] As an implementation of the embodiment of the present invention, step S9 includes:

[0083] S9.1 The learning rate adjustment formula (13) is as follows:

[0084] Learning Rate new =Learning Rate initial ×decay epoch (13)

[0085] Among them, decay is the decay rate and epoch is the number of iterations.

[0086] Embodiment 3:

[0087] The embodiment of the present invention further provides a road rutting depth prediction device, comprising:

[0088] An acquisition module is used to obtain historical road rutting depth data;

[0089] A normalization module is used to normalize the road rutting depth data;

[0090] A training module, used for training a neural network model optimized by a genetic algorithm according to normalized road rutting depth data;

[0091] The prediction module is used to input the road rutting depth data to be processed into the trained neural network model to predict the rutting depth.

[0092] As an implementation of the embodiment of the present invention, the historical road rutting depth data includes: temperature, vehicle axle load L, and actually measured historical rutting depth data.

[0093] As an implementation of an embodiment of the present invention, the training module includes:

[0094] A construction unit, used for establishing a neural network model according to the normalized road rutting depth data;

[0095] Optimization unit, used for neural network models optimized based on genetic algorithms.

[0096] Embodiment 4:

[0097] An embodiment of the present invention further provides a road rutting depth prediction system, comprising: a memory and a processor, wherein the memory stores a computer program executed by the processor, and the computer program executes a road rutting depth prediction method when executed by the processor.

[0098] Embodiment 5:

[0099] An embodiment of the present invention further provides a storage medium, on which a computer program is stored, and the computer program executes the road rutting depth prediction method when running.

[0100] The embodiments described above are only descriptions of the preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.

Claims

1. A method for predicting road rutting depth, characterized in that: The steps include: Step S10, obtaining historical road rutting depth data; Step S20, normalizing the road rutting depth data; Step S30, training a neural network model optimized by a genetic algorithm according to the normalized road rutting depth data; Step S40: input the road rutting depth data to be processed into the trained neural network model to predict the rutting depth.

2. The road rutting depth prediction method according to claim 1, characterized in that: The historical road rutting depth data includes: temperature, vehicle axle load L, and actually measured historical rutting depth data.

3. The road rutting depth prediction method according to claim 2, characterized in that: Step S30 includes: A neural network model is established based on the normalized road rutting depth data; Neural network model optimized based on genetic algorithm.

4. A road rutting depth prediction device, characterized in that: include: An acquisition module is used to obtain historical road rutting depth data; A normalization module is used to normalize the road rutting depth data; A training module, used for training a neural network model optimized by a genetic algorithm according to normalized road rutting depth data; The prediction module is used to input the road rutting depth data to be processed into the trained neural network model to predict the rutting depth.

5. The road rutting depth prediction device according to claim 4, characterized in that: The historical road rutting depth data includes: temperature, vehicle axle load L, and actually measured historical rutting depth data.

6. The road rutting depth prediction device according to claim 5, characterized in that: The training modules include: A construction unit, used for establishing a neural network model according to the normalized road rutting depth data; Optimization unit, used for neural network models optimized based on genetic algorithms.

7. A road rutting depth prediction system, characterized in that: include: A memory and a processor, wherein the memory stores a computer program executed by the processor, and when the computer program is executed by the processor, the road rutting depth prediction method according to any one of claims 1 to 3 is executed.

8. A storage medium, characterized in that: The storage medium stores a computer program, which executes the road rutting depth prediction method according to any one of claims 1 to 3 when running.

Images