Multi-model fusion asphalt mixture fatigue life prediction method and system
Through the multi-model fusion method, combined with gray correlation analysis and genetic algorithm to optimize the BP neural network, the problems of low accuracy and insufficient applicability of asphalt mixture fatigue life prediction in the prior art are solved, and efficient and accurate fatigue life prediction is achieved.
Patent Information
- Application Number
- CN202411563634.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-05
- Publication Date
- 2025-02-28
AI Technical Summary
The prior art is difficult to find a balance between efficiency, accuracy and applicability. The methods used to predict the fatigue life of asphalt mixtures have problems such as long periods, high costs and low prediction accuracy.
The multi-model fusion method is used to screen out the main influencing factors closely related to the fatigue life of asphalt mixture through gray correlation analysis, and these factors are used as input features of the BP neural network model, and the initial weight and threshold of the BP neural network are optimized using genetic algorithms to improve the prediction accuracy and generalization ability of the model.
It significantly improves the accuracy and efficiency of fatigue life prediction of asphalt mixture, adapts to different types of asphalt mixtures and complex construction conditions, and solves the problems of low prediction accuracy and insufficient applicability in the prior art.
Smart Images
Figure CN119541720A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the fields of road engineering and artificial intelligence technology, and in particular to a multi-model fusion asphalt mixture fatigue life prediction method and system. Background Art
[0002] Asphalt mixture is a commonly used road paving material, and its fatigue life directly determines the durability and service life of the road surface. Accurately predicting the fatigue life of asphalt mixture has important engineering application value for road design, construction and maintenance. At present, common fatigue life prediction methods mainly include laboratory test method, empirical formula method and numerical simulation method. However, these methods have significant shortcomings: the laboratory test method has a long cycle and high cost, and it is difficult to simulate the real road environment; although the empirical formula method is simple, it lacks adaptability and is difficult to cope with different materials and construction conditions; although the numerical simulation method can simulate complex conditions, it consumes a lot of computing resources and requires high accuracy of the model, resulting in a limited scope of application. Therefore, it is difficult for existing technologies to find a balance between efficiency, accuracy and applicability. Summary of the invention
[0003] In view of the many problems existing in the above-mentioned prior art, the present invention provides a method and system for predicting fatigue life of asphalt mixture with multi-model fusion. The present invention selects the main influencing factors closely related to fatigue life of asphalt mixture through grey correlation analysis, takes these influencing factors as input features of BP neural network model, and optimizes the initial weights and thresholds of BP neural network through genetic algorithm to improve the prediction accuracy and generalization ability of the model. Through multiple iterations, accurate prediction of fatigue life of asphalt mixture is finally achieved, which significantly improves the efficiency and accuracy of prediction.
[0004] A multi-model fusion asphalt mixture fatigue life prediction method includes the following steps:
[0005] collecting indirect tensile fatigue test data of asphalt mixture, and normalizing the test data of the asphalt mixture to generate a normalized test data set;
[0006] The normalized experimental data set is analyzed based on the grey correlation analysis method, the grey correlation degree between each input parameter and the fatigue life of the asphalt mixture is calculated, the main influencing factors most relevant to the fatigue life of the asphalt mixture are screened out, and the main influencing factors are used as input features of the BP neural network model;
[0007] Constructing a BP neural network model, wherein the BP neural network model includes an input layer, a hidden layer and an output layer, the number of nodes in the input layer is the number of the main influencing factors, and the number of nodes in the output layer is 1, which is used to output the fatigue life prediction value of the asphalt mixture;
[0008] The initial weights and thresholds of the BP neural network model are optimized by using a genetic algorithm, wherein the genetic algorithm includes initializing a population, setting a crossover probability and a mutation probability, and selecting individuals with higher fitness to perform crossover and mutation operations according to a fitness function until the weights and thresholds of the optimized BP neural network model are obtained;
[0009] The optimized BP neural network model is used to predict the fatigue life of the asphalt mixture experimental data, and the fatigue life prediction result of the asphalt mixture is output.
[0010] Preferably, the normalization process performed in the data preprocessing step is performed using the following formula:
[0011]
[0012] Among them, x is the original experimental data; x min and x max are the minimum and maximum values of the experimental data respectively; x * is the normalized data, and its range is [0,1].
[0013] Preferably, in the grey correlation analysis method, the calculation formula of the grey correlation coefficient is:
[0014]
[0015] Among them, ξ i (k) is the grey correlation coefficient of the i-th feature on the k-th index; i (k)=|y(k)-x i (k)|, ρ is the resolution coefficient, ρ∈(0,∞), the smaller ρ is, the greater the resolution is, the value range of ρ is (0,1), when ρ≤0.5463, the resolution is best, usually ρ=0.5.
[0016] Preferably, the threshold range of the grey correlation coefficient in the grey correlation analysis method is set to 0.5 to 1.0.
[0017] Preferably, the number of hidden layer nodes of the BP neural network model is determined by the following formula:
[0018] hiddennum=sqrt(m+n)+a
[0019] Among them, m is the number of input layer nodes; n is the number of output layer nodes; a is an integer between 1 and 10.
[0020] Preferably, the BP neural network model is trained using a back propagation algorithm, which adjusts network weights and thresholds by minimizing the mean square error.
[0021] Preferably, the fitness function of the genetic algorithm is the inverse of the prediction error of the BP neural network model to ensure that the fitness value is proportional to the prediction accuracy.
[0022] Preferably, the population size of the genetic algorithm is set to 30, the crossover probability is set to 0.8, the mutation probability is set to 0.2, and the number of evolutionary iterations is set to 50.
[0023] Preferably, the prediction result is compared with actual experimental data, the prediction error is calculated, and the parameters of the BP neural network model are further optimized and adjusted.
[0024] A system for implementing the multi-model fusion asphalt mixture fatigue life prediction method, the system comprising:
[0025] A data acquisition module is used to collect indirect tensile fatigue test data of asphalt mixture, including temperature, asphalt-stone ratio, stress level, asphalt content, void ratio and load times, and normalize the test data of the asphalt mixture to generate a normalized test data set;
[0026] A feature analysis module is used to analyze the normalized experimental data set based on a grey correlation analysis method, calculate the grey correlation between each input parameter and the fatigue life of the asphalt mixture, and screen out the main influencing factors most relevant to the fatigue life of the asphalt mixture;
[0027] A neural network construction module is used to construct a BP neural network model, wherein the BP neural network model includes an input layer, a hidden layer and an output layer, wherein the number of nodes in the input layer is the number of the main influencing factors, and the number of nodes in the output layer is 1, which is used to output the fatigue life prediction value of the asphalt mixture;
[0028] An optimization module, used for optimizing the initial weights and thresholds of the BP neural network model by using a genetic algorithm, wherein the optimization module includes components for initializing the population, setting the crossover probability and the mutation probability, and selecting individuals with higher fitness to perform crossover and mutation operations according to the fitness function until the weights and thresholds of the optimized BP neural network model are obtained;
[0029] The prediction module is used to use the optimized BP neural network model to perform fatigue life prediction on the experimental data of the asphalt mixture and output the fatigue life prediction result of the asphalt mixture.
[0030] Compared with the prior art, the advantages and beneficial effects of the present invention are:
[0031] The present invention achieves the effect of improving the accuracy, efficiency and adaptability of asphalt mixture fatigue life prediction by means of grey correlation analysis, genetic algorithm and BP neural network technology.
[0032] Compared with the prior art, the present invention screens out key influencing factors related to fatigue life through grey correlation analysis, thereby reducing the redundancy of input features;
[0033] The present invention uses a genetic algorithm to optimize the weights and thresholds of the BP neural network, thereby avoiding the defect that the neural network is prone to fall into a local optimum.
[0034] The present invention significantly improves the prediction accuracy and the generalization ability of the model through the method of multi-model fusion.
[0035] The present invention not only effectively solves the problems of long experimental period, high cost and low prediction accuracy in the prior art, but also can adapt to different types of asphalt mixtures and complex construction conditions. BRIEF DESCRIPTION OF THE DRAWINGS
[0036] Figure 1 is a flow chart of the method of the present invention;
[0037] Figure 2 This is a comparison chart of predicted values and measured values in an embodiment of the present invention;
[0038] Figure 3 4 is a comparison chart of prediction error values in an embodiment of the present invention. DETAILED DESCRIPTION
[0039] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the present disclosure. In the following detailed description, for ease of explanation, many specific details are set forth to provide a comprehensive understanding of the embodiments of the present disclosure. However, it is obvious that one or more embodiments may also be implemented without these specific details. In addition, in the following description, descriptions of known structures and technologies are omitted to avoid unnecessary confusion of the concepts of the present disclosure.
[0040] The terms used herein are only for describing specific embodiments and are not intended to limit the present disclosure. The terms "comprise", "include", etc. used herein indicate the existence of the features, steps, operations and / or components, but do not exclude the existence or addition of one or more other features, steps, operations or components.
[0041] All terms (including technical and scientific terms) used herein have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification, and should not be interpreted in an idealized or overly rigid manner.
[0042] The method of the present invention comprises:
[0043] (1) Establish a data set. The data are collected from the indirect tensile fatigue test data of asphalt mixture and the mixture composition parameters, including temperature, asphalt-stone ratio, stress level, asphalt content, void ratio, load number and other parameters.
[0044] (2) Data preprocessing: normalize the collected experimental data to obtain a normalized data set.
[0045] Furthermore, the normalization calculation formula described in (2) is:
[0046]
[0047] Among them, x is the original sample data value; x max is the maximum value of the original sample data; x min is the minimum value of the original sample data; x * is the data value after normalization, and its range is [0,1].
[0048] (3) Feature selection: The preprocessed data is analyzed based on the grey correlation analysis method, and the grey correlation between each parameter and fatigue life is calculated. Then, the parameters are sorted according to the size of the grey correlation, and the key influencing parameters closely related to fatigue life are screened out as the input features of the neural network model.
[0049] Furthermore, the grey correlation analysis method used in (3) calculates the correlation between each influencing factor and fatigue life. The correlation represents the overall similarity between the feature sequence and the parent sequence, that is, the average value of the correlation coefficient. Its calculation formula is:
[0050]
[0051] Among them i (k)=|y(k)-x i (k)|, ρ is the resolution coefficient, ρ∈(0,∞), the smaller ρ is, the greater the resolution is. Generally, the value interval of ρ is (0,1), and the specific value is determined by the situation. When ρ≤0.5463, the resolution is the best, and ρ=0.5 is usually taken. ξ i (k) is the comparison sequence x i Grey correlation coefficient on the kth index, where r i is the correlation between the ith feature sequence and the parent sequence, and m is the number of observations.
[0052] (4) Divide the original data set into a training set and a test set. The training set is used for training and learning the neural network model, and the test set is used to verify the prediction accuracy of the neural network model.
[0053] (5) Establishment and training of BP neural network. Construct a BP neural network model, including input layer, hidden layer and output layer. The number of nodes in the input layer is equal to the number of main influencing parameters, and the output parameter is 1 (fatigue life, i.e., number of loads).
[0054] Furthermore, the number of hidden layers in (5) is determined using an empirical formula:
[0055] hiddennum=sqrt(m+n)+a
[0056] Where m is the number of neurons in the input layer, n is the number of neurons in the output layer, and a is usually an integer between 1 and 10. Through continuous updating and adjustment, when the mean square error value of the training set reaches the minimum, the optimal number of hidden layer nodes is determined.
[0057] The range of the number of hidden layer neurons j is:
[0058]
[0059] Where a is the dimension of the input layer, b is the dimension of the output layer, and c is a constant between [1,10].
[0060] (6) Genetic algorithm (GA) optimization: construct a genetic algorithm model and set parameters such as population size, crossover probability, and mutation probability. Define the fitness function and use the prediction error of the BP neural network as the fitness function to evaluate the fitness of each individual in the population. Iteratively update the population and generate a new population through selection, crossover, and mutation operations until the preset number of iterations or fitness threshold is reached to obtain the optimal BP neural network weights and thresholds. Initialize the BP neural network using the optimal weights and thresholds obtained by genetic algorithm optimization. Use the training data set to train the BP neural network and adjust the network weights and thresholds through the back propagation algorithm to minimize the prediction error.
[0061] Furthermore, the genetic algorithm optimization process in (6) is as follows:
[0062] Step 1. Initialize the population: randomly generate a certain number of individuals (i.e., neural network weights and biases) to form an initial population, and encode the individuals into a real number string. The encoding length is calculated according to the following formula:
[0063] D=a×b+b+b×c+c
[0064] In the formula, a is the number of input layer nodes set in the BP neural network, b is the number of hidden layer nodes, c is the number of output layer nodes, and D is the encoding length.
[0065] (7) Fatigue life prediction: Input the parameters of the asphalt mixture to be predicted into the trained BP neural network model, and output the corresponding fatigue life prediction value. Verify and evaluate the prediction results to verify the accuracy and reliability of the model.
[0066] Step 2. Fitness evaluation: For each individual, the BP neural network model is used for training, and the fitness value is calculated based on its performance on the training set.
[0067] Step 3. Selection operation: select parent individuals according to fitness values. The roulette method is used for selection operation, so that individuals with higher fitness have a higher probability of being selected, and the probability is:
[0068]
[0069] Among them, f i is the inverse of the fitness value, and N is the population size;
[0070] Step 4. Crossover operation: Randomly select two from the selected parent individuals for crossover operation to generate new offspring individuals. The crossover operation selects two crossover points to merge and exchange the genetic information of the parent individuals. The crossover operation is:
[0071] C nl =C nl (1-b)+C ml b
[0072] C ml =C ml (1-b)+C nl b
[0073] Where b is a random number in the range [0,1], C nl Refers to the gene at position l on the nth chromosome, C ml Refers to the gene at position l on the mth chromosome.
[0074] Step 5. Mutation: Randomly perform genetic mutation operations on offspring individuals to increase the diversity of the population. The mutation operation can be a small random adjustment of the weights or biases, or a larger random change.
[0075] Step 6. Update the population: Replace the original parent individuals with the newly generated offspring individuals to form a new population.
[0076] Step 7. Repeat steps 2 to 6 until the specified number of iterations is reached or the termination condition is met.
[0077] Furthermore, the genetic algorithm parameters in (6) are set as follows: the number of evolutionary iterations is set to 50, the population size is set to 30, the crossover function parameter is set to 0.8, and the mutation function parameter is set to 0.2.
[0078] This example uses the indirect tensile fatigue test data of asphalt mixture as the data set. The flow chart of the present invention is shown in the attached Figure 1 The prediction process of the method of the present invention mainly includes the following steps:
[0079] Step 1. Collection and preparation of experimental data
[0080] First, collect the experimental data of asphalt mixture, which usually includes the following aspects:
[0081] Input variables: Physical properties of asphalt mixture (such as void ratio, asphalt content, aggregate size distribution, etc.), experimental conditions such as temperature and loading frequency, and fatigue test conditions such as stress level and loading mode. Output variables: Fatigue life of asphalt mixture (expressed in the number of loads)
[0082] Step 2. Grey Relational Analysis (GRA)
[0083] (1): Construct comparison sequence and reference sequence
[0084] The fatigue life of asphalt mixture (number of loads) is set as the reference sequence, and other input variables are set as comparison sequences.
[0085] (2): Dimensionless data
[0086] All data were normalized to eliminate the effect of dimension.
[0087] (3): Calculate the correlation coefficient
[0088] The correlation coefficient between each input variable and the output variable is calculated to measure the influence of different input variables on fatigue life.
[0089] (4): Calculate the correlation
[0090] The correlation coefficient is used to calculate the correlation between each input variable and fatigue life, and the main influencing factors are determined. The main influencing factors here are temperature, stress ratio, asphalt-stone ratio, asphalt content, stability, and loading frequency.
[0091] Step 3. Genetic Algorithm (GA) Optimize BP Neural Network
[0092] (1) Constructing BP neural network model: using the main influencing factors screened by GRA as the input layer nodes of BP network and fatigue life as the output layer nodes.
[0093] (2) Encoding: Encode the weights and threshold parameters of the BP neural network to form the initial population of the genetic algorithm.
[0094] (3) Genetic operations: selection, crossover, and mutation operations optimize the parameters of the BP network. Fitness function: Taking the prediction error of the BP network as the target, define the fitness function and calculate the fitness of each individual.
[0095] (4) Iterative evolution: The genetic algorithm runs iteratively until a preset stopping condition (such as maximum number of generations or prediction accuracy) is reached.
[0096] Step 4. Optimize model training
[0097] (1) Initialization: Initialize the BP neural network using the weights and thresholds optimized by GA.
[0098] (2) Forward propagation: Input the training set data into the BP neural network and calculate the output value.
[0099] (3) Back propagation: Calculate the prediction error and use the gradient descent method to adjust the network weights and thresholds.
[0100] (4) Training iteration: Continuously adjust network parameters until the network error converges or reaches the preset accuracy requirements.
[0101] Step 5. Prediction Validation
[0102] (1) Prediction: Input the new experimental data into the trained BP neural network to obtain the fatigue life prediction value of the asphalt mixture.
[0103] (2) Verification: Compare the predicted values with the measured values to evaluate the prediction accuracy of the model.
[0104] Step 6. Result analysis
[0105] (1) Results (see Table 1): Comparison of predicted values and measured values Figure 2 , the model prediction error comparison is shown in Figure 3 .
[0106] Table 1. Results of some predicted values and error values
[0107]
[0108]
[0109] (2) Error analysis: The model error evaluation indicators are shown in Table 2.
[0110] Table 2 is the error evaluation index table
[0111]
[0112] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware.
[0113] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.
Claims
1. A multi-model fusion asphalt mixture fatigue life prediction method, characterized in that: The following steps are involved: collecting indirect tensile fatigue test data of asphalt mixture, and normalizing the test data of the asphalt mixture to generate a normalized test data set; The normalized experimental data set is analyzed based on the grey correlation analysis method, the grey correlation degree between each input parameter and the fatigue life of the asphalt mixture is calculated, the main influencing factors most relevant to the fatigue life of the asphalt mixture are screened out, and the main influencing factors are used as input features of the BP neural network model; Constructing a BP neural network model, wherein the BP neural network model includes an input layer, a hidden layer and an output layer, the number of nodes in the input layer is the number of the main influencing factors, and the number of nodes in the output layer is 1, which is used to output the fatigue life prediction value of the asphalt mixture; The initial weights and thresholds of the BP neural network model are optimized by using a genetic algorithm, wherein the genetic algorithm includes initializing a population, setting a crossover probability and a mutation probability, and selecting individuals with higher fitness to perform crossover and mutation operations according to a fitness function until the weights and thresholds of the optimized BP neural network model are obtained; The optimized BP neural network model is used to predict the fatigue life of the asphalt mixture experimental data, and the fatigue life prediction result of the asphalt mixture is output.
2. The method for predicting fatigue life of asphalt mixture by multi-model fusion according to claim 1 is characterized in that: The normalization process performed in the data preprocessing step is performed using the following formula: Among them, x is the original experimental data; x min and x max are the minimum and maximum values of the experimental data respectively; x * is the normalized data, and its range is [0,1].
3. The method for predicting fatigue life of asphalt mixture by multi-model fusion according to claim 1 is characterized in that: In the grey correlation analysis method, the calculation formula of the grey correlation coefficient is: Among them, ξ i (k) is the grey correlation coefficient of the i-th feature on the k-th index; i (k)=|y(k)-x i (k)|, ρ is the resolution coefficient, ρ∈(0,∞), the smaller ρ is, the greater the resolution is, the value range of ρ is (0,1), when ρ≤0.5463, the resolution is best, take ρ=0.
5.
4. The method for predicting fatigue life of asphalt mixture by multi-model fusion according to claim 3 is characterized in that: The threshold range of the grey correlation coefficient in the grey correlation analysis method is set to 0.5 to 1.
0.
5. The method for predicting fatigue life of asphalt mixture by multi-model fusion according to claim 1 is characterized in that: The number of hidden layer nodes of the BP neural network model is determined by the following formula: hiddennum=sqrt(m+n)+a Among them, m is the number of input layer nodes; n is the number of output layer nodes; a is an integer between 1 and 10.
6. The method for predicting fatigue life of asphalt mixture by multi-model fusion according to claim 5 is characterized in that: The BP neural network model is trained using a back propagation algorithm, which adjusts network weights and thresholds by minimizing mean square error.
7. The method for predicting fatigue life of asphalt mixture by multi-model fusion according to claim 1 is characterized in that: The fitness function of the genetic algorithm is the inverse of the prediction error of the BP neural network model to ensure that the fitness value is proportional to the prediction accuracy.
8. The method for predicting fatigue life of asphalt mixture by multi-model fusion according to claim 7 is characterized in that: The population size of the genetic algorithm is set to 30, the crossover probability is set to 0.8, the mutation probability is set to 0.2, and the number of evolutionary iterations is set to 50.
9. The method for predicting fatigue life of asphalt mixture by multi-model fusion according to claim 1, characterized in that: The prediction results are compared with actual experimental data, the prediction errors are calculated, and the parameters of the BP neural network model are further optimized and adjusted.
10. A system for implementing the asphalt mixture fatigue life prediction method of multiple model fusion according to any one of claims 1 to 9, characterized in that: The system comprises: A data acquisition module is used to collect indirect tensile fatigue test data of asphalt mixture, including temperature, asphalt-stone ratio, stress level, asphalt content, void ratio and load times, and normalize the test data of the asphalt mixture to generate a normalized test data set; A feature analysis module is used to analyze the normalized experimental data set based on a grey correlation analysis method, calculate the grey correlation between each input parameter and the fatigue life of the asphalt mixture, and screen out the main influencing factors most relevant to the fatigue life of the asphalt mixture; A neural network construction module is used to construct a BP neural network model, wherein the BP neural network model includes an input layer, a hidden layer and an output layer, wherein the number of nodes in the input layer is the number of the main influencing factors, and the number of nodes in the output layer is 1, which is used to output the fatigue life prediction value of the asphalt mixture; An optimization module, used for optimizing the initial weights and thresholds of the BP neural network model by using a genetic algorithm, wherein the optimization module includes components for initializing the population, setting the crossover probability and the mutation probability, and selecting individuals with higher fitness to perform crossover and mutation operations according to the fitness function until the weights and thresholds of the optimized BP neural network model are obtained; The prediction module is used to use the optimized BP neural network model to perform fatigue life prediction on the experimental data of the asphalt mixture and output the fatigue life prediction result of the asphalt mixture.