WO2024001719A1 - Parameter optimization method and system for vibration compaction of high-speed rail filler - Google Patents

Abstract

The present invention relates to a parameter optimization method for vibration compaction of a high-speed rail filler, comprising the following steps: (1) fitting dry density in a compaction process by using a hyperbolic model, and calculating and constructing original data; (2) on the basis of obtaining training data, establishing a dry density increment prediction model by using a BP neural network; (3) evaluating the compaction quality by using a compaction degree index, and establishing a compaction degree constraint condition; (4) solving a vibration parameter optimization process on the basis of a GA algorithm, and establishing a GA-based dynamic optimization model; and (5) determining the dynamic optimization result after the parameter optimization processing as an optimal scheme for vibration compaction. The present invention further relates to a parameter optimization system for vibration compaction of a high-speed rail filler using the method. According to the method of the present invention, the problems that in a conventional vibration compaction process, compaction vibration parameters cannot be adjusted in real time, excitation energy cannot be well controlled, vibration compaction equipment is prone to "jumping", particle crushing is increased, and roadbed uneven settlement diseases are formed are solved.

Description

A parameter optimization method and system for vibration compaction of high-speed rail filler Technical field

The invention relates to the technical field of civil engineering, and in particular to a parameter optimization method and system for vibration compaction of high-speed railway fillers.

Background technique

Subgrade performance is closely related to filler composition and vibration compaction process control. In order to ensure the safety of high-speed railway subgrade and improve the comfort of train operation, there are strict standards for the gradation, maximum particle size and shape of subgrade filler. Traditional vibration compaction quality inspection of high-speed railway foundation generally adopts the test pit sampling method, which ignores the problem of filler discreteness, resulting in low compaction efficiency, uneven compaction and other problems.

The compaction state of the filler will continue to change with the compaction process. Constantly adjusting the vibration parameters according to the compaction state of the filler instead of using fixed compaction parameters can improve the compaction efficiency. To achieve this goal, intelligent compaction (IC), which consists of continuous compaction control (CCC) and intelligent compaction (IC), has gradually become a hot spot in recent research. The current research results on vibration parameters (amplitude, frequency, time) are still about dynamic response issues under different vibration parameters. They mainly reveal the impact of a single vibration parameter on the compaction effect, and there is a lack of improvement results in vibration parameters during the compaction process. It cannot meet the requirements of real-time adjustment of intelligent compaction vibration parameters. With the emergence of stepless amplitude-modulated and frequency-modulated road rollers, the roller can be continuously frequency-modulated and amplitude-modulated during the compaction process. However, adjusting the vibration parameters to achieve the maximum compaction degree next time is a partial improvement, while the vibration parameters during the compaction process need to be adjusted Improved in real time to make the filler reach the optimal dense state better and faster. At the same time, excessive excitation energy can easily cause "vibration jump" in the vibration compaction equipment and reduce the service life of the vibration compaction equipment. At the same time, it will also cause an increase in particle breakage and cause uneven settlement of the roadbed.

Contents of the invention

The main purpose of the present invention is to provide a parameter optimization method and system for vibration compaction of high-speed railway bed fill to solve the problem in the prior art that the compaction vibration parameters cannot be adjusted in real time and the excitation energy cannot be very good during the traditional vibration compaction process. Control and vibration compaction equipment are prone to "jumping vibration" phenomena, causing increased particle breakage and uneven settlement of the roadbed.

The present invention is a parameter optimization method for vibration compaction of high-speed rail fillers, which includes the following steps:

(1) Use the hyperbolic model to fit the dry density during the compaction process, and calculate and construct the original data;

(2) On the basis of obtaining training data, use BP neural network to establish a dry density incremental prediction model;

(3) Use compaction degree indicators to evaluate compaction quality and establish compaction degree constraints;

(4) Solve the vibration parameter optimization process based on the GA algorithm and establish a dynamic optimization model based on GA;

(5) Determine the dynamic optimization results described after parameter optimization processing as the best solution for vibration compaction.

Further, the formula of the hyperbolic model in step (1) is:

In the formula, ρ _d is the current state density; ρ _max is the compaction steady state dry density value under the current working conditions; ρ ₀ is the initial state dry density value; n is the number of vibrations; a and b are model parameters.

Further, the steps of establishing the dry density increment prediction model in step (2) include:

a. Construct a high-speed rail fill vibration compaction parameter data set, preprocess the data, and divide the training set, verification set and test set;

b. Build the BP neural network model architecture, input the training set and verification set to perform model training, and determine whether the verification set error is less than the set error. If it is greater, modify the model architecture and retrain. If it is less, proceed to the next step;

c. Use the test set to verify the prediction ability of the BP neural network model, and determine whether the test set error is less than the set error. If it is greater, return to step b to retrain the model. If it is less, save the final dry density increment prediction model;

Further, the dry density incremental prediction model constructs a nonlinear functional relationship between the model input and output. The function is expressed as follows:

In the formula, f is the mapping relationship, A _0i and _fi are the amplitude and frequency at the current moment respectively, p _di is the dry density at the previous moment, is the dry density increment at the current moment.

Further, the data is preprocessed using the data normalization method, and the formula is as follows:

x _scaled = x _std *(max-min)+min

In the formula, x is the data to be normalized, x _min(axis=0) is a row vector composed of the minimum value in each column, x _max(axis=0) is a row vector composed of the maximum value in each column, max is the maximum value of the interval to be mapped to, the default is 1, min is the minimum value of the interval to be mapped, the default is 0, x _std is the standardized result, and x _scaled is the normalized result.

Further, the BP neural network is an optimized BP neural network, and the optimized BP neural network is a traditional BP neural network optimized by introducing a learning rate improver AdamOptimizer optimization algorithm;

The AdamOptimizer improved algorithm is an improved algorithm that uses the first-order moment estimate and the second-order moment estimate of the gradient to dynamically adjust the learning rate of each parameter in a conventional optimization algorithm. The improved formula of the AdamOptimizer improved algorithm is:

In the formula, l _r0 is the initial learning rate, lr _t is the learning rate at time t, is the exponential decay rate estimated by the first moment, is the exponential decay rate estimated for the second moment.

Further, in the step (3): the formula of the compaction constraint is as follows:

In the formula, is the dry density value after vibration n times, is the maximum value of dry density.

Further, the GA-based dynamic optimization model in step (4) is obtained by the following steps:

Set the initial parameters of the GA algorithm and randomly initialize the population to complete the initial configuration;

Iteratively calculate the dry density increment, and determine whether the individual compaction degree and the number of iterations meet the requirements. If not, continue to calculate the dry density increment. If the conditions are met, save the compaction degree sequence, and then calculate the fitness of the individual population. value and sort;

Execute iterative optimization processes such as genetic algorithm selection, crossover, and mutation, and output the final results.

Further, the fitness value of the individual population is calculated using the compaction total energy function, and the formula is as follows:

In the formula, fitness is the fitness function, n is the number of vibrations required for the entire compaction process, fi is the dry density value of the i-1th state, A0i is the vibration parameter selected for the i-th state, and E _i is the vibration parameter of the i-th state. Vibration compaction energy, W is the static load, M _p is the eccentricity.

A parameter optimization system for vibration compaction of high-speed rail filler is also provided, which includes a data input terminal, a background operation and processing terminal and a data display terminal. The background operation and processing terminal adopts the method described in any one of claims 1-9. The computational processing end of the algorithm is used to optimize the parameters of vibration compaction of high-speed railway fillers.

The invention discloses a method for optimizing vibration compaction parameters of high-speed rail filler based on the energy minimum principle. Through a comprehensive deep learning algorithm and target optimization algorithm, the vibration compaction parameters are improved in real time during the compaction process. First, a hyperbolic model is constructed through experiments and mathematical theory, and the original data is calculated and constructed; BP neural network is used to establish a dry density increase model. Pre-quantity Based on the energy minimum principle, a genetic algorithm was used to construct a vibration parameter improvement method for the compaction process. The compaction parameters were adjusted in real time based on the compaction status of the filler, effectively improving the compaction quality and efficiency, making the filler better and more efficient. To quickly reach the optimal compaction state, a vibration compaction parameter optimization system for high-speed rail fillers based on the energy minimum principle was developed. The method of the invention solves the problem that during the traditional vibration compaction process, the compaction vibration parameters cannot be adjusted in real time, the excitation energy cannot be well controlled, and the vibration compaction equipment is prone to "jumping vibration", resulting in increased particle breakage and uneven settlement of the roadbed. Disease and other issues.

The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments. Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention.

Description of drawings

The drawings that form a part of the present invention are used to assist the understanding of the present invention. The contents provided in the drawings and their relevant descriptions in the present invention can be used to explain the present invention, but do not constitute an improper limitation of the present invention. In the attached picture:

Figure 1 is a flow chart of a parameter optimization method for vibration compaction of high-speed rail fillers.

Figure 2 is a flow chart for building a dry density incremental prediction model based on the improved BP neural network.

Figure 3 shows the BP neural network model architecture based on the improved BP algorithm.

Figure 4 is the flow chart of the GA dynamic optimization model.

Figure 5 is a 6-bit binary encoding diagram of the variable.

Figure 6 shows the system interface of the high-speed rail filler vibration compaction parameter optimization system based on the energy minimum principle.

Figure 7 shows the training set and validation set loss function curves of the two types of dry density incremental models.

Figure 8 is a comparison chart of the test results of the dry density increment prediction model before and after the improvement.

Figure 9 shows the final vibration parameter optimization results.

Figure 10 shows the vibration optimization parameters and compaction curves in each period.

Detailed ways

The present invention will be clearly and completely described below in conjunction with the accompanying drawings. A person of ordinary skill in the art will be able to implement the present invention based on these descriptions. Before describing the present invention in conjunction with the accompanying drawings, it should be particularly pointed out that:

The technical solutions and technical features provided in each part of the present invention, including the following description, can be combined with each other if there is no conflict.

In addition, the embodiments of the present invention mentioned in the following description are generally only some embodiments of the present invention, rather than all the embodiments. Therefore, based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts should fall within the scope of protection of the present invention.

Regarding terms and units in this invention. In the description and claims of the present invention and related parts, the term "includes "include", "have" and any variations thereof are intended to cover non-exclusive inclusion.

The steps of a parameter optimization method for vibration compaction of high-speed rail fillers are shown in Figure 1:

Step (1): Collect and construct original data, specifically using the hyperbolic model to fit the dry density of the compaction process, and calculate and construct the original data;

Step (2) Establish a dry density increment prediction model. Specifically, on the basis of obtaining training data, a BP neural network is used to establish a dry density increment prediction model;

Step (3) Establish constraint conditions, specifically using compaction degree indicators to evaluate compaction quality and establish compaction degree constraint conditions;

Step (4) Establish a dynamic optimization model based on GA, specifically solve the vibration parameter optimization process based on the GA algorithm, and establish a dynamic optimization model based on GA;

Step (5) outputs the best solution, specifically determining the dynamic optimization result described after parameter optimization processing as the best solution for vibration compaction.

As shown in Figure 2, the construction of the dry density incremental prediction model based on the improved BP neural network includes data set construction, model training, and model testing. First, the data set components are included in the following steps: original data input, data preprocessing, and data set partitioning, where the data set is divided into test set, verification set, and training set; then model training is performed, including the steps in order: optimizing the BP neural network architecture. , continue training at the breakpoint, and then determine whether the prediction error of the verification set is less than the set error value. If not, return to the optimization of the BP neural network architecture. If yes, enter the model test; finally perform the model test, including the steps: Save Model, model calling, model testing, determine whether the test error is less than the set error value, if so, save the final model, if not, return to model training.

On the basis of obtaining training data, an improved BP neural network is used to establish a dry density increment prediction model;

Use compaction degree indicators to evaluate the compaction quality and establish compaction degree constraints to ensure the compaction quality of the filler;

The vibration parameter optimization process is solved based on the GA algorithm, and a dynamic optimization model based on GA is established based on the principle of minimizing energy in the vibration compaction process to achieve real-time improvement and optimization of vibration compaction parameters;

The dynamic optimization results after parameter optimization processing are determined as the best solution for vibration compaction.

In the described step (1): the calculation and construction of original data are mainly divided into three steps: 1) Determine the hyperbolic functional relationship between the dry density and vibration parameters of the high-speed rail filler vibration compaction process; 2) Determine the model parameters a and b through experiments ;3) Calculate the dry density data sets under different combinations of vibration frequency f, amplitude A and corresponding dry density p, and use this as the original data.

In the described step 1): Determining the hyperbolic functional relationship between the dry density and vibration parameters of the high-speed rail filler vibration compaction process is divided into three steps:

①Set the basic equation of the hyperbolic function as follows:

In the formula, a and b are variable parameters respectively.

②In order to eliminate the dimensional relationship of different variables, the dry density is normalized. The processing process is as follows:

In the formula, y ^* is the standardized value; y, y _min , and y _max are the actual value of the variable and the maximum and minimum values of the standardized interval, respectively.

③ Substitute the above formula into the formula in ① to get the fitting model of dry density and vibration number, as shown below:

In the formula, ρ _d is the current state density; ρ _max is the compaction steady state dry density value under the current working conditions; ρ ₀ is the initial state dry density value; n is the number of vibrations; a and b are model parameters.

In the described step 2): determine the model parameters a and b through experiments, and obtain the model parameters and fitting correlation coefficients under each working condition as shown in Table 1 below:

Table 1 Model parameters and correlation coefficients

As shown in Figure 2, in step (2): the construction of the dry density increment prediction model is mainly divided into three steps:

①'To construct a high-speed railway filler vibration compaction parameter data set, first preprocess the data to normalize it to between [0,1], the specific formula is as follows; and according to the data volume proportion is 70%, The data set is randomly divided into training set, validation set and test set at a ratio of 15% and 15%.

x _scale d＝x _std *(max-min)+min

In the formula, x is the data to be normalized, x _{min (axis=0)} is a row vector composed of the minimum value in each column, x _{max (axis=0)} is a row vector composed of the maximum value in each column, max is the maximum value of the interval to be mapped, the default is 1, min is the minimum value of the interval to be mapped, the default is 0, x _std is the standardized result, x _scaled is the normalized result.

②’ Build an optimized BP neural network model architecture, input the training set and verification set to perform model training, and determine whether the verification set error is less than the set error. If it is greater, modify the model architecture and retrain. If it is less, proceed to the next step.

As shown in Figure 3, in step ②': the optimized BP neural network model architecture can be divided into an input layer, a hidden layer and an output layer. The number of neurons in the input layer is 3, and the amplitude A _0i and A at the current moment are input at the same time. Frequency f _i and dry density p _di at the previous moment, the number of neurons in the output layer is 1, and the dry density increment at the current moment is output. Through the dry density incremental prediction model, the nonlinear functional relationship between the model input and output can be constructed. The function is expressed as follows:

In the formula, f is the mapping relationship, A _0i and _fi are the amplitude and frequency at the current moment respectively, p _di is the dry density at the previous moment, is the dry density increment at the current moment

In the step ②': Compared with the traditional BP algorithm, the improved BP neural network algorithm has the characteristics of fast convergence speed and high prediction accuracy. On the basis of the traditional BP neural network, the learning rate improver AdamOptimizer optimization algorithm is introduced, which can improve the traditional BP neural network algorithm. The BP algorithm is prone to problems such as local optimality, sample dependence, and unadjustable learning rate. Reasonably adjust the learning rate according to changes in the loss function to accelerate model convergence and improve prediction accuracy.

In the step ②': the AdamOptimizer improved algorithm is an optimized version of the conventional optimization algorithm (such as SGD, AdaGrad algorithm). On the basis of the SGD algorithm, first-order momentum and second-order momentum are added. On the basis of the AdaGrad algorithm, Correction of deviations has the characteristics of fast convergence and high stability; and the first-order moment estimation and second-order moment estimation of the gradient are used to dynamically adjust the learning rate of each parameter. The main improvement formula of the AdamOptimizer improvement algorithm is as follows:

In the formula, l _r0 is the initial learning rate, lr _t is the learning rate at time t, is the exponential decay rate estimated by the first moment, is the exponential decay rate estimated for the second moment.

③'Use the divided test set to verify the prediction ability of the optimized BP neural network model that has been trained. Use MSE and MAE (as shown in the following formula) to determine whether the test set error is less than the set error. If it is greater, return to step 2) Retrain the model, and if less than, save the final dry density incremental prediction model.

In the formula, N represents the number of predicted samples, y _t and represent actual values and predicted values respectively.

In the described step (3): the compaction degree index is used to evaluate the compaction quality. According to the requirements for the compaction degree of the bottom layer of the base bed specified in the "High-speed Railway Subgrade Specification" to reach more than 95%, it is determined that the individual compaction degree should meet the requirements of greater than 95%. The constraint condition of 0.95 is to ensure the compaction quality of the filler. The formula is as follows:

In the formula, is the dry density value after vibration n times, is the maximum value of dry density.

As shown in Figure 4, in step (4): the GA-based dynamic optimization model is obtained by the following steps:

Set the initial parameters of the GA algorithm such as population size (PS), maximum genetic algebra (MAXGEN), selection probability SP, crossover probability CP, mutation probability Mu and use binary coding to randomly initialize the population. Each chromosome in the population includes two genes (vibration frequency, amplitude), as shown in Figure 5, each variable occupies a 6-bit binary code;

Iteratively calculate the dry density increment to determine whether the individual compaction degree is greater than 0.95 or whether the number of iterations reaches the number of chromosomes. If not, continue to calculate the dry density increment. If the conditions are met, save the compaction degree sequence and calculate again. Calculate the fitness value of the individual population according to the total compaction energy function (as shown in the following formula), and sort them according to the energy minimum principle;

In the formula, fitness is the fitness function, n is the number of vibrations required for the entire compaction process, fi is the dry density value of the i-1th state, A0i is the vibration parameter selected for the i-th state, and E _i is the vibration parameter of the i-th state. Vibration compaction energy, W is the static load, M _p is the eccentricity.

Execute iterative optimization processes such as genetic algorithm selection, crossover, and mutation, and output the final vibration parameter optimization results;

The embodiments and implementation processes of the complete method according to the content of the present invention are as follows:

Calculate and construct raw data:

The hyperbolic model was used to fit the dry density during the compaction process and the original data was obtained. The data volume was 7500 groups. Some of the data are shown in Table 2 below;

Table 2

Establish a dry density increment prediction model:

First, the dry density increment is calculated based on the current dry density, and the overall data is normalized. Part of the processed data is shown in Table 3 below.

table 3

Secondly, the data set is randomly divided into three parts: training set, verification set, and test set, with the data volume accounting for 70%, 15%, and 15% respectively. Import the Keras deep learning library and other third-party libraries in the python3.8.1 environment, set the number of input layer neurons to 3, the number of hidden layer neurons to 100, the number of output layer neurons to 1, and the initial learning rate to 0.001 , the number of training iterations is 100, build an improved BP neural network dry density incremental prediction model architecture, then import the training set and verification set to perform model training, and compare it with the traditional BP neural network model to obtain the model's performance in the training process The change rules of the training set loss value (referred to as loss) and the verification set loss value (referred to as val_loss) are shown in Figure 7. It is easy to know that the loss and val_loss of the traditional BP neural network model decrease slowly and continue to oscillate during the training process. The loss curve has not converged at the end of 100 rounds; while the improved BP neural network model has a decrease in loss and val_loss during the training process. high speed. After the fifth iteration, loss and val_loss began to converge without oscillation. This is because the improved BP neural network introduced the momentum method to accelerate convergence. At the same time, during the entire iterative operation process, val_loss is smaller than loss. At the end of the iteration, both loss and val_loss have converged, and both values are close to 0, indicating that the improved BP neural network model has better training effect, and in the verification set It has good generalization ability. In summary, the improved BP neural network model proposed by the present invention has the characteristics of adjustment and fast convergence, and solves the problems of the traditional BP neural network model that converges slowly and easily falls into the local optimum.

Finally, save the above-trained improved BP neural network dry density incremental prediction model, input 1125 sets of test sets that have not participated in the training into the dry density incremental prediction model, conduct model testing, and use MSE and MAE to evaluate the prediction results, and Compared with the traditional BP neural network model, the prediction results are compared with the actual values, as shown in Figure 8. It is easy to know that the traditional BP neural network model can predict most of the data well, but the prediction effect is not good for some prominent points of dry density increment; while the improved BP neural network model has strong prediction ability, and for dry density Incremental ups and downs can be well predicted. The MSE and MAE values calculated by the improved BP neural network model are 4.5×10 ^-6 and 1.4×10 ^-3 respectively, while the MSE and MAE values calculated by the traditional BP neural network model are 6.35×10 ^-4 and 5.1×10 respectively. ^-3 , indicating that the prediction accuracy of the improved BP neural network is higher than that of the traditional BP neural network. In summary, the improved BP neural network model proposed by the present invention not only performs well on the training set and test set, but also has high prediction accuracy and strong generalization ability on the test set, and can be well adapted to dry density increment Prediction.

Finally, save the final dry density incremental prediction model based on the improved BP algorithm and apply it in the subsequent dynamic optimization process.

Establish a compaction constraint higher than 0.95.

Establish a dynamic optimization model based on GA.

First, implement the GA algorithm in the python3.8.1 environment, analyze the parameters of the genetic algorithm through experiments, determine the reasonable parameters of the genetic algorithm, and set the optimal population size (PS) of the GA algorithm to 150 and the maximum genetic generation (MAXGEN) to 200 , the selection probability SP is 0.9, the crossover probability CP is 0.6, and the mutation probability Mu is 0.05. Using binary coding, 500 chromosomes are established. Each chromosome includes two genes (vibration frequency, nominal amplitude), and each variable occupies a 6-bit binary code.

Secondly, call the dry density increment prediction model, iteratively calculate the dry density increment, and determine whether the individual compaction degree is greater than 0.95 or whether the number of iterations reaches the number of chromosomes. If not, continue to calculate the dry density increment. If the conditions are met, , then save the compaction degree sequence, then calculate the fitness value of the individual population according to the total compaction energy function, and sort them according to the energy minimum principle.

Finally, iterative optimization processes such as selection, crossover, and mutation are executed. After reaching the maximum number of iterations, the final vibration parameter optimization results are output, as shown in Figure 9. To achieve the specified compaction degree, 9 complete vibrations are required, which means a total of 180 vibration compactions. As can be seen from Figure 9(a), the optimized vibration frequency gradually increases from the initial 10.15Hz to 27.58Hz with the vibration compaction process. This optimization result is in line with the existing two-degree-of-freedom vibration compaction model for compaction equipment. The relationship with the optimal vibration frequency between soil masses; Figure 9(b) shows the adjustment scheme of the nominal amplitude after the improvement. It can be seen that the optimized vibration amplitudes are smaller than the amplitude scheme before the improvement, and as the vibration compaction process progresses The amplitude increased from 0.15mm to 120 times of continuous vibration, and the amplitude change gradually stabilized, finally reaching 0.34mm. Figure 9(c) shows the comparison of energy output before and after the improvement. It can be seen that the improvement The energy output after the improvement was less than the energy output before the improvement throughout the entire compaction process, and the energy was ultimately reduced by 127.58J, accounting for 25.61% of the energy output before the improvement; therefore, the GA algorithm proposed by the present invention The dynamic optimization method can select appropriate vibration parameters according to the current compaction state of the soil, and effectively reduce the energy during the vibration process while ensuring 95% compaction, which can effectively improve the compaction efficiency and reduce the wear on the instrument. .

The dynamic optimized structure after parameter optimization is determined as the best solution for vibration compaction, as shown in Table 4 below.

Table 4 Vibration parameter optimization scheme

Correspondingly, parameters are input into the vibration compaction parameter optimization system. After running the system, the vibration optimization parameters and compaction curves for each period can be output on the interface, as shown in Figure 10.

The test results based on the optimization of vibration compaction parameters of high-speed rail fillers show that under the action of this optimization method, the energy output is effectively reduced by 127.58J, accounting for 25.61% of the energy output before improvement, that is, this method can guarantee 95% compaction degree. It can effectively reduce the energy in the vibration process, effectively improve the compaction efficiency and reduce the wear and tear on the instrument. It can be used in intelligent vibration compaction of high-speed rail fillers and serves the optimization of vibration compaction parameters.

The invention discloses a method for optimizing vibration compaction parameters of high-speed rail filler based on the energy minimum principle. Through a comprehensive deep learning algorithm and target optimization algorithm, the vibration compaction parameters are improved in real time during the compaction process. First, a hyperbolic model is constructed through experiments and mathematical theory, and the original data is calculated and constructed; BP neural network is used to establish a dry density increase model. The volume prediction model is used, and then a genetic algorithm is used to construct a method for improving the vibration parameters of the compaction process based on the energy minimum principle. The compaction parameters are adjusted in real time based on the compaction status of the filler, effectively improving the compaction quality and efficiency, making the filler better and more efficient. To reach the optimal compaction state faster, a vibration compaction parameter optimization system for high-speed rail fillers based on the energy minimum principle was developed. The method of the invention solves the problem that during the traditional vibration compaction process, the compaction vibration parameters cannot be adjusted in real time, the excitation energy cannot be well controlled, and the vibration compaction equipment is prone to "jumping vibration", resulting in increased particle breakage and uneven settlement of the roadbed. Disease and other issues.

The relevant contents of the present invention have been described above. A person of ordinary skill in the art will be able to implement the present invention based on these descriptions. Based on the above contents of the present invention, all other embodiments obtained by those of ordinary skill in the art without any creative work should fall within the scope of protection of the present invention.

Claims (7)

1. A parameter optimization method for vibration compaction of high-speed rail filler, which is characterized by: including the following steps:

   (1) Use the hyperbolic model to fit the dry density during the compaction process, and calculate and construct the original data;

   (2) On the basis of obtaining training data, use BP neural network to establish a dry density incremental prediction model;

   (3) Use compaction degree indicators to evaluate compaction quality and establish compaction degree constraints;

   (4) Solve the vibration parameter optimization process based on the GA algorithm and establish a dynamic optimization model based on GA;

   (5) Determine the dynamic optimization results described after parameter optimization processing as the best solution for vibration compaction;

   The steps for establishing the dry density increment prediction model in step (2) include:

   a. Construct a high-speed railway filler vibration compaction parameter data set, including the steps: input original data, preprocess the data, and divide the training set, verification set and test set;

   b. Build the BP neural network model architecture, input the training set and verification set to perform model training, and determine whether the verification set error is less than the set error. If it is greater, modify the model architecture and retrain. If it is less, proceed to the next step;

   c. Use the test set to verify the prediction ability of the BP neural network model, and determine whether the test set error is less than the set error. If it is greater, return to step b to retrain the model. If it is less, save the final dry density increment prediction model;

   In the described step (3): the formula of the compaction degree constraint is as follows:
   

   In the formula, is the dry density value after vibration n times, is the maximum value of dry density;

   The GA-based dynamic optimization model in step (4) is obtained by the following steps:

   A. Set the initial parameters of the GA algorithm and randomly initialize the population to complete the initialization configuration;

   B. Call the dry density increment prediction model, iteratively calculate the dry density increment, and determine whether the individual compaction degree meets the compaction degree constraint or whether the number of iterations reaches the number of chromosomes. If not, continue the dry density increment calculation. , if the conditions are met, the compaction degree sequence is saved, and then the fitness values of the population individuals are calculated and sorted;

   C. Execute iterative optimization processes such as genetic algorithm selection, crossover, and mutation, and output the final results.
2. A parameter optimization method for vibration compaction of high-speed rail filler as claimed in claim 1, characterized in that: the formula of the hyperbolic model in step (1) is:
   

   In the formula, ρ _d is the current state density; ρ _max is the compaction steady state dry density value under the current working conditions; ρ ₀ is the initial state. State dry density value; n is the number of vibrations; a and b are model parameters.
3. A parameter optimization method for vibration compaction of high-speed rail filler as claimed in claim 1, characterized in that: the dry density increment prediction model constructs a nonlinear functional relationship between model input and output, and the function is expressed as follows:
   

   In the formula, f is the mapping relationship, A _0i and _fi are the amplitude and frequency at the current moment respectively, p _di is the dry density at the previous moment, is the dry density increment at the current moment.
4. A parameter optimization method for vibration compaction of high-speed rail filler as claimed in claim 1, characterized in that:

   The data is preprocessed using the data normalization method, the formula is as follows:
   
   x _scaled = x _std *(max-min)+min

   In the formula, x is the data to be normalized, x _{min (axis=0)} is a row vector composed of the minimum value in each column, x _{max (axis=0)} is a row vector composed of the maximum value in each column, max is the maximum value of the interval to be mapped, the default is 1, min is the minimum value of the interval to be mapped, the default is 0, x _std is the standardized result, x _scaled is the normalized result.
5. A parameter optimization method for vibration compaction of high-speed railway filler as claimed in claim 1, characterized in that: the BP neural network is an optimized BP neural network, and the optimized BP neural network is a traditional BP neural network by introducing The learning rate improver is optimized by the AdamOptimizer optimization algorithm;

   The AdamOptimizer improved algorithm is an improved algorithm that uses the first-order moment estimate and the second-order moment estimate of the gradient to dynamically adjust the learning rate of each parameter in a conventional optimization algorithm. The improved formula of the AdamOptimizer improved algorithm is:
   

   In the formula, l _r0 is the initial learning rate, lr _t is the learning rate at time t, is the exponential decay rate estimated by the first moment, is the exponential decay rate estimated for the second moment.
6. A parameter optimization method for vibration compaction of high-speed rail filler as claimed in claim 1, characterized in that: calculating the fitness value of individuals in the population is calculated using the compaction total energy function, and the formula is as follows:
   

   In the formula, fitness is the fitness function, n is the number of vibrations required for the entire compaction process, fi is the dry density value of the i-1th state, A0i is the vibration parameter selected for the i-th state, and E _i is the vibration parameter of the i-th state. Vibration compaction energy, W is the static load, M _p is the eccentricity.
7. A parameter optimization system for vibration compaction of high-speed rail filler, which is characterized in that it includes a data input terminal, a background operation and processing terminal and a data display terminal. The background operation and processing terminal adopts the method described in any one of claims 1-6. A parameter optimization method for high-speed rail filler vibration compaction is used to perform computational processing on the computational processing end.