Permeable grout grouting effect prediction method and system based on PSO-RBF

Abstract

The invention discloses a permeable grout grouting effect prediction method and system based on PSO-RBF, and relates to the field of grouting repair engineering.The method comprises the following steps that the porosity and the water content of different pavement base layers are collected, grouting influence factor data are obtained, the grouting influence factor data comprise the single-hole grouting amount, the grouting hole depth and the grouting hole distance, and the grouting influence factor data are obtained; the porosity, the moisture content and the grouting influence factor data are divided into a training set and a test set, a grouting coring sample compressive strength prediction model is generated through a particle swarm optimization RBF neural network model, a predicted value of the grouting compressive strength is obtained through prediction, and a grout diffusion depth predicted value is obtained through the same steps. And calculating a comprehensive grouting performance evaluation coefficient according to the two groups of predicted values and measured values, comparing the comprehensive grouting performance evaluation coefficient with a preset error threshold value, and judging that a predicted result meets a precision requirement if the comprehensive grouting performance evaluation coefficient does not exceed the error threshold value. Through the particle swarm optimization-based RBF neural network, permeation type slurry grouting restoration can be realized, and accurate grouting effect prediction can be realized.

Description

Technical Field

[0001] The invention discloses a method and a system for predicting the grouting effect of a permeable slurry based on PSO-RBF, and belongs to the technical field of grouting repair of underground diseases. Background Art

[0002] Grouting, as one of the main technical means for reinforcement and repair of underground engineering defects, has been widely used in water conservancy engineering, transportation engineering, tunnel engineering and mining engineering. The essence of grouting is the process of slurry penetrating and diffusing in the medium and then consolidating with the medium. The penetration and diffusion process depends to a large extent on the characteristics of the grouting material. Therefore, the selection of suitable grouting materials plays a key role in the treatment of highway defects. In engineering practice, due to the lack of scientific theoretical and technical guidance, the grouting parameters of permeable polymer slurry are mostly determined by field experience, which will cause great uncertainty in the treatment effect of engineering defects. For example, excessive grouting pressure will cause loosening of the pavement foundation, while too low grouting pressure will cause the area to be treated to fail to reach the designed strength; improper selection of grouting hole type and hole spacing will not only affect the reinforcement effect of the area to be treated, but also affect the construction period and the cost of defect repair, etc. These have seriously hindered the promotion and application of permeable polymer grouting technology in highway defect treatment and the standardization of process flow. Summary of the invention

[0003] The technical problem solved by the present invention is that it is difficult to accurately predict the grouting effect when grouting is used to reinforce and repair underground diseases of existing highways. A method and system for predicting the grouting effect of permeable slurry based on PSO-RBF is provided.

[0004] The present invention is implemented by the following technical solutions:

[0005] The present invention first discloses a method for predicting the effect of permeable slurry grouting based on PSO-RBF, which comprises the following steps:

[0006] S1, setting up multiple groups of permeable polymer grouting test groups, obtaining the moisture content and porosity of the highway base where the different grouting test groups are located, and the grouting influencing factor data of each grouting test group, wherein the grouting influencing factor data includes the single hole grouting amount, the grouting hole depth and the grouting hole spacing, and mapping the numbers of each grouting test group and the grouting influencing factor data one by one to form a first data set;

[0007] S2, obtaining the measured parameters of the grouting test effect after grouting of each grouting test group, wherein the measured parameters of the grouting test effect include the measured value of the compressive strength of the grouting core sample and the measured value of the slurry diffusion depth;

[0008] S3, select RBF neural network to build a grouting effect prediction model, and optimize the initialization parameters of RBF neural network through PSO algorithm;

[0009] S4. The predicted values of the grouting effect include the predicted compressive strength of the grouting core samples and the predicted grout diffusion depth. Select the water content of the road base where the grouting test group is located, combine it with the first data set and the measured compressive strength of the grouting core samples to construct the first training set and the first test set. Input the first training set into the grouting effect prediction model for model training, and evaluate the grouting effect prediction model through the first test set to obtain the grouting sample compressive strength prediction model. Select the porosity of the road base where the grouting test group is located, combine it with the first data set and the measured grout diffusion depth to construct the second training set and the second test set. Input the second training set into the grouting effect prediction model for model training, and evaluate the grouting effect prediction model through the second test set to obtain the grouting slurry diffusion depth prediction model.

[0010] S5. Use the grouting sample compressive strength prediction model and the grouting slurry diffusion depth prediction model to predict the actual grouting effect of the permeable grout.

[0011] In the method for predicting the grouting effect of the permeable grout based on PSO-RBF of the present invention, further, in step S3, the RBF neural network kernel function selects the Gaussian kernel function, initializes the RBF neural network structure, uses the mean square error as the fitness function, optimizes the key parameters of the RBF neural network by minimizing the fitness function, takes the key parameters of the RBF neural network as each particle in the PSO, updates the individual optimal position and the global optimal position of the particles through the PSO algorithm, iterates step by step until convergence, and then constructs the grouting effect prediction model using the RBF neural network optimized by the PSO parameters.

[0012] In the method for predicting the grouting effect of the permeable grout based on PSO-RBF of the present invention, further, the key parameters of the RBF neural network model include the center parameter, the width parameter, and the weight parameter.

[0013] In the method for predicting the grouting effect of the permeable grout based on PSO-RBF of the present invention, further, the PSO algorithm updates the individual optimal position and the global optimal position of the RBF neural network key parameter particles through the following formula:

[0014] v k+1 =v k +c 1 *rand()*(pbest k -present k )+c 2 *rand()*(gbest k -present k )(1)

[0015] present k+1 = present k + v k+1 (2)

[0016] v k 、v k+1 are the particle velocities at the current time k and the next time k + 1 respectively. persent k 、persent k+1 are the current particle and the particle at the next time. The rand() function generates a random number between (0, 1), c 1 , c 2 is the learning factor. In each iteration, each particle is updated with two optimal values. Among them, pbest k is the individual optimal position of the current particle, and gbest k is the current optimal value tracked by the particle swarm optimizer, representing the global optimal position of any particle in the particle swarm.

[0017] In the method for predicting the grouting effect of permeable grout based on PSO - RBF of the present invention, further, the steps of setting the training set and the test set in step S4 are as follows: Select N grouting test groups from M grouting test groups to obtain the training set. The input variables of the first training set and the second training set are the grouting pressure, single - hole grouting volume, grouting hole spacing of the corresponding grouting test group, and the porosity or water content of the highway subgrade, which are four influencing factors. Among them, the first training set selects porosity as the fourth influencing factor, and the second training set selects water content as the fourth influencing factor. The target output is the predicted value of the grouting effect of the corresponding grouting test group. Among them, the first training set outputs the predicted value of the compressive strength of the grouting core sample, and the second training set outputs the predicted value of the grout diffusion depth; Obtain the test set from the remaining M - N grouting test groups.

[0018] In the method for predicting the grouting effect of permeable grout based on PSO - RBF of the present invention, further, in step S4, the training processes of the grouting sample compressive strength prediction model and the grout diffusion depth prediction model are as follows:

[0019] The grouting effect prediction model constructed by using the optimized RBF neural network model parameters is as follows:

[0020]

[0021] Among them, x is the input variable, is the output variable of the model prediction. For the training of the compressive strength prediction model of the grouting sample, the input variables are the moisture content, single-hole grouting volume, grouting hole depth, and grouting hole spacing in the first training set, and the output variable is the predicted value of the compressive strength of the grouting core sample. For the training of the grouting slurry diffusion depth prediction model, the input variables are the porosity, single-hole grouting volume, grouting hole depth, and grouting hole spacing in the second training set, and the output variable is the predicted value of the slurry diffusion depth, ω j is the weight parameter of the model, c j is the center parameter of the model, σ j is the width parameter of the model, Φ is the Gaussian kernel function, and K is the number of hidden layer nodes of the model;

[0022] The predicted values of the grouting effect output by the compressive strength prediction model of the grouting sample and the grouting slurry diffusion depth prediction model according to the first test set or the second test set are compared with the measured parameters of the corresponding grouting test effect. The error evaluation indexes including the mean square error, mean absolute error, and determination coefficient are used to evaluate the model performance, and the RBF neural network model structure of the compressive strength prediction model of the grouting sample and the grouting slurry diffusion depth prediction model is adjusted according to the error evaluation indexes.

[0023] In the penetration-type slurry grouting effect prediction method based on PSO-RBF of the present invention, further, in the step S4, during the training process of the grouting effect prediction model, the k-fold cross-validation is used to evaluate the model performance for the predicted values of the grouting effect output by the first training set and the second training set.

[0024] In the penetration-type slurry grouting effect prediction method based on PSO-RBF of the present invention, further, in the step S4, the predicted value of the compressive strength generated by the compressive strength prediction model of the grouting sample for the first test set and the predicted value of the slurry diffusion depth generated by the grouting slurry diffusion depth prediction model for the second test set are obtained, and the mean square error, mean absolute error, and relative root mean square error between them and the measured parameters of the grouting test effect are calculated respectively. The comprehensive grouting performance evaluation coefficient Y is obtained through the following formula n ,

[0025] Y n = aMSE 1 + bMSE 2 + cMAE 1 + dMAE 2 + eRRMSE 1 + fRRMAE 2 ,

[0026] where, MSE 1 is the root mean square error between the predicted value and the measured value of the compressive strength, MAE 1 is the mean absolute error between the predicted value and the measured value of the compressive strength, RRMSE1 is the relative root mean square error between the predicted value and the measured value of the compressive strength, MSE 2 is the root mean square error between the predicted value and the measured value of the slurry diffusion depth, MAE 2 is the mean absolute error between the predicted value and the measured value of the slurry diffusion depth, RRMSE 2 is the relative root mean square error between the predicted value and the measured value of the slurry diffusion depth. a, b, c, d, e, f are the proportionality coefficients set for each parameter respectively, satisfying a + b + c + d + e + f = 1;

[0027] The calibration error threshold is Y threshold , and the comprehensive grouting performance evaluation coefficient Y n are compared as follows:

[0028] If Y n ≤Y threshold , it is determined that the accuracy rate of the predicted value meets the accuracy requirements, and the compressive strength prediction model of the grouting sample and the slurry diffusion depth prediction model of the grouting slurry are used as the prediction references for the grouting effect in the actual grouting process;

[0029] If Y n >Y threshold , it is determined that the accuracy rate of the predicted value does not meet the accuracy requirements, and the training set test samples of the compressive strength prediction model of the grouting sample and the slurry diffusion depth prediction model of the grouting slurry are continuously increased until the accuracy rate of the predicted value meets the accuracy requirements.

[0030] The present invention also discloses a penetration-type slurry grouting effect prediction system based on PSO-RBF, which is used to execute the above-mentioned penetration-type slurry grouting effect prediction method of the present invention, including

[0031] a highway subgrade sample collection module, which obtains the moisture content and porosity of the highway subgrade where different grouting test groups are located;

[0032] a grouting influencing factor collection module, which obtains the grouting influencing factor data of each grouting test group. The grouting influencing factor data includes the single-hole grouting volume, the grouting hole depth, and the grouting hole spacing, and maps the numbers of each grouting test group and the grouting influencing factor data one by one to form a first data set;

[0033] a test data collection module, which obtains the measured parameters of the grouting test effect after grouting for each grouting test group. The measured parameters of the grouting test effect include the measured value of the compressive strength of the grouting core sample and the measured value of the slurry diffusion depth;

[0034] The compressive strength prediction model generation module selects an RBF neural network optimized by the PSO algorithm to construct a grouting effect prediction model. It selects the moisture content of the road base where the grouting test group is located, combines it with the first data set and the measured compressive strength values of the grouting core samples to construct a first training set and a first test set. The first training set is input into the grouting effect prediction model for model training, and the grouting effect prediction model is evaluated through the first test set to obtain a grouting sample compressive strength prediction model;

[0035] The diffusion depth prediction model generation module selects an RBF neural network optimized by the PSO algorithm to construct a grouting effect prediction model. It selects the porosity of the road base where the grouting test group is located, combines it with the first data set and the measured values of the slurry diffusion depth to construct a second training set and a second test set. The second training set is input into the grouting effect prediction model for model training, and the grouting effect prediction model is evaluated through the second test set to obtain a grouting slurry diffusion depth prediction model;

[0036] The prediction error calculation module calculates the mean square error, mean absolute error, and relative root mean square error between the predicted compressive strength value and the predicted slurry diffusion depth value and the measured parameters of the grouting test effect in the test set;

[0037] The prediction performance evaluation module calculates a comprehensive grouting performance evaluation coefficient based on the mean square error, mean absolute error, and relative root mean square error generated by the prediction error calculation module, and compares it with the error threshold to evaluate the prediction effect of the slurry grouting effect.

[0038] The present invention also discloses a computer device, including:

[0039] One or more processors;

[0040] A memory storing one or more computer programs;

[0041] Wherein, the processor calls the computer program to implement the above-mentioned permeable slurry grouting effect prediction method of the present invention.

[0042] Compared with the prior art, the present invention combines a radial basis function neural network (RBF) based on particle swarm optimization (PSO) with the prediction of the grouting effect of underground diseases on roads. From the perspective of the pore diseases generated underground according to the basic properties of the semi-rigid road base collected, permeable slurries are mostly used for the grouting repair of pore media. According to the characteristics of the permeable slurry, the present invention selects the porosity and moisture content of the road base soil, which directly affect the grouting effect of the permeable slurry. The porosity reflects the size of the void space in the soil. The higher the porosity, the stronger the permeability of the soil, and the slurry can more easily penetrate into the pores of the soil to fill the voids and enhance the grouting effect. The moisture content determines the degree of wetness of the soil and affects the fluidity and distribution of the slurry in the soil. When the moisture content of the soil is moderate, the fluidity of the slurry is the best, and it can effectively diffuse and penetrate into all parts of the soil; if the moisture content is too low, the slurry may be difficult to penetrate, while too high a moisture content may affect the stability of the slurry. Therefore, the porosity and moisture content are the key parameters determining whether the permeable slurry can fully penetrate and repair underground diseases. Combining these two parameters can comprehensively evaluate the permeability of the soil and the adaptability of the slurry, thereby more accurately predicting the repair effect of the slurry, optimizing the grouting design and construction process.

[0043] The present invention simultaneously obtains the data of the influencing factors affecting the grouting effect, including the single-hole grouting volume, grouting depth, and grouting hole spacing. The moisture content, single-hole grouting volume, grouting depth, and grouting hole spacing data of multiple groups of permeable polymer grouting experimental groups are divided into two parts of data sets as the first training set and the first test set. The grouting sample compressive strength prediction model is trained through the first training set, and the grouting sample compressive strength model is evaluated through the first test set; using the same steps, a grouting diffusion depth prediction model is established according to the porosity, single-hole grouting volume, grouting depth, and grouting hole distance of multiple groups of permeable polymer grouting experimental groups, and the predicted diffusion depth value is generated. According to the two groups of predicted values and the measured parameters of the grouting effect of the corresponding grouting test groups, the mean square error, mean absolute error, and relative root mean square error are calculated. The comprehensive grouting performance evaluation coefficient is generated from the mean square error, mean absolute error, and relative root mean square error of each group. Compared with the pre-set error threshold, if it does not exceed the error threshold, it is judged that the performance of the prediction model is good and meets the accuracy requirements; otherwise, it is judged that the performance of the prediction model does not meet the accuracy requirements. The grouting effect of the permeable polymer is predicted through the combined action of various factors, improving the prediction accuracy.

[0044] In the present invention, the PSO-RBF model selected for predicting the grouting effect of permeable grout is a hybrid optimization algorithm that combines the Particle Swarm Optimization (PSO) algorithm with the Radial Basis Function (RBF) neural network, and is used to improve the performance of the RBF neural network in predicting the grouting effect through permeable grout parameters. The main process includes initializing the particle swarm, defining the fitness function, iteratively updating, training the RBF neural network model, and finally outputting the optimal model parameters. The Particle Swarm Optimization (PSO) algorithm can perform random search and evolution by simulating biological activities. During the movement of particles, the optimal solution is selected through local and global optimal control to establish the RBF model. This algorithm avoids the problem that traditional optimization methods are prone to falling into local optima and significantly improves the performance of the RBF neural network. Its advantages lie in the high efficiency of the optimization process, the absence of manual parameter setting, and strong global search ability, making it particularly suitable for complex non-linear problems such as prediction, classification, and system modeling. The RBF neural network can effectively simulate the non-linear seepage behavior of grout in soil through the characteristics of its local basis functions, and accurately capture the complex interaction between grout and soil. Its structure is simple, the training speed is fast, and it can handle the complex relationship between soil parameters (such as porosity, water content, etc.) and grouting effect. In addition, the RBF neural network has a high response sensitivity and can make accurate predictions for different soil types and grout characteristics, especially suitable for materials such as permeable grout with strong diffusivity and adaptability. After combining with the PSO algorithm, the optimization process of the RBF further enhances the global search ability and avoids the problem of local optimal solutions, thus providing high-precision and reliable prediction results in grouting effect prediction. This enables the PSO-RBF model to not only train efficiently in predicting the grouting effect of permeable grout, but also accurately model complex non-linear systems, improving the accuracy of engineering prediction and the actual application effect of the project.

[0045] In summary, the method and system for predicting the grouting effect of permeable grout based on PSO-RBF provided by the present invention predict the grouting effect of permeable grout by optimizing the RBF neural network based on the particle swarm. The PSO-RBF model can well match the characteristics of highway underground diseases suitable for permeable grout, achieve a relatively accurate prediction of the grouting effect of using permeable grout for repair, and ensure the standardization of the permeable grout repair process and the certainty of the grouting effect. Description of the Drawings

[0046] Figure 1 It is a schematic diagram of the step flow of the method for predicting the grouting effect of permeable grout based on PSO-RBF of the present invention.

[0047] Figure 2This is a schematic structural diagram of the penetration grout injection effect prediction system based on PSO-RBF of the present invention. Detailed implementation manners

[0048] To make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below in conjunction with specific embodiments.

[0049] Embodiment 1

[0050] As Figure 1 shown, the specific steps of the penetration grout injection effect prediction method based on PSO-RBF of the present invention are as follows.

[0051] S1. Set multiple groups of penetration polymer grouting test groups, ensure that the basic property parameters of the roadbeds of each test group are not completely the same, obtain the moisture content and porosity of the roadbeds where different grouting test groups are located, as well as the grouting influencing factor data of each grouting test group. The grouting influencing factor data includes single-hole grouting volume, grouting hole depth, and grouting hole spacing. Map the numbers of each grouting test group and the grouting influencing factor data one by one to form a first data set.

[0052] For example, in this embodiment, 16 grouting tests are set. The label of the i-th group is S i , and the porosity in the S i -th group of tests is marked as n i , the moisture content is marked as e i , and i = 1, 2, 3,..., 16;

[0053] Record the basic property parameters of the roadbed soil samples used in each grouting test group, as shown in Table 1:

[0054] Table 1 Basic properties of roadbed soil samples used in grouting tests

[0055] Grouting test group number Porosity / % Water content / % 1 25 10 2 21 8 3 14 11 4 37 12 5 42 20 6 23 15 7 34 19 8 12 13 9 36 11 10 44 32 11 41 12 12 19 14 13 23 16 14 38 23 15 29 34 16 40 25

[0056] Set 16 groups of grouting test samples for establishing a grouting model.

[0057] In addition, the influencing factors of the grouting effect of the permeable polymer slurry include, in addition to the characteristics of the grouted medium (i.e., the road base), the single-hole grouting volume, the spacing between grouting holes, and the depth of grouting holes. Among them, the grouting volume is an important index to be considered during grouting. A small grouting volume may result in the grouting effect not meeting the expected requirements, the slurry being unable to completely fill the voids, weakening the overall mechanical properties of the solidified body, and possibly causing secondary leakage problems. If the grouting volume is large, it will lead to waste of materials and increased costs, and will also cause uneven diffusion of the slurry, even forming incomplete blocking areas, weakening the overall blocking performance. At the same time, excessive grouting will also result in insufficient density and uneven strength of the local solidified body. After the excess slurry solidifies, it may form a loose structure, resulting in a relatively high surface strength of the solidified body but a fragile internal structure.

[0058] The spacing between grouting holes is also an important index to be considered. When the spacing between grouting holes is too small, the diffusion range of the slurry may overlap, resulting in excessive concentration of the slurry, causing waste of materials and affecting the construction efficiency at the same time; when the spacing between grouting holes is too large, it is difficult for the diffusion range of the slurry to completely cover the target area, reducing the overall diffusion depth, and will also cause uneven distribution areas in the solidified body, resulting in a decline in the structural integrity, and thus affecting the compressive strength and durability.

[0059] For the 16 groups of grouting test groups set, record the single-hole grouting volume, the depth of grouting holes, and the spacing between grouting holes in each test group, as shown in Table 2:

[0060] Table 2 First data set

[0061]

[0062]

[0063] Generate the first data set. The data in the first data set is used to combine the grouting parameters adopted in the grouting test and the basic property data of the road base soil layer to complete the establishment of the prediction model.

[0064] S2. Obtain the measured parameters of the grouting test effect after grouting for each grouting test group. The measured parameters of the grouting test effect include the measured compressive strength of the grouting core sample and the measured diffusion depth of the slurry. The specific operation process for obtaining the measured compressive strength of the grouting core sample and the measured diffusion depth of the slurry is as follows.

[0065] Detection: First, use ground penetrating radar for detection before construction. Determine the road disease area and disease degree through the analysis of the detected data. Sample and analyze the porosity n i and moisture content e i .

[0066] Mark the grouting holes: Arrange the grouting holes according to the design requirements.

[0067] Drilling: According to the designed drilling depth, use a percussion drill to drill to the designed depth at the marked grouting hole positions. The drilling is required to be vertical. During drilling, the road surface must be kept clean without being polluted. The drill bit should pass through the treated iron basin to drill at the grouting hole positions to prevent dust diffusion, and the drilling area should be cleaned in time to prevent dust pollution to the environment. The drill bit should be aligned with the mark, then gently press and slowly advance. After the drill bit enters the ground, it can be drilled normally to prevent accidental injury to personnel.

[0068] Hole cleaning: Before inserting the grouting pipe, use high-pressure air from an air compressor to blow the accumulated soil in the formed grouting holes from the bottom to facilitate effective grouting.

[0069] Inserting the grouting pipe: According to the technical requirements of polymer grouting and the on-site conditions of the test section, use cutting tools to cut the PPI pipe to about 30 cm, and install a pressure-holding valve at the top. The main function of the pressure-holding valve is to make the slurry continue to penetrate under continuous pressure after the grouting gun is pulled out, and it is also for the purpose of being able to close at any time for pressure-holding and re-grouting. Insert the PPI pipe with the installed grouting head into the grouting hole.

[0070] Injecting permeable polymer slurry: According to the construction and grouting technical requirements, to prevent the polymer from spraying onto the road surface and causing pollution to the road surface, use a clamp to firmly hold the injection gun and the injection cap. Combining the road surface detection data, the specific on-site conditions and engineering experience, implement grouting at the grouting location according to the design. When grouting, mix the A and B component materials in a mass ratio of 1:1 at the grouting equipment, and then transport them through the grouting PPI pipe to the road surface disease area, where a chemical reaction occurs and the material changes from liquid to solid. The grouting pressure during grouting should not be greater than 0.5 Mpa.

[0071] Grout according to the single-hole grouting volume designed in the grouting method table. Use the two-component permeable polymer 9100-2A and 9100-2B of Wanhua Company for grouting. The grouting pressure at the hole mouth during grouting should not be greater than 0.5 Mpa. The single-hole grouting volumes are set to 500 g, 1000 g, 1500 g, and 2000 g respectively. After the single-hole reaches the response grouting volume, stop grouting and close the grouting valve and the pressure-holding valve. The single-hole grouting is completed, and continue to the next grouting hole for grouting. After the valve is closed, the grouting hole is in a pressure-holding state. At this time, the slurry continues to diffuse under the action of pressure until the pressure gradually dissipates. After 30 minutes, open the pressure-holding switch of the grouting hole, and there should be no obvious back-pressure slurry spraying out of the grouting hole. The grouting is completed.

[0072] Maintenance: After grouting, heavy vehicles are prohibited from passing, and traffic is prohibited for 2 - 3 hours.

[0073] Detect the grouting effect after grouting: Use a special tool to pull out the grouting pipe. Mainly use ground penetrating radar for post-grouting detection. The falling weight deflectometer can be used as an auxiliary detection means in special sections to analyze the grouting filling and reinforcement effect. If the requirements are met, the grouting is completed; if not, additional grouting is required until the requirements are met. The detection method and location are the same as before grouting. After the detection meets the standard, core samples are taken from this section, and core samples are drilled in the grouting sections under different working conditions to check the grouting cementation situation and compactness. The measured compressive strength of the grouting core sample and the measured grouting diffusion depth are obtained from the core sample detection.

[0074] Seal the holes: To prevent rainwater erosion, damage the road surface, and maintain the overall image of the road surface, use special rubber plugs or sealants, etc. to seal the holes.

[0075] Clean the environment: Use an iron brush to treat the grouting holes and the polluted road surface, use a broom to clean the construction operation area, and then use a blower for cleaning.

[0076] S3. Select the RBF neural network to construct a grouting effect prediction model, and optimize the initial parameters of the RBF neural network through the PSO algorithm.

[0077] In the step S3, the RBF neural network kernel function selects the Gaussian kernel function, initializes the RBF neural network structure, uses the mean square error as the fitness function, optimizes the key parameters of the RBF neural network, including the center parameter, width parameter, and weight parameter, by minimizing the fitness function. Take each parameter of the RBF neural network as each particle in the PSO, update the individual optimal position and global optimal position of the particle through the PSO algorithm, iterate step by step until convergence, and then use the parameters optimized by the PSO to construct a grouting effect prediction model.

[0078] The RBF neural network, also known as the radial basis function neural network, is a three-layer feedforward neural network for local approximation, consisting of an input layer, a hidden layer, and an output layer. The function of the hidden layer is to complete the non-linear transformation from input to output, map low-dimensional data to high-dimensional space, and thus realize the linear separability of the input data. The activation functions of the hidden layer neurons and the output layer neurons are different. The hidden layer uses the radial basis function as the transformation function, and the node activation function of the output layer is a linear function to realize the space mapping from the hidden layer to the output layer. In this embodiment, the Gaussian function is selected as the kernel function. Perform a non-linear transformation on the input data. The specific expression of the Gaussian function is as follows:

[0079]

[0080] s is the diffusion speed of the kernel function, corresponding to the width parameter σ optimized by the PSO. j, the default value is 1, r is the Euclidean distance from the sample of the input data set to the sample center point, and the size of r depends on the sample data characteristics. The output Q(x i ) of the hidden layer of the RBF neural network is:

[0081]

[0082] s is the diffusion speed of the kernel function, x M is the input vector, C i is the center vector of the Gaussian function, corresponding to the center parameter c optimized by PSO j

[0083] The output Y of the output layer is:

[0084]

[0085] w i is the weight, corresponding to the weight parameter ω optimized by PSO j .

[0086] The larger the diffusion speed s, the smoother the fitting function, but the approximation error increases; correspondingly, the smaller s is, the smaller and more accurate the approximation error is, but the function is not smooth enough and the performance is poor. Therefore, the selection of the diffusion speed parameter s becomes the key to improving the prediction accuracy of the radial basis neural network.

[0087] By minimizing the fitness function, the key parameters of the RBF neural network are optimized, including the center parameter, width parameter and weight parameter. Each parameter of the RBF neural network is used as each particle in the PSO algorithm, and the individual optimal position and global optimal position of the particle are updated through the PSO algorithm.

[0088] The PSO algorithm is also called the particle swarm optimization algorithm, which can perform random search and evolution by simulating biological activities. During the particle movement process, the optimal solution is selected through local optimal and global optimal control and a model is established. Each particle has an adaptability determined by the model and a vector determining the direction and distance. Then all particles will search for the best target value. In the particle swarm optimization algorithm, the system is initialized as a group of random solutions and the optimal solution is searched through update iterations. The particle swarm algorithm maintains a group of particles flying at a certain speed in the n-dimensional search space. The particles have no weight and no volume. Assuming that the position of the i-th particle in the n-dimensional space is the vector X = (x 1 ,x 2 ,…,x n ), and its velocity vector is V = (v 1 ,v 2 ,…,v n ), the PSO algorithm updates the individual optimal position and global optimal position of the particle through the following formula:

[0089] vk+1 = v k + c 1 * rand() * (pbest k - present k ) + c 2 * rand() * (gbest k - present k );

[0090] present k+1 = present k + v k+1 .

[0091] v k 、v k+1 are the particle velocities at the current time k and the next time k + 1 respectively, persent k 、persent k+1 are the current particle and the particle at the next time, the rand() function generates a random number between (0, 1), c 1 , c 2 is the learning factor. In each iteration, each particle is updated with two optimal values. In each iteration, each particle is updated with two optimal values. Among them, pbest k is the individual optimal position of the current particle, gbest k is the current optimal value tracked by the particle swarm optimizer, representing the global optimal position of any particle in the particle swarm.

[0092] The parameters in the RBF neural network model are gradually iterated by the PSO algorithm until convergence, and then the grouting effect prediction model is constructed using the parameters optimized by PSO as follows:

[0093]

[0094] Among them, ω j is the weight parameter of the grouting effect prediction model, corresponding to the weight w i of the kernel function, c j is the center parameter of the grouting effect prediction model, corresponding to the center vector C i of the kernel function, σ j is the width parameter of the grouting effect prediction model, corresponding to the diffusion speed s of the kernel function, Φ is the Gaussian kernel function K is the number of hidden layer nodes, and x is the input vector.

[0095] S4. The predicted values of the grouting effect include the predicted compressive strength of the grouting core samples and the predicted grout diffusion depth. The optimized grouting effect prediction model in step S3 is used to predict the compressive strength of the grouting core samples and the grout diffusion depth respectively, and the comprehensive grouting effect is predicted.

[0096] The water content of the road base where the grouting test group is located is selected and combined with the measured compressive strength values of the grouting core samples in the first dataset to be divided into a first training set and a first test set. The compressive strength of the grouting core samples in the first training set and the first test set is predicted through the grouting effect prediction model to generate predicted compressive strength values. Among them, the first training set is input into the grouting effect prediction model for model training, and the grouting effect prediction model is evaluated through the first test set to obtain the grouting sample compressive strength prediction model.

[0097] The steps of the first training set and the first test set include selecting N grouting test groups from the M grouting test groups for training to obtain the training set. Among them, the input variables of the first training set are the column vectors of the N grouting test groups. The first training set for predicting the compressive strength of the grouting core samples includes four influencing factors of grouting pressure, single-hole grouting volume, grouting hole spacing in the first dataset of these N grouting test groups, and the corresponding road base water content. The target output is the predicted compressive strength value of the grouting core samples in the measured parameters of the grouting test effect of the corresponding grouting test group. The remaining M - N grouting test groups obtain the first test set according to the same data classification.

[0098] In the grouting effect prediction model in step S3, the input variables x are selected as the water content, single-hole grouting volume, grouting hole depth, and grouting hole spacing in the first training set, and the target output is the predicted compressive strength value of the grouting core samples. The data of the first training set is used as input to train the optimized grouting effect prediction model in step S4. During the model training process, the mean square error (MSE) is used as the fitness function, as follows:

[0099]

[0100] where, y i is the measured compressive strength value of the grouting core samples of the i-th sample in the first training set, is the predicted compressive strength value of the i-th sample output by the model, and N is the number of samples in the first training set. According to the training results, the positions and velocities of the particle swarm optimization algorithm are dynamically updated, and the center parameters, width parameters, and weight parameters of the RBF neural network are optimized through multiple iterations to finally obtain the optimal parameter combination.

[0101] Combine the data in the first data set with the moisture content and the measured compressive strength values of the grouting core samples of each test sample to generate a first training set and a first test set. Extract the first 12 groups of experimental group samples in Table 3, the data table of the compressive strength prediction model, as the first training set, and use the sample data of the remaining four experimental groups as the first test set.

[0102] Table 3 Data table of the compressive strength prediction model

[0103]

[0104]

[0105] In Table 3, the data table of the compressive strength prediction model, P1, P2, P3, and P4 respectively represent the predicted compressive strength values obtained by predicting the first test set according to the grouting sample compressive strength prediction model that has completed training. During the training process of RBF, the first training set is not only used for optimizing the model parameters but also outputs corresponding predicted values. The main functions of these predicted values are mainly reflected in error feedback and model adjustment: First, RBFNN performs a non-linear mapping on the input data of the first training set through the Gaussian radial basis function, calculates the predicted values and compares them with the actual values to obtain the error for adjusting the weights to optimize the network performance; Second, the predicted values of the first training set can be used to monitor the training effect of the model to ensure that the error gradually converges, and at the same time help to judge whether there are overfitting or underfitting problems; In addition, analyzing the prediction results of the first training set can also guide the adjustment of the radial basis center, width parameters, and network structure, making the generalization ability of the model on the test set stronger and improving the prediction accuracy of the compressive strength.

[0106] In this embodiment, the cross-validation method is used to evaluate the model performance and verify its prediction accuracy and generalization ability. The grouting effect prediction model obtains the grouting sample compressive strength prediction model through the above model training. Input the data of the first test set into the grouting sample compressive strength prediction model to predict the grouting sample compressive strength, compare it with the measured value of the grouting core sample compressive strength, and verify the generalization ability of the grouting sample compressive strength prediction model through evaluation indicators (such as MSE, MAE, R2) to ensure its high accuracy and strong adaptability in predicting the grouting effect.

[0107] Among them, the mean square error (MSE) evaluates the square error between the predicted compressive strength value and the measured value of the grouting core sample compressive strength. The formula is as follows:

[0108]

[0109] The mean absolute error (MAE) measures the average deviation between the predicted compressive strength value and the measured value of the grouting core sample compressive strength. The formula is as follows:

[0110]

[0111] The coefficient of determination (R 2 ) quantifies the ability of the model to explain the variation of the independent variables, and the formula is as follows:

[0112]

[0113] where n is the number of test samples in the second test set, and y i is the measured compressive strength of the grouting core sample of the i-th sample in the training set, is the predicted compressive strength of the i-th sample.

[0114] Select the porosity of the road base where the grouting test group is located, combine it with the first data set and the measured value of the slurry diffusion depth to construct the second training set and the second test set. Predict the slurry diffusion depth in the second training set and the second test set through the grouting effect prediction model to generate the predicted value of the slurry diffusion depth. Input the second training set into the grouting effect prediction model for model training, and evaluate the grouting effect prediction model through the second test set to obtain the grouting slurry diffusion depth prediction model.

[0115] The prediction of the slurry diffusion depth also uses the grouting effect prediction model optimized by the particle swarm optimization algorithm in step S3. Different from the prediction of the compressive strength of the grouting core sample, in this step, for the training of the grouting slurry diffusion depth prediction model, the input variables include the porosity, single-hole grouting volume, grouting hole depth, and grouting hole spacing in the second training set, and the output variable is the predicted value of the slurry diffusion depth.

[0116] The training and evaluation process of the grouting slurry diffusion depth prediction model is the same as the process of predicting the predicted value of the compressive strength of the grouting sample by the grouting sample compressive strength prediction model, and will not be elaborated here.

[0117] Generate the grouting sample compressive strength prediction model and the grouting slurry diffusion depth prediction model, and analyze the applicability and generalization ability of the model according to the distribution of the training and test errors. If the test error is large, adjust the parameters of the particle swarm optimization algorithm (such as the inertia weight, acceleration factor) or the RBF neural network structure (such as the number of hidden layer nodes). After analysis and verification, the optimized grouting sample compressive strength prediction model can be used to predict the compressive strength of the grouting site with unseen data or new working conditions, providing a scientific basis for the optimization of the construction parameters of the grouting project.

[0118] Adjust the structure of the RBF neural network model to optimize the performance of the prediction model for the compressive strength of grouting samples and the prediction model for the diffusion depth of grouting slurry. The adjustment includes changing the number of centers of the RBF network, the width parameter of the radial basis function, the weight optimization strategy, and the selection of the training algorithm to improve the fitting ability and generalization performance of the model. The specific adjustment method is usually optimized according to the error evaluation index (such as the mean square error MSE, the mean absolute error MAE, and the coefficient of determination R 2 ) to ensure that the prediction model can more accurately reflect the actual change trend of the test data. The above respectively evaluates the models of the prediction model for the compressive strength of grouting samples and the prediction model for the diffusion depth of grouting slurry. This embodiment also comprehensively evaluates the prediction effect of the grouting effect by the mean square error, the mean absolute error, and the relative root mean square error between the predicted compressive strength values and the predicted diffusion depth values of the grouting slurry predicted by the first test set and the second test set and the measured parameters of the grouting test effect.

[0119] The mean square error represents the average value of the square of the error between the predicted value and the actual value. The mean absolute error represents the average value of the absolute error between the predicted value and the actual value. The relative root mean square error reflects the ratio of the error to the actual value. In machine learning and modeling, all three are important indicators for evaluating model performance. According to the mean square deviation, the root mean square error, and the relative root mean square error of each generated predicted value, calculate the overall comprehensive grouting performance evaluation coefficient of the grouting effect, and compare it with the pre-set error threshold. If it does not exceed the error threshold, it is judged that the error between the generated predicted value and the measured value meets the expected requirements, and the generated predicted value can be used as a reference for the grouting effect in the actual grouting process. Otherwise, it is judged that the accuracy of the predicted value does not meet the precision requirements.

[0120] In step S4, obtain the predicted compressive strength value generated by the prediction model for the compressive strength of grouting samples for the first test set and the predicted diffusion depth value of the grouting slurry generated by the prediction model for the diffusion depth of grouting slurry for the second test set, calculate the mean square error, the mean absolute error, and the relative root mean square error between them and the measured parameters of the grouting test effect, and then calculate the comprehensive grouting performance evaluation coefficient Y through the following formula n ,

[0121] Y n =aMSE 1 +bMSE 2 +cMAE 1 +dMAE 2 +eRRMSE 1 +fRRMAE 2 ,

[0122] where, MSE 1 is the mean square error between the predicted compressive strength value and the measured compressive strength value, MAE 1is the mean absolute error between the predicted compressive strength and the measured compressive strength, RRMSE 1 is the relative root mean square error between the predicted compressive strength and the measured compressive strength, MSE 2 is the mean square error between the predicted grout diffusion depth and the measured grout diffusion depth, MAE 2 is the mean absolute error between the predicted grout diffusion depth and the measured grout diffusion depth, RRMSE 2 is the relative root mean square error between the predicted grout diffusion depth and the measured grout diffusion depth. a, b, c, d, e, f are the proportionality coefficients set for each parameter respectively, satisfying a + b + c + d + e + f = 1.

[0123] The calibration error threshold is Y threshold , and the error thresholds of porosity and moisture content need to be adjusted according to the specific requirements of the project and the actual characteristics of the soil to ensure that the difference between the predicted grouting effect and the actual repair effect is within an acceptable range, guaranteeing the safety and effectiveness of the project. Usually, the porosity error is ±3% to ±10%, and for projects with high requirements, the error threshold can be reduced to ±3%; the moisture content error is ±2% to ±5%, and for projects with high-precision requirements, the error threshold should be controlled within ±2% as much as possible.

[0124] Compare the comprehensive grouting performance evaluation coefficient Y n with the error threshold of Y threshold and make a comparison

[0125] If Y n ≤Y threshold , it is determined that the accuracy rate of the predicted value meets the accuracy requirements, and the compressive strength prediction model of the grouting sample and the grout diffusion depth prediction model of the grouting slurry are used as the prediction references for the grouting effect in the actual grouting process;

[0126] If Y n >Y threshold , it is determined that the accuracy rate of the predicted value does not meet the accuracy requirements and cannot be used as a reference for the grouting effect in the actual grouting process. Check whether there are problems with the parameters of each prediction model or continue to increase the training set test samples to improve the prediction accuracy. If high-precision prediction results are required, the set training set test samples can be increased, or the k-fold cross-validation method can be used. The data is divided into k subsets, and each time one subset is selected as the test set, and the rest are used as the training set, and loop k times to obtain more stable results.

[0127] S5. Use the compressive strength prediction model of the grouting sample and the grout diffusion depth prediction model of the grouting slurry to predict the actual grouting effect of the permeable grout.

[0128] When predicting the actual grouting effect, the water content of the road base is selected and combined with the single-hole grouting volume, grouting hole depth, and grouting hole spacing to input into the compressive strength prediction model of the grouting sample, and the predicted value of the compressive strength of the grouting sample is output; the porosity of the road base is selected and combined with the single-hole grouting volume, grouting hole depth, and grouting hole spacing to input into the grouting slurry diffusion depth prediction model, and the predicted value of the grouting slurry diffusion depth is output. The predicted values of the compressive strength of the grouting sample and the grouting slurry diffusion depth are summarized to obtain the prediction of the actual grouting effect.

[0129] Example Two

[0130] See Figure 2 , which discloses a PSO-RBF-based prediction system for the grouting effect of permeable slurry of the present invention. This system is used to execute the prediction method for the grouting effect of permeable slurry in Example One, and specifically includes a road base sample collection module, a grouting influencing factor collection module, a test data collection module, a compressive strength prediction model generation module, a diffusion depth prediction model generation module, a prediction error calculation module, and a prediction performance evaluation module.

[0131] The road base sample collection module obtains the water content and porosity of the road base where different grouting test groups are located.

[0132] The grouting influencing factor collection module obtains the grouting influencing factor data of each grouting test group. The grouting influencing factor data includes the single-hole grouting volume, grouting hole depth, and grouting hole spacing. The numbers of each grouting test group and the grouting influencing factor data are mapped one by one to form a first data set.

[0133] The test data collection module obtains the measured parameters of the grouting test effect after grouting for each grouting test group. The measured parameters of the grouting test effect include the measured value of the compressive strength of the grouting core sample and the measured value of the slurry diffusion depth.

[0134] The compressive strength prediction model generation module selects an RBF neural network optimized by the PSO algorithm to construct a grouting effect prediction model. The water content of the road base where the grouting test group is located is selected and combined with the first data set and the measured value of the compressive strength of the grouting core sample to construct a first training set and a first test set. The first training set is input into the grouting effect prediction model for model training, and the grouting effect prediction model is evaluated through the first test set to obtain a compressive strength prediction model for the grouting sample.

[0135] The diffusion depth prediction model generation module selects an RBF neural network optimized by the PSO algorithm to construct a grouting effect prediction model. It selects the porosity of the highway base layer where the grouting test group is located, combines it with the first data set and the measured values of the slurry diffusion depth to construct a second training set and a second test set. The second training set is input into the grouting effect prediction model for model training, and the grouting effect prediction model is evaluated through the second test set to obtain a grouting slurry diffusion depth prediction model.

[0136] The prediction error calculation module calculates the mean square error, mean absolute error, and relative root mean square error between the predicted compressive strength and the predicted slurry diffusion depth and the measured parameters of the grouting test effect in the test set.

[0137] The prediction performance evaluation module calculates a comprehensive grouting performance evaluation coefficient based on the mean square error, mean absolute error, and relative root mean square error generated by the prediction error calculation module, and compares it with the error threshold to evaluate the prediction effect of the slurry grouting effect.

[0138] Embodiment III

[0139] The present invention also provides a computer device, including: one or more processors and a memory storing one or more computer programs; wherein, the processor calls the computer program to implement the steps of the PSO-RBF-based permeable slurry grouting effect prediction method described in Embodiment I, and the specific implementation process can refer to the description of Embodiment I.

[0140] For the specific implementation process of each step, please refer to the description of the foregoing method.

[0141] It should be understood that in the embodiments of the present invention, the so-called processor may be a central processing unit (CPU), and this processor may also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or this processor may also be any conventional processor, etc. The memory may include a read-only memory and a random access memory, and provide instructions and data to the processor. A part of the memory may also include a non-volatile random access memory. For example, the memory may also store information about the device type.

[0142] In this text, the orientation or positional relationships indicated by terms such as "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", "vertical", "horizontal", etc. are based on the orientation or positional relationships shown in the drawings, and are only for the sake of clarity in expressing the technical solution and convenience in description. Therefore, they should not be construed as limitations on the present invention.

[0143] In this text, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion. In addition to the elements listed, it may also include other elements not specifically listed.

[0144] The above is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. The prediction method of permeable slurry grouting effect based on PSO-RBF is characterized by: The steps include: S1, setting up multiple groups of permeable polymer grouting test groups, obtaining the moisture content and porosity of the highway base where the different grouting test groups are located, and the grouting influencing factor data of each grouting test group, wherein the grouting influencing factor data includes the single hole grouting amount, the grouting hole depth and the grouting hole spacing, and mapping the numbers of each grouting test group and the grouting influencing factor data one by one to form a first data set; S2, obtaining the measured parameters of the grouting test effect after grouting of each grouting test group, wherein the measured parameters of the grouting test effect include the measured value of the compressive strength of the grouting core sample and the measured value of the slurry diffusion depth; S3, select RBF neural network to build a grouting effect prediction model, and optimize the initialization parameters of RBF neural network through PSO algorithm; S4. The predicted values ​​of the grouting effect include the predicted values ​​of the compressive strength of the grouting core sample and the predicted values ​​of the slurry diffusion depth. The moisture content of the highway base where the grouting test group is located is selected in combination with the first data set and the measured values ​​of the compressive strength of the grouting core sample to construct a first training set and a first test set. The first training set is input into the grouting effect prediction model for model training. The grouting effect prediction model is evaluated through the first test set to obtain the grouting sample compressive strength prediction model. The porosity of the highway base where the grouting test group is located is selected in combination with the first data set and the measured values ​​of the slurry diffusion depth to construct a second training set and a second test set. The second training set is input into the grouting effect prediction model for model training. The grouting effect prediction model is evaluated through the second test set to obtain the grouting slurry diffusion depth prediction model. S5. Use the grouting sample compressive strength prediction model and the grouting slurry diffusion depth prediction model to predict the actual grouting effect of the permeable slurry.

2. The method for predicting the effect of grouting of permeable slurry based on PSO-RBF according to claim 1 is characterized in that: In step S3, the RBF neural network kernel function selects a Gaussian kernel function, initializes the RBF neural network structure, takes the mean square error as the fitness function, optimizes the key parameters of the RBF neural network by minimizing the fitness function, takes the key parameters of the RBF neural network as each particle in the PSO, updates the individual optimal position and the global optimal position of the particle by the PSO algorithm, gradually iterates until convergence, and then uses the RBF neural network after the PSO optimization parameters to construct a grouting effect prediction model.

3. The method for predicting the effect of grouting of permeable slurry based on PSO-RBF according to claim 2 is characterized in that: The key parameters of the RBF neural network model include center parameter, width parameter and weight parameter.

4. The method for predicting the effect of grouting of permeable slurry based on PSO-RBF according to claim 3 is characterized in that: The PSO algorithm updates the individual optimal position and global optimal position of the key parameter particles of the RBF neural network through the following formula: v k+1 =v k +c1*rand()*(pbest k -present k )+c2*rand()*(gbest k -present k )(1) present k+1 =present k +v k+1 (2) v k 、v k+1 are the particle speeds at the current time k and the next time k+1, respectively, k 、persent k+1 The rand() function generates random numbers between (0,1) for the current particle and the next moment particle. c1 and c2 are learning factors. In each iteration, each particle is updated with two optimal values, where pbest k is the individual optimal position of the current particle, gbest k It is the current optimal value tracked by the particle swarm optimizer, which represents the global optimal position of any particle in the particle swarm.

5. The method for predicting the effect of grouting of permeable slurry based on PSO-RBF according to claim 1 is characterized in that: The steps of setting the training set and the test set in step S4 are as follows: select N groups of grouting test groups from the M groups of grouting test groups to obtain the training set, the input variables of the first training set and the second training set are the grouting pressure, single-hole grouting amount and grouting hole spacing of the corresponding grouting test group, and the porosity or water content of the highway base layer. The first training set selects porosity as the fourth influencing factor, and the second training set selects water content as the fourth influencing factor. The target output is the grouting effect prediction value of the corresponding grouting test group, wherein the first training set outputs the compressive strength prediction value of the grouting core sample, and the second training set outputs the slurry diffusion depth prediction value; obtain the test set from the remaining MN groups of grouting test groups.

6. The method for predicting the effect of grouting of permeable slurry based on PSO-RBF according to claim 5 is characterized in that: In step S4, the training process of the grouting sample compressive strength prediction model and the grouting slurry diffusion depth prediction model is as follows: The grouting effect prediction model constructed using the optimized RBF neural network model parameters is as follows: Among them, x is the input variable, is the output variable predicted by the model. For the training of the grouting sample compressive strength prediction model, the input variables are the water content, single hole grouting amount, grouting hole depth and grouting hole spacing in the first training set, and the output variable is the predicted value of the compressive strength of the grouting core sample. For the training of the grouting slurry diffusion depth prediction model, the input variables are the porosity, single hole grouting amount, grouting hole depth and grouting hole spacing in the second training set, and the output variable is the predicted value of the slurry diffusion depth. j is the weight parameter of the model, c j is the central parameter of the model, σ j is the width parameter of the model, Φ is the Gaussian kernel function, and K is the number of hidden layer nodes of the model; The grouting sample compressive strength prediction model and the grouting slurry diffusion depth prediction model are based on the grouting effect prediction value output by the first test set or the second test set, compared with the corresponding measured parameters of the grouting test effect, and the model performance is evaluated using error evaluation indicators including mean square error, mean absolute error and determination coefficient.

7. The method for predicting the effect of grouting of permeable slurry based on PSO-RBF according to claim 1 is characterized in that: In the step S4, during the training process of the grouting effect prediction model, the grouting effect prediction values ​​output by the first training set and the second training set are evaluated for model performance using k-fold cross validation.

8. The method for predicting the effect of grouting of permeable slurry based on PSO-RBF according to claim 1 is characterized in that: In step S4, the compressive strength prediction value generated by the grouting sample compressive strength prediction model for the first test set and the grouting slurry diffusion depth prediction value generated by the grouting slurry diffusion depth prediction model for the second test set are obtained, and the mean square error, mean absolute error and relative root mean square error between the measured parameters of the grouting test effect are calculated respectively, and the comprehensive grouting performance evaluation coefficient Y is obtained by the following formula n , Y n =aMSE1+bMSE2+cMAE1+dMAE2+eRRMSE1+fRRMAE2, Among them, MSE1 is the root mean square error between the predicted value and the measured value of compressive strength, MAE1 is the mean absolute error between the predicted value and the measured value of compressive strength, RRMSE1 is the relative root mean square error between the predicted value and the measured value of compressive strength, MSE2 is the root mean square error between the predicted value and the measured value of slurry diffusion depth, MAE2 is the mean absolute error between the predicted value and the measured value of slurry diffusion depth, RRMSE2 is the relative root mean square error between the predicted value and the measured value of slurry diffusion depth, a, b, c, d, e, f are the proportional coefficients corresponding to the settings of each parameter, satisfying a+b+c+d+e+f=1; The calibration error threshold is Y threshold , and the comprehensive grouting performance evaluation coefficient Y n In comparison: If Y n ≤Y threshold , then the accuracy of the predicted value is judged to meet the accuracy requirements, and the grouting sample compressive strength prediction model and the grouting slurry diffusion depth prediction model are used as the prediction reference of the grouting effect in the actual grouting process; If Y n >Y threshold , it is judged that the accuracy of the predicted value does not meet the precision requirement, and the training set test samples of the grouting sample compressive strength prediction model and the grouting slurry diffusion depth prediction model are continued to be increased until the accuracy of the predicted value meets the precision requirement.

9. A permeable slurry grouting effect prediction system based on PSO-RBF, used to execute the method according to any one of claims 1 to 8, characterized in that: include Highway base sample collection module, to obtain the moisture content and porosity of the highway base in different grouting test groups; A grouting influencing factor acquisition module is used to obtain the grouting influencing factor data of each grouting test group, wherein the grouting influencing factor data includes the single hole grouting amount, the grouting hole depth and the grouting hole spacing, and each grouting test group number and the grouting influencing factor data are mapped one by one to form a first data set; The test data acquisition module obtains the measured parameters of the grouting test effect after grouting of each grouting test group, wherein the measured parameters of the grouting test effect include the measured value of the compressive strength of the grouting core sample and the measured value of the slurry diffusion depth; The compressive strength prediction model generation module selects the RBF neural network optimized by the PSO algorithm to construct the grouting effect prediction model, selects the moisture content of the highway base where the grouting test group is located, combines the first data set and the measured compressive strength of the grouting core sample to construct the first training set and the first test set, inputs the first training set into the grouting effect prediction model for model training, and evaluates the grouting effect prediction model through the first test set to obtain the grouting sample compressive strength prediction model; The diffusion depth prediction model generation module selects the RBF neural network optimized by the PSO algorithm to construct the grouting effect prediction model, selects the porosity of the highway base where the grouting test group is located, combines the first data set and the measured value of the slurry diffusion depth to construct the second training set and the second test set, inputs the second training set into the grouting effect prediction model for model training, and evaluates the grouting effect prediction model through the second test set to obtain the grouting slurry diffusion depth prediction model; The prediction error calculation module calculates the mean square error, mean absolute error and relative root mean square error between the predicted values ​​of compressive strength and slurry diffusion depth and the actual measured parameters of the grouting test effect in the test set according to the predicted values; The prediction performance evaluation module calculates the comprehensive grouting performance evaluation coefficient based on the mean square error, mean absolute error and relative root mean square error generated by the prediction error calculation module, and compares it with the error threshold to evaluate the prediction effect of the grouting effect.

10. A computer device, characterized in that: include: one or more processors; a memory storing one or more computer programs; The processor calls the computer program to implement: The method for predicting the grouting effect of permeable slurry based on PSO-RBF as described in any one of claims 1 to 8.

Images