Data knowledge-driven intelligent design method for mix proportion of foam light soil

Abstract

The invention relates to a data knowledge-driven intelligent design method for the mix proportion of foam light soil. The method comprises the steps of obtaining foamed light soil performance data, and performing preprocessing to obtain a complete data set; respectively establishing an LSTM-based learner model, an SVM-based learner model and an XGBoost-based learner model, and establishing a foam light soil performance prediction model; training the foam light soil performance prediction model by adopting a complete data set to obtain a performance prediction value, calculating a precision difference between the performance prediction value and a performance true value, and optimizing the foam light soil performance prediction model according to the precision difference; inputting a formula ratio to be predicted into the trained foam light soil performance prediction model to obtain a foam light soil performance prediction value; and constructing a multi-objective design optimization model of the foam light soil, and solving the multi-objective design optimization model of the foam light soil by adopting a multi-objective optimization algorithm to obtain an intelligent design scheme of the foam light soil.

Description

Technical Field

[0001] The present application relates to the fields of materials science and computer technology, and in particular to a data knowledge-driven intelligent design method for foam lightweight soil mix ratio. Background Art

[0002] Foamed lightweight soil is a porous material. It is made of cement, water, aggregate, admixtures and other raw materials mixed in a specific proportion to form a slurry, which is then evenly mixed with fine foam (made of a foaming agent aqueous solution) and finally cast and cured. Due to its advantages such as light weight, good fluidity, easy construction, stable performance and little impact on the environment, foamed lightweight soil has been widely used in the construction of infrastructure such as roads, railways and airports in recent years.

[0003] In the prior art, the mix design of foam lightweight soil materials usually relies on experience and traditional tests. However, the complexity of foam lightweight soil components leads to nonlinear and discrete properties of foam lightweight soil. Traditional methods are difficult to accurately predict and adjust the properties of foam lightweight soil to adapt to various complex and special environments, resulting in instability and uncertainty in the design results.

[0004] Therefore, in the related technology, there is an urgent need for a method that can improve the accuracy and stability of foam lightweight soil mix design. Summary of the invention

[0005] Based on this, it is necessary to provide a data knowledge-driven intelligent design method for foam lightweight soil mix ratio that can improve the accuracy and stability of foam lightweight soil mix ratio design in response to the above technical problems.

[0006] In a first aspect, the present application provides a data knowledge-driven intelligent design method for foam lightweight soil mix ratio. The method comprises:

[0007] Obtain foam lightweight soil performance data and preprocess it to obtain a complete data set;

[0008] Establishing an LSTM-based learner model, an SVM-based learner model, and an XGBoost-based learner model respectively, and establishing a foam lightweight soil performance prediction model based on the LSTM-based learner model, the SVM-based learner model, and the XGBoost-based learner model;

[0009] Using a complete data set to train the foam lightweight soil performance prediction model to obtain a performance prediction value, calculating the accuracy difference between the performance prediction value and the actual performance value, and optimizing the foam lightweight soil performance prediction model according to the accuracy difference;

[0010] Input the formula ratio to be predicted into the trained foam lightweight soil performance prediction model to obtain the predicted value of the foam lightweight soil performance;

[0011] A multi-objective design optimization model of foam lightweight soil is constructed with cost and foam lightweight soil performance as the objectives, and a multi-objective optimization algorithm is used to solve the multi-objective design optimization model of foam lightweight soil based on the predicted value of the foam lightweight soil performance to obtain an intelligent design scheme for foam lightweight soil.

[0012] Optionally, in one embodiment of the present application, the preprocessing includes:

[0013] The water-cement ratio and foam injection rate are standardized, and the corresponding fluidity, standard settlement rate and compressive strength are transformed into normal distribution and reduced in dimension.

[0014] Optionally, in one embodiment of the present application, the LSTM-based learner model includes a forget gate, an input gate, and an output gate, wherein the forget gate is controlled by a sigmoid gate, and the output gate includes a sigmoid function and a tanh function.

[0015] Optionally, in one embodiment of the present application, the linear function expression of the SVM-based learner model is:

[0016]

[0017] Among them, α n and is the Lagrange multiplier, n is the number of support vectors, K(x n ,x) is the kernel function and b is the bias value.

[0018] Optionally, in one embodiment of the present application, the objective function of the XGBoost based learner model is:

[0019]

[0020] Where i is each instance, is the predicted value, y i is the true value, l is the loss function, Ω(f k ) is the complexity of each decision tree.

[0021] Optionally, in one embodiment of the present application, the establishing of a foam lightweight soil performance prediction model based on the LSTM-based learner model, the SVM-based learner model, and the XGBoost-based learner model includes:

[0022] Partitioning the complete data set to obtain a training set, a validation set, and a test set;

[0023] Input the training set and the test set into the LSTM base learner model, the SVM base learner model and the XGBoost base learner model to obtain feature values ​​and predicted values;

[0024] A linear regression LR model is used for training based on the characteristic values, and prediction is performed based on the predicted values ​​to obtain a foam lightweight soil performance prediction model.

[0025] Optionally, in one embodiment of the present application, the adopting a multi-objective optimization algorithm to solve the foam lightweight soil multi-objective design optimization model includes:

[0026] Initialize the population particles and define the design parameters of the foam lightweight soil multi-objective optimization model represented by each particle;

[0027] Calculating a fitness value associated with a cost-performance optimization model based on the design parameters;

[0028] Performing Pareto dominance updates on the design parameters to construct a Pareto frontier;

[0029] updating the speed and position of the design parameters according to the solution in the Pareto front;

[0030] The optimal solution for the multi-objective design of foam lightweight soil is determined based on the solution set of the Pareto frontier.

[0031] In a second aspect, the present application also provides a data knowledge-driven foam lightweight soil mix ratio intelligent design device. The device comprises:

[0032] The data set processing module is used to obtain the foam lightweight soil performance data and perform preprocessing to obtain a complete data set;

[0033] A foam lightweight soil performance prediction model establishment module is used to establish an LSTM-based learner model, an SVM-based learner model, and an XGBoost-based learner model, respectively, and establish a foam lightweight soil performance prediction model based on the LSTM-based learner model, the SVM-based learner model, and the XGBoost-based learner model;

[0034] A foam lightweight soil performance prediction model training module is used to train the foam lightweight soil performance prediction model using a complete data set to obtain a performance prediction value, calculate the accuracy difference between the performance prediction value and the actual performance value, and optimize the foam lightweight soil performance prediction model according to the accuracy difference;

[0035] The foam lightweight soil performance prediction module is used to input the formula ratio to be predicted into the trained foam lightweight soil performance prediction model to obtain the foam lightweight soil performance prediction value;

[0036] The foam lightweight soil intelligent design scheme determination module is used to construct a foam lightweight soil multi-objective design optimization model with cost and foam lightweight soil performance as the objectives, and adopt a multi-objective optimization algorithm to solve the foam lightweight soil multi-objective design optimization model based on the predicted value of the foam lightweight soil performance to obtain the foam lightweight soil intelligent design scheme.

[0037] In a third aspect, the present application further provides a computer device, wherein the computer device comprises a memory and a processor, wherein the memory stores a computer program, and the processor executes the steps of the methods described in the above embodiments.

[0038] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the method described in each of the above embodiments are implemented.

[0039] The above-mentioned data knowledge-driven foam lightweight soil mix ratio intelligent design method first obtains the foam lightweight soil performance data and performs preprocessing to obtain a complete data set; then, respectively establishes an LSTM-based learner model, an SVM-based learner model and an XGBoost-based learner model, and establishes a foam lightweight soil performance prediction model based on the LSTM-based learner model, the SVM-based learner model and the XGBoost-based learner model; then, the foam lightweight soil performance prediction model is trained using the complete data set to obtain a performance prediction value, the accuracy difference between the performance prediction value and the actual performance value is calculated, and the foam lightweight soil performance prediction model is optimized according to the accuracy difference; then, the formula ratio to be predicted is input into the trained foam lightweight soil performance prediction model to obtain the foam lightweight soil performance prediction value; finally, a foam lightweight soil multi-objective design optimization model is constructed with cost and foam lightweight soil performance as the objectives, and a multi-objective optimization algorithm is used to solve the foam lightweight soil multi-objective design optimization model based on the foam lightweight soil performance prediction value to obtain a foam lightweight soil intelligent design scheme. In other words, an intelligent design method for foam lightweight soil mix ratio is proposed. Through multi-module coordination, the structural design indicators are used to guide the mix ratio design of foam lightweight soil for widening roadbed, while meeting the design safety and economic requirements. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 An application environment diagram of a data knowledge-driven foam lightweight soil mix intelligent design method in one embodiment;

[0041] Figure 2 A schematic diagram of a flow chart of a data knowledge-driven foam lightweight soil mix intelligent design method in one embodiment;

[0042] Figure 3A schematic flow chart of specific steps of a data knowledge-driven foam lightweight soil mix intelligent design method in one embodiment;

[0043] Figure 4 A structural block diagram of a data knowledge-driven foam lightweight soil mix ratio intelligent design device in one embodiment;

[0044] Figure 5 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0045] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0046] The embodiment of the present application provides a data knowledge driven foam lightweight soil mix ratio intelligent design method, which can be applied to Figure 1 In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be integrated on the server 104, or it can be placed on the cloud or other network servers. Among them, the terminal 102 can be but is not limited to various personal computers, laptops, smart phones, tablets, Internet of Things devices and portable wearable devices. The Internet of Things devices can be smart speakers, smart TVs, smart air conditioners, smart car-mounted devices, etc. Portable wearable devices can be smart watches, smart bracelets, head-mounted devices, etc. The server 104 can be implemented with an independent server or a server cluster consisting of multiple servers.

[0047] In one embodiment, Figure 2 As shown in the figure, a data knowledge-driven intelligent design method for foam lightweight soil mix is ​​provided. Figure 1 The server in the example is used to illustrate the following steps:

[0048] S201: Acquire foam lightweight soil performance data and perform preprocessing to obtain a complete data set.

[0049] In the present application, firstly, the performance data of foamed lightweight soil are obtained through experiments, numerical simulation and literature research, including the water-cement ratio of foamed lightweight soil, the foam injection rate (V 泡沫 / V 水泥浆液×100%), fly ash, fluidity, standard settlement rate, compressive strength, etc., and then preprocess them to obtain a complete data set. A sample data set X is established based on parameter data such as the water-cement ratio of foamed lightweight soil and foam injection rate. The fluidity, standard settlement rate and compressive strength of foamed lightweight soil are used as target feature variables. A complete data set is established based on the target feature variables, in which the target feature variables correspond to the relevant operating parameter data in the sample data set X. Specifically, let X = {x1, x2, ...x i ,...,x N} represents a data set containing N samples, let Y = {y1, y2, ...y i ,...,y N}, which represents a data set containing N sample target variable values, where y i is the target variable corresponding to the i-th sample, x i and i Is a one-to-one correspondence, thus forming a complete data set:

[0050] U={(X,Y)}={(x1,y1),(x2,y2),...(x i ,y i ),...(x N ,y N )}.

[0051] Specifically, in one embodiment of the present application, the preprocessing includes:

[0052] The water-cement ratio and foam injection rate are standardized, and the corresponding fluidity, standard settlement rate and compressive strength are transformed into normal distribution and reduced in dimension.

[0053] In one embodiment of the present application, the sample data set X, i.e., the water-cement ratio and the foam injection rate, is standardized by a normalization method to eliminate the dimension of the feature, the standardized sample data set X is updated, and the complete data set is updated synchronously. Afterwards, the complete data set is processed by Box-Cox transformation to make the data in the complete data set close to the normal distribution, and to make the skewed distribution that does not conform to the model assumptions as close to the normal distribution as possible after the transformation. The complete data set is subjected to dimensionality reduction processing by the Pearson product-moment correlation coefficient method, and the feature variables with no correlation and weak correlation are screened out to achieve dimensional pruning. Specifically, the sample data set X is standardized according to the following formula:

[0054]

[0055] in, is the value of the sample data set X in the jth dimension, is the maximum value of the sample data set X in the jth dimension, It is the minimum value of the sample data set X in the jth dimension.

[0056] The updated complete dataset is:

[0057] v={(X,Y)}={(x1,y1),(x2,y2),...(x i ,y i ),...(x N ,y N )}.

[0058] The Box-Cox transformation of the complete data set v is performed according to the following formula:

[0059]

[0060] Among them, v i is any data in the complete data set v, is the data obtained after Box-Cox transformation, and γ is the value that minimizes the likelihood function.

[0061] The complete data set v is reduced in dimension according to the following formula:

[0062]

[0063] Among them, μ xj is the average value of all values ​​of the relevant operating parameter data in the sample data set X in the j dimension, μ y is the average value of all values ​​of the target variable, r j is the correlation coefficient between the relevant operating parameter data in the sample data set X and the target variable.

[0064] Among them, when the correlation coefficient is 0-0.09, the correlation coefficient is no correlation, when it is 0.1-0.3, the correlation coefficient is weak correlation, when it is 0.3-0.5, the correlation coefficient is medium correlation, and when it is 0.5-1.0, the correlation coefficient is strong correlation; by sorting the correlation coefficients, the correlation coefficients corresponding to no correlation and weak correlation are screened out.

[0065] S203: Establishing an LSTM-based learner model, an SVM-based learner model, and an XGBoost-based learner model respectively, and establishing a foam lightweight soil performance prediction model based on the LSTM-based learner model, the SVM-based learner model, and the XGBoost-based learner model.

[0066] In the embodiment of the present application, the complete data set after preprocessing is respectively used with the LSTM-based learner model, the SVM-based learner model, and the XGBoost-based learner model to establish a model, and obtain the LSTM-based learner model, the SVM-based learner model, and the XGBoost-based learner model. Afterwards, the LSTM-based learner model, the SVM-based learner model, and the XGBoost-based learner model are fused by the Stacking ensemble learning method to obtain a foam lightweight soil performance prediction model, wherein the input data is the water-cement ratio and the foam injection rate, and the output data is the fluidity, the standard settlement rate, and the compressive strength.

[0067] Specifically, in one embodiment of the present application, the LSTM-based learner model includes a forget gate, an input gate, and an output gate, wherein the forget gate is controlled by a sigmoid gate, and the output gate includes a sigmoid function and a tanh function.

[0068] In one embodiment of the present application, the network structure of the LSTM-based learner model includes three gate control structures, namely the forget gate, input gate, and output gate. The first step of LSTM is to determine which information will be discarded from the node state. This step is determined by the forget gate, which is controlled by the sigmoid gate. It will pass or partially pass the output of the previous moment. The forget gate outputs f t as follows:

[0069] f t =σ(W f ·[h t-1 ,x t ]+b f )

[0070] Among them, W f is the weight matrix between the hidden layer and the output layer, b f is the bias of the output layer, h t-1 is the transmission state at time t-1, that is, the hidden state, x t is the state of the transfer state cell at time t.

[0071] The output gate consists of two parts. One part uses the sigmoid function to decide which values ​​to update, denoted as i t ; The other part uses the tanh function to generate new candidate values, denoted as C t . They are output as follows:

[0072] i t =σ(W i ·[h t-1 ,x t ]+b i )

[0073] Ct =tanh(W c ·[h t-1 ,x t ]+b c )

[0074] Among them, σ represents the sigmoid function, W i and W c Both represent weight matrices, b i and b c Both represent offset.

[0075] The current cell state is input by the forget gate f t The product of the state at the previous moment plus the product of the two parts of the input gate is expressed as C t ,Right now:

[0076] C t =f t ·C t-1 +i t ·C t

[0077] Finally, the output gate stage runs a sigmoid gate to determine the output of the cell state, and the cell state is processed by tanh and multiplied by the sigmoid gate output to obtain the final output part h t .

[0078] o t =σ(W o ·[h t-1 ,x t ]+b o )

[0079] h t =o t ·tanh(c t )

[0080] Among them, y is the output of the sigmoid gate, W o represents the weight, b o is the bias of the sigmoid gate.

[0081] In one embodiment of the present application, the linear function expression of the SVM-based learner model is:

[0082]

[0083] Among them, α n and is the Lagrange multiplier, n is the number of support vectors, K(x n ,x) is the kernel function and b is the bias value.

[0084] In one embodiment of the present application, SVM is suitable for obtaining the global optimal value through regression prediction under limited samples, and has good generalization ability for unknown samples. The principle is to determine the following function expression by mapping the function into a high-dimensional space for linear regression:

[0085]

[0086] Among them, ω is the weight coefficient vector in the feature space, is the mapping from low dimension to high dimension, and b is the bias value.

[0087] In order to reduce the risk of overfitting and the impact of outliers, the insensitive function ε and the slack variable ξ are introduced. i and and regularization parameters, the regression problem is optimized as follows:

[0088]

[0089] st(b+W T X i )-y i ≤ε+ξ i ,i=1,2,...,n

[0090] y i -(b+W T X i )≤ε+ξ i ,i=1,2,...,n

[0091]

[0092] By introducing Lagrange multipliers and defining the Lagrange function, it is transformed into a dual problem that is equivalent to the original problem and easier to solve. At the same time, the kernel function is introduced to obtain the linear function as follows:

[0093]

[0094] Among them, α n and is the Lagrange multiplier, n is the number of support vectors, K(x n ,x) is the kernel function and b is the bias value.

[0095] In one embodiment of the present application, the objective function of the XGBoost based learner model is:

[0096]

[0097] Where i is each instance, is the predicted value, y iis the true value, l is the loss function, Ω(f k ) is the complexity of each decision tree.

[0098] In one embodiment of the present application, according to the cleaned data D={(x i ,y i )}(|D|=n,x i ∈R m ,y i =R), based on the CART algorithm to train the decision tree, the tree integration model is:

[0099]

[0100] Among them, F is the set space of decision trees, x i is the feature vector of the i-th data point, y i is the true value corresponding to each data point, f k For decision tree.

[0101] The objective function of the XGBoost base learner model is:

[0102]

[0103] Where i is each instance, is the predicted value, y i is the true value, l is the loss function, Ω(f k ) is the complexity of each decision tree. The first part is the error between the predicted value and the target true value, and the second part is the sum of the complexity of each tree, which is used to control the regularization term of the model complexity.

[0104] The objective function is trained using the incremental training method. Each time, a new function is added to the model while preserving the original model. The incremental function added in each round minimizes the objective function as much as possible. The process is as follows:

[0105]

[0106]

[0107]

[0108] …

[0109]

[0110] in, is the model prediction value of the ith sample in the tth round, Keep the model prediction value of round t-1 Then, add a new function f t (xi ) until the objective function reaches the desired target minimum.

[0111] In one embodiment of the present application, the foam lightweight soil performance prediction model is established based on the LSTM-based learner model, the SVM-based learner model and the XGBoost-based learner model, including:

[0112] S301: Divide the complete data set into a training set, a validation set and a test set.

[0113] S303: Input the training set and the test set into the LSTM-based learner model, the SVM-based learner model and the XGBoost-based learner model to obtain feature values ​​and prediction values.

[0114] S305: A linear regression LR model is used for training based on the characteristic values, and prediction is performed based on the predicted values ​​to obtain a foam lightweight soil performance prediction model.

[0115] In one embodiment of the present application, the complete data set is partitioned into a training set, a validation set, and a test set in a ratio of 3:1:2. Then, the training set is divided into five parts, and the training sets are named Train1, Train2, Train3, Train4, and Train5 respectively; fA1(X) represents the LSTM base learner model, fA2(X) represents the SVM base learner model, and fA3(X) represents the XGBoost base learner model; any one of the training sets is used as the validation set and the other four as the training set, and the model is trained using 5-fold cross validation, and prediction is performed on the test set; fA1(X), fA2(X ), fA3(X) respectively get 5 prediction values ​​on the training set, overlap and merge the 5 prediction values ​​vertically to get A1, A2, A3, and get prediction values ​​B1, B2, B3 on the test set; the prediction results A1, A2, A3 of the three trained base learners are used as the feature values ​​of the next layer of meta learners, and the linear regression LR model is used for training; the trained linear regression LR model is used to make predictions based on B1, B2 and B3, and the meta learner is constructed by linear regression as a prediction model for the performance of foam lightweight soil. By observing data from different data space angles and data structure angles, it has the advantage of strong model generalization ability.

[0116] S205: Using a complete data set to train the foam lightweight soil performance prediction model to obtain a performance prediction value, calculating the accuracy difference between the performance prediction value and the actual performance value, and optimizing the foam lightweight soil performance prediction model according to the accuracy difference.

[0117] In the embodiment of the present application, the foam lightweight soil performance prediction model established by training the complete data set after pretreatment is used, and different mix ratios, that is, different water-cement ratios and foam injection rates are input into the model to obtain performance prediction values, including fluidity, standard settlement rate and compressive strength. Afterwards, the precision difference i between the performance prediction value and the actual performance value is calculated, and the foam lightweight soil performance prediction model is optimized according to the precision difference i, wherein, when the precision difference i is greater than the preset standard precision difference, the complete data set v after pretreatment is re-modeled with the LSTM base learner model, the SVM base learner model, and the XGBoost base learner model, and the LSTM base learner model, the SVM base learner model, and the XGBoost base learner model are obtained, and the LSTM base learner model, the SVM base learner model, and the XGBoost base learner model are re-fused by the Stacking ensemble learning method to obtain a second foam lightweight soil performance prediction model, and the second foam lightweight soil performance prediction model is used to predict the performance of foam lightweight soil with different ratios.

[0118] S207: Input the formulation ratio to be predicted into the trained foam lightweight soil performance prediction model to obtain a predicted value of the foam lightweight soil performance.

[0119] In the embodiment of the present application, a foam lightweight soil performance prediction model is used to input different mix ratios, i.e., proportions of formula components, and output corresponding foam lightweight soil performance prediction values.

[0120] S209: constructing a multi-objective design optimization model for foam lightweight soil with cost and foam lightweight soil performance as objectives, and solving the multi-objective design optimization model for foam lightweight soil based on the predicted value of the foam lightweight soil performance by using a multi-objective optimization algorithm to obtain an intelligent design scheme for foam lightweight soil.

[0121] In the embodiment of the present application, a multi-objective design optimization model of foamed lightweight soil is constructed with cost and foamed lightweight soil performance (fluidity, standard settlement rate, compressive strength) as the target, aiming to achieve the balance and optimization of cost and performance to the greatest extent. Afterwards, a multi-objective optimization algorithm is used to solve the multi-objective design optimization model of foamed lightweight soil to obtain an intelligent design scheme of foamed lightweight soil.

[0122] In specific applications, the foamed lightweight soil required at a certain work site is used as a research case to form a multi-objective optimization framework and establish a multi-objective optimization model that comprehensively considers cost-performance (fluidity, standard settlement rate and compressive strength). First, the cost objective function of the foamed lightweight soil is established:

[0123]

[0124] Among them, c = 1, 2, 3, which are cement, water, and foaming agent, respectively. F0 is the average mixing and transportation cost of a cubic meter of foam lightweight soil, which usually decreases with the expansion of the production scale of foam lightweight soil. Here, it is taken as a fixed value F0 = 25. cost It is the production cost of one cubic meter of foam lightweight soil.

[0125] After that, the constraints of the foam lightweight soil mix design are determined:

[0126] Unilateral quality constraints: Wherein, CV is the total volume of each material component, and i=1, 2, 3 are the single dosages of cement, water, and foaming agent respectively.

[0127] Q l ≤Q≤Q u

[0128] R L ≤R≤R u

[0129] PS i,l ≤PS i

[0130] PPL i ≤PPL≤PPL u

[0131] PC i,l ≤PC i

[0132] Among them, Q = [q1, q2, q3], which are the single cubic meter dosage of cement, water and foaming agent respectively, R is the wet density of foamed lightweight soil, PS i is the compressive strength of foam lightweight soil on the i-th day, PPL is the fluidity of foam lightweight soil, PC i is the standard settlement rate of foam lightweight soil, Q l , R L , P.S. i,l , PL i,l , PC i,l is the lower limit of each value, Q u , R u , PPL u The upper limit allowed for each value.

[0133] In one embodiment of the present application, the adopting of a multi-objective optimization algorithm to solve the foam lightweight soil multi-objective design optimization model includes:

[0134] S401: Initialize the population particles and define the design parameters of the foam lightweight soil multi-objective optimization model represented by each particle.

[0135] S403: Calculate the fitness value related to the cost-performance optimization model based on the design parameters.

[0136] S405: Perform Pareto dominance update on the design parameters to construct a Pareto frontier.

[0137] S407: Update the speed and position of the design parameters according to the solution in the Pareto front.

[0138] S409: Determine the optimal solution for the multi-objective design of foam lightweight soil based on the solution set of the Pareto frontier.

[0139] In one embodiment of the present application, first, the population particles are initialized, and the positions and speeds of the population particles are randomly initialized in the solution space, that is, the design parameters of the lightweight soil multi-objective optimization model represented by each particle are defined; then, for each particle, based on its design parameters, the fitness value related to the cost-performance optimization model is calculated; then, each design parameter is Pareto dominance updated, including whether each parameter is dominated by other parameters in the population, if a solution is not dominated, the optimal solution of each parameter is updated, and all undominated solutions are constructed into a Pareto frontier; then, for each design parameter, its speed and position are updated according to the solutions in the individual and global (or local) Pareto frontiers, specifically, the speed update formula is:

[0140] v i (t+1)=w·v i (t)+c1·r1·(pbest i -x i (t))+c2·r2·(pbest-x i (t))

[0141] The position update formula is:

[0142] x i (t+1)=x i (t)+v i (t+1)

[0143] Among them, v i (t+1) is the velocity of particle i at time t+1, w is the inertia weight, which is used to control the particle velocity, v i (t) is the velocity of particle i at time t, c1 and c2 are learning factors (or acceleration constants), which control the particle to its individual historical optimal position pbest i and the weight of the global optimal position pbest, r1 and r2 are random numbers in the interval [0,1], pbest i is the historical optimal position of particle i, x i(t) is the position of particle i at time t, and pbest is the best position of all particles in the group.

[0144] Afterwards, check whether the termination condition is met, that is, the maximum number of iterations is reached or the quality of the solution is no longer significantly improved, then it is considered to have converged. If the termination condition is not met, repeat the above steps to continue iterating. Finally, analyze the solution set of the Pareto frontier and select the solution that best suits the multi-objective optimization model of foam concrete. After finding the Pareto optimal solution of the multi-objective optimization model, continue to sort the solutions to obtain a unique equilibrium solution in order to select the optimal solution. In one embodiment of the present application, a method based on a combination of combined weights and grey correlation analysis is selected to determine the weights of the multi-objective model. Specifically, first, construct a target matrix, expressed as:

[0145]

[0146] Among them, m is the number of Pareto optimal solutions, and n is the number of target problems.

[0147] Different indicators may have different units and value ranges. To avoid unfair and inaccurate comparison of solutions caused by these differences, the actual indicator values ​​need to be converted into a unified evaluation value. The calculation method of the evaluation value usually varies according to the type of indicator. The most commonly used evaluation methods are forward and reverse evaluation. In the target model, cost belongs to reverse and performance belongs to forward.

[0148] Positive

[0149] Reversal

[0150] As the evaluation criterion of the scheme, the correlation degree can be obtained by calculating the distance or similarity between the scheme index vector and the optimal scheme index vector. Then, the multiple schemes are sorted according to the size of the correlation degree. The larger the correlation degree, the better the scheme, and vice versa.

[0151] Positive: in, is the positive value of the jth feature, n is the number of training samples, ω i is the weight of the i-th training sample, δ is the relationship degree, x 0(j) is the 0th feature of the jth training sample, x i(j) is the jth feature of the i-th training sample.

[0152] In the above-mentioned data knowledge-driven intelligent design method for foam lightweight soil mix ratio, first, the performance data of foam lightweight soil is obtained and preprocessed to obtain a complete data set; then, an LSTM-based learner model, an SVM-based learner model and an XGBoost-based learner model are established respectively, and a foam lightweight soil performance prediction model is established based on the LSTM-based learner model, the SVM-based learner model and the XGBoost-based learner model; then, the foam lightweight soil performance prediction model is trained using the complete data set to obtain a performance prediction value, the accuracy difference between the performance prediction value and the actual performance value is calculated, and the foam lightweight soil performance prediction model is optimized according to the accuracy difference; then, the formula ratio to be predicted is input into the trained foam lightweight soil performance prediction model to obtain a foam lightweight soil performance prediction value; finally, a multi-objective design optimization model for foam lightweight soil is constructed, and a multi-objective optimization algorithm is used to solve the multi-objective design optimization model for foam lightweight soil to obtain an intelligent design scheme for foam lightweight soil. In other words, an intelligent design method for foam lightweight soil mix ratio is proposed. Through multi-module coordination, the structural design indicators are used to guide the mix ratio design of foam lightweight soil for widening roadbed, while meeting the design safety and economic requirements.

[0153] The following is a specific example to illustrate the specific implementation steps of a data knowledge driven foam lightweight soil mix intelligent design method of the present application. Figure 3 As shown, first, S501, the performance data of foam lightweight soil is obtained and preprocessed to obtain a complete data set. Specifically, S503, the water-cement ratio and foam injection rate are standardized, and the corresponding fluidity, standard settlement rate, and compressive strength are converted to normal distribution and reduced in dimension.

[0154] Afterwards, S505, respectively establish an LSTM-based learner model, an SVM-based learner model and an XGBoost-based learner model, and establish a foam lightweight soil performance prediction model based on the LSTM-based learner model, the SVM-based learner model and the XGBoost-based learner model. Among them, S507, the LSTM-based learner model includes a forget gate, an input gate and an output gate, wherein the forget gate is controlled by a sigmoid gate, and the output gate includes a sigmoid function and a tanh function. S509, the linear function expression of the SVM-based learner model is:

[0155]

[0156] Among them, α n and is the Lagrange multiplier, n is the number of support vectors, K(x n ,x) is the kernel function and b is the bias value.

[0157] S511, the objective function of the XGBoost base learner model is:

[0158]

[0159] Where i is each instance, is the predicted value, y i is the true value, l is the loss function, Ω(f k ) is the complexity of each decision tree.

[0160] S513-S517, divide the complete data set into a training set, a validation set and a test set; input the training set and the test set into the LSTM-based learner model, the SVM-based learner model and the XGBoost-based learner model to obtain characteristic values ​​and predicted values; train with a linear regression LR model based on the characteristic values, and make predictions based on the predicted values ​​to obtain a foam lightweight soil performance prediction model.

[0161] Then, S519, the foam lightweight soil performance prediction model is trained with a complete data set to obtain a performance prediction value, the accuracy difference between the performance prediction value and the actual performance value is calculated, and the foam lightweight soil performance prediction model is optimized according to the accuracy difference. Then, S521, the formula ratio to be predicted is input into the trained foam lightweight soil performance prediction model to obtain the foam lightweight soil performance prediction value. Finally, S523, a foam lightweight soil multi-objective design optimization model is constructed with cost and foam lightweight soil performance as the objectives, and a multi-objective optimization algorithm is used to solve the foam lightweight soil multi-objective design optimization model based on the foam lightweight soil performance prediction value to obtain a foam lightweight soil intelligent design solution. S525-S533, initialize population particles, define the design parameters of the foam lightweight soil multi-objective optimization model represented by each particle; calculate the fitness value related to the cost-performance optimization model based on the design parameters; perform Pareto dominance update on the design parameters and construct a Pareto frontier; update the speed and position of the design parameters according to the solution in the Pareto frontier; determine the optimal solution for the foam lightweight soil multi-objective design based on the solution set of the Pareto frontier.

[0162] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0163] Based on the same inventive concept, the embodiment of the present application also provides a data knowledge driven foam lightweight soil mix intelligent design device for realizing the data knowledge driven foam lightweight soil mix intelligent design method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more data knowledge driven foam lightweight soil mix intelligent design device embodiments provided below can refer to the limitations of a data knowledge driven foam lightweight soil mix intelligent design method above, and will not be repeated here.

[0164] In one embodiment, Figure 4 As shown, a data knowledge driven foam lightweight soil mix intelligent design device 400 is provided, comprising: a data set processing module 401, a foam lightweight soil performance prediction model establishment module 403, a foam lightweight soil performance prediction model training module 405, a foam lightweight soil performance prediction module 407 and a foam lightweight soil intelligent design scheme determination module 409, wherein:

[0165] The data set processing module 401 is used to obtain foam lightweight soil performance data and perform preprocessing to obtain a complete data set.

[0166] The foam lightweight soil performance prediction model establishment module 403 is used to establish an LSTM-based learner model, an SVM-based learner model and an XGBoost-based learner model respectively, and establish a foam lightweight soil performance prediction model based on the LSTM-based learner model, the SVM-based learner model and the XGBoost-based learner model.

[0167] The foam lightweight soil performance prediction model training module 405 is used to train the foam lightweight soil performance prediction model using a complete data set to obtain a performance prediction value, calculate the accuracy difference between the performance prediction value and the actual performance value, and optimize the foam lightweight soil performance prediction model according to the accuracy difference;

[0168] The foam lightweight soil performance prediction module 407 is used to input the formula ratio to be predicted into the trained foam lightweight soil performance prediction model to obtain the foam lightweight soil performance prediction value;

[0169] The foam lightweight soil intelligent design scheme determination module 409 is used to construct a foam lightweight soil multi-objective design optimization model with cost and foam lightweight soil performance as objectives, and adopt a multi-objective optimization algorithm to solve the foam lightweight soil multi-objective design optimization model based on the predicted value of the foam lightweight soil performance to obtain the foam lightweight soil intelligent design scheme.

[0170] In one embodiment of the present application, the data set processing module is further used to:

[0171] The water-cement ratio and foam injection rate are standardized, and the corresponding fluidity, standard settlement rate and compressive strength are transformed into normal distribution and reduced in dimension.

[0172] In one embodiment of the present application, the LSTM-based learner model includes a forget gate, an input gate, and an output gate, wherein the forget gate is controlled by a sigmoid gate, and the output gate includes a sigmoid function and a tanh function.

[0173] In one embodiment of the present application, the linear function expression of the SVM-based learner model is:

[0174]

[0175] Among them, α n and is the Lagrange multiplier, n is the number of support vectors, K(x n ,x) is the kernel function and b is the bias value.

[0176] In one embodiment of the present application, the objective function of the XGBoost based learner model is:

[0177]

[0178] Where i is each instance, is the predicted value, y i is the true value, l is the loss function, Ω(f k ) is the complexity of each decision tree.

[0179] In one embodiment of the present application, the foam lightweight soil performance prediction model establishment module is also used for:

[0180] Partitioning the complete data set to obtain a training set, a validation set, and a test set;

[0181] Input the training set and the test set into the LSTM base learner model, the SVM base learner model and the XGBoost base learner model to obtain feature values ​​and predicted values;

[0182] A linear regression LR model is used for training based on the characteristic values, and prediction is performed based on the predicted values ​​to obtain a foam lightweight soil performance prediction model.

[0183] In one embodiment of the present application, the foam lightweight soil intelligent design solution determination module is also used to:

[0184] Initialize the population particles and define the design parameters of the foam lightweight soil multi-objective optimization model represented by each particle;

[0185] Calculating a fitness value associated with a cost-performance optimization model based on the design parameters;

[0186] Performing Pareto dominance updates on the design parameters to construct a Pareto frontier;

[0187] updating the speed and position of the design parameters according to the solution in the Pareto front;

[0188] The optimal solution for the multi-objective design of foam lightweight soil is determined based on the solution set of the Pareto frontier.

[0189] Each module in the above-mentioned data knowledge-driven foam lightweight soil mix intelligent design device can be implemented in whole or in part by software, hardware and a combination thereof. Each of the above-mentioned modules can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to each of the above modules.

[0190] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Figure 5As shown. The computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, a data knowledge-driven intelligent design method for foam lightweight soil mix ratio is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covered on the display screen, or a key, trackball or touchpad set on the computer device shell, or an external keyboard, touchpad or mouse.

[0191] Those skilled in the art will understand that Figure 5 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0192] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0193] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0194] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0195] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0196] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

[0197] The technical features of the above embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0198] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A data knowledge-driven foam lightweight soil mix intelligent design method, characterized in that: The method comprises: Obtain foam lightweight soil performance data and preprocess it to obtain a complete data set; Establishing an LSTM-based learner model, an SVM-based learner model, and an XGBoost-based learner model respectively, and establishing a foam lightweight soil performance prediction model based on the LSTM-based learner model, the SVM-based learner model, and the XGBoost-based learner model; Using a complete data set to train the foam lightweight soil performance prediction model to obtain a performance prediction value, calculating the accuracy difference between the performance prediction value and the actual performance value, and optimizing the foam lightweight soil performance prediction model according to the accuracy difference; Input the formula ratio to be predicted into the trained foam lightweight soil performance prediction model to obtain the foam lightweight soil performance prediction value; A multi-objective design optimization model of foam lightweight soil is constructed with cost and foam lightweight soil performance as the objectives, and a multi-objective optimization algorithm is used to solve the multi-objective design optimization model of foam lightweight soil based on the predicted value of the foam lightweight soil performance to obtain an intelligent design scheme for foam lightweight soil.

2. According to claim 1, a data knowledge-driven foam lightweight soil mix intelligent design method is characterized in that: The pre-processing comprises: The water-cement ratio and foam injection rate are standardized, and the corresponding fluidity, standard settlement rate and compressive strength are transformed into normal distribution and reduced in dimension.

3. According to claim 1, a data knowledge driven foam lightweight soil mix intelligent design method is characterized in that: The LSTM-based learner model includes a forget gate, an input gate and an output gate, wherein the forget gate is controlled by a sigmoid gate, and the output gate includes a sigmoid function and a tanh function.

4. The data knowledge driven foam lightweight soil mix intelligent design method according to claim 1 is characterized in that: The linear function expression of the SVM-based learner model is: Among them, α n and is the Lagrange multiplier, n is the number of support vectors, K(x n ,x) is the kernel function and b is the bias value.

5. The data knowledge driven foam lightweight soil mix intelligent design method according to claim 1 is characterized in that: The objective function of the XGBoost base learner model is: Where i is each instance, is the predicted value, y i is the true value, l is the loss function, Ω(f k ) is the complexity of each decision tree.

6. The data knowledge driven foam lightweight soil mix intelligent design method according to claim 1 is characterized in that: The foam lightweight soil performance prediction model established based on the LSTM-based learner model, the SVM-based learner model and the XGBoost-based learner model includes: Partitioning the complete data set to obtain a training set, a validation set, and a test set; Input the training set and the test set into the LSTM base learner model, the SVM base learner model and the XGBoost base learner model to obtain feature values ​​and predicted values; A linear regression LR model is used for training based on the characteristic values, and prediction is performed based on the predicted values ​​to obtain a foam lightweight soil performance prediction model.

7. The data knowledge driven foam lightweight soil mix intelligent design method according to claim 1 is characterized in that: The adopting of a multi-objective optimization algorithm to solve the multi-objective design optimization model of foam lightweight soil includes: Initialize the population particles and define the design parameters of the foam lightweight soil multi-objective optimization model represented by each particle; Calculating a fitness value associated with a cost-performance optimization model based on the design parameters; Performing Pareto dominance updates on the design parameters to construct a Pareto frontier; updating the speed and position of the design parameters according to the solution in the Pareto front; The optimal solution for the multi-objective design of foam lightweight soil is determined based on the solution set of the Pareto frontier.

8. A data knowledge driven foam lightweight soil mix intelligent design device, characterized in that: The device comprises: The data set processing module is used to obtain the foam lightweight soil performance data and perform preprocessing to obtain a complete data set; A foam lightweight soil performance prediction model establishment module is used to establish an LSTM-based learner model, an SVM-based learner model, and an XGBoost-based learner model, respectively, and establish a foam lightweight soil performance prediction model based on the LSTM-based learner model, the SVM-based learner model, and the XGBoost-based learner model; A foam lightweight soil performance prediction model training module is used to train the foam lightweight soil performance prediction model using a complete data set to obtain a performance prediction value, calculate the accuracy difference between the performance prediction value and the actual performance value, and optimize the foam lightweight soil performance prediction model according to the accuracy difference; The foam lightweight soil performance prediction module is used to input the formula ratio to be predicted into the trained foam lightweight soil performance prediction model to obtain the foam lightweight soil performance prediction value; The foam lightweight soil intelligent design scheme determination module is used to construct a multi-objective design optimization model for foam lightweight soil, and adopts a multi-objective optimization algorithm to solve the multi-objective design optimization model for foam lightweight soil to obtain an intelligent design scheme for foam lightweight soil.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images