Grouting material strength prediction method and system based on causal inference and machine learning

Abstract

The invention provides a grouting material strength prediction method and system based on causal inference and machine learning, and belongs to the technical field of grouting material performance prediction.The grouting material strength prediction method comprises the steps that key factor data having potential influences on the compressive strength of a grouting material are collected; the collected data are preprocessed; variables related to the strength performance of the grouting material are selected from the preprocessed data, and a causal structure diagram is preliminarily constructed; processing the preliminarily constructed causal structure diagram to obtain a complete causal network diagram, and obtaining main influence factors based on the complete causal network diagram; constructing a data set based on the main influence factors; training a deep neural network prediction model based on the data to obtain a trained deep neural network prediction model; feature data of a to-be-tested material is preprocessed, it is ensured that the data format is consistent with the model, the trained deep neural network prediction model is loaded for prediction, and a predicted value of compressive strength is generated.

Description

Technical Field

[0001] The present invention belongs to the technical field of performance prediction of grouting materials, and particularly relates to a method and system for predicting the strength of grouting materials based on causal inference and machine learning. Background Art

[0002] The statements in this part merely provide background technical information related to the present invention and do not necessarily constitute prior art.

[0003] With the rapid development of infrastructure construction and underground engineering technology, grouting materials are increasingly widely used in geotechnical engineering and tunnel construction. However, although existing methods for predicting the performance of grouting materials have gradually introduced advanced machine learning technologies such as deep learning, they still face some challenges. Although current deep learning models can learn through a large amount of data, they often lack effective feature selection and causal analysis mechanisms, especially in dealing with complex situations where multiple factors are intertwined. This makes it possible for the model to fail to automatically screen out the most influential variables when facing a large number of features, resulting in limitations in the accuracy and reliability of prediction.

[0004] In the prior art, many methods have used deep learning models (such as multi-layer perceptrons, convolutional neural networks, etc.) to predict the strength of grouting materials. Although they show strong capabilities in dealing with complex non-linear relationships and large-scale data. However, these methods often rely on data-driven and lack clear screening and analysis of the key causal factors affecting material performance. Therefore, simply relying on the method of model training may ignore some important variable relationships, resulting in poor generalization ability of the model. Especially when facing variable material and environmental conditions in engineering practical applications, the prediction effect may not reach the expected accuracy. Summary of the Invention

[0005] To overcome the deficiencies of the above prior art, the present invention provides a method for predicting the strength of grouting materials based on causal inference and machine learning, and finally realizes accurate prediction of the compressive strength performance of grouting materials.

[0006] To achieve the above object, one or more embodiments of the present invention provide the following technical solutions:

[0007] In the first aspect, a method for predicting the strength of grouting materials based on causal inference and machine learning is disclosed, including:

[0008] Collecting data on key factors that potentially affect the compressive strength of grouting materials;

[0009] Preprocessing the collected data;

[0010] Selecting variables related to the strength performance of grouting materials from the preprocessed data and initially constructing a causal structure diagram;

[0011] Process the preliminarily constructed causal structure diagram to obtain a complete causal network diagram, and obtain the main influencing factors based on the complete causal network diagram;

[0012] Construct a data set based on the main influencing factors;

[0013] Train a deep neural network prediction model based on the data to obtain a trained deep neural network prediction model;

[0014] Preprocess the characteristic data of the material to be tested to ensure that the data format is consistent with the model, and load the trained deep neural network prediction model for prediction to generate a predicted value of the compressive strength.

[0015] As a further technical solution, when preliminarily constructing the causal structure diagram, select variables related to the strength performance of the grouting material, including the water-cement ratio, the type and dosage of admixtures, the type of cement, and the characteristic factors of the material curing conditions, and assume the causal relationship between the variables to preliminarily construct the causal structure diagram.

[0016] As a further technical solution, process the preliminarily constructed causal structure diagram to obtain a complete causal network diagram. Specifically:

[0017] Given the data set X = {x 1 , x 2 , …, x n} that affects the strength performance of the grouting material, respectively initialize the factor candidate set M(x) = {} of the target node y and the main factor set M(y) = {};

[0018] Search for the node x i that has the greatest dependence on the y node, set a threshold m, and include M(x) in order from high to low according to the degree of dependence; remove most of the nodes that have no direct causal relationship with the target node y;

[0019] Remove the non-causal nodes in the candidate node set M(x) until all nodes are iterated;

[0020] Determine the causal direction between the candidate factor set and the target node y, iterate all nodes, and obtain a complete causal network diagram.

[0021] As a further technical solution, obtain the main influencing factors based on the complete causal network diagram, including: water-cement ratio, type of admixture, dosage of admixture, type of cement, and material curing conditions, and output these x types of factors and their corresponding causal effect intensities to display the relationship between variables in the form of an intuitive causal network diagram.

[0022] As a further technical solution, the deep neural network prediction model takes the obtained n main factors as inputs, where the number of neurons in the input layer is n, and the corresponding features are: water-cement ratio, type of admixture, dosage of admixture, type of cement, and material curing conditions. The number of neurons in the output layer is 1, representing the prediction target of the compressive strength of the grouting material.

[0023] As a further technical solution, when training the deep neural network prediction model based on data, it includes:

[0024] Set the hidden layer to m layers, and the number of neurons is i respectively;

[0025] Each layer uses the ReLU function as the activation function to introduce non-linearity into the network, enabling the model to learn complex non-linear relationships in the data;

[0026] Set Dropout to prevent the model from overfitting.

[0027] In the second aspect, a grouting material strength prediction system based on causal inference and machine learning is disclosed, including:

[0028] A data collection and preprocessing module, which is configured to: collect data on key factors that potentially affect the compressive strength of the grouting material;

[0029] Preprocess the collected data;

[0030] A causal network diagram construction module, which is configured to: select variables related to the strength performance of the grouting material from the preprocessed data and preliminarily construct a causal structure diagram;

[0031] Process the preliminarily constructed causal structure diagram to obtain a complete causal network diagram;

[0032] A data set construction module, which is configured to: obtain the main influencing factors based on the complete causal network diagram; construct a data set based on the main influencing factors;

[0033] A deep neural network prediction model training module, which is configured to: train the deep neural network prediction model based on the data to obtain a trained deep neural network prediction model;

[0034] A prediction module, which is configured to: preprocess the characteristic data of the material to be tested to ensure that the data format is consistent with the model, and load the trained deep neural network prediction model for prediction to generate a predicted value of the compressive strength.

[0035] The above one or more technical solutions have the following beneficial effects:

[0036] The technical solution of the present invention discloses a prediction method for the strength performance of grouting materials based on causal inference and machine learning. This method first collects and organizes the data of key factors affecting the compressive strength, and performs data cleaning and preprocessing. Then, a causal structure diagram is constructed, and the causal inference algorithm (MRCI) is used to evaluate the causal effect of variables on the strength performance, and the main influencing factors are screened out. Next, a deep neural network (DNN+Self-Attention) model is constructed based on the selected features, the model is trained and verified, and the hyperparameters are optimized, and finally the accurate prediction of the compressive strength performance of the grouting materials is realized.

[0037] The technical solution of the present invention is a method combining machine learning and causal inference analysis, which can fully explore the potential information in historical data while helping the model more accurately identify the key factors that have the greatest impact on the strength of grouting materials through the clear screening of causal relationships, thereby improving the accuracy and reliability of prediction.

[0038] Advantages of additional aspects of the present invention will be given in part in the following description, become apparent in part from the following description, or be learned by practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] The accompanying drawings forming a part of this specification are used to provide a further understanding of the present invention. The schematic embodiments and descriptions thereof of the present invention are used to explain the present invention and do not constitute an improper limitation of the present invention.

[0040] Figure 1 It is a flow chart of the overall framework of the technical solution of the embodiment of the present disclosure;

[0041] Figure 2 It is a causal network framework diagram of the technical solution of the embodiment of the present disclosure;

[0042] Figure 3 It is a prediction model diagram of the deep neural network (DNN) in the present disclosure. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0043] It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present invention belongs.

[0044] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present invention.

[0045] In the case of no conflict, the embodiments in the present invention and the features in the embodiments may be combined with each other.

[0046] Embodiment 1

[0047] See the appendix Figure 1 As shown, this embodiment discloses a grouting material strength prediction method based on causal inference and machine learning, including:

[0048] Step 1: Causal inference analysis to screen out the main features;

[0049] Step 2: Machine learning to predict the strength performance of grouting materials.

[0050] In this embodiment, regarding Step 1: Causal inference analysis to screen out the main features, it specifically includes the following process:

[0051] Step S1-1: First, according to engineering experience, relevant literature research, and preliminary test results, determine the key factors that potentially affect the compressive strength of grouting materials, such as: types of cement, water-cement ratio, types and dosages of admixtures, curing conditions, etc., and collect data.

[0052] The data source can be: collecting data from a series of tests designed in the laboratory; collecting actual data at the construction site; collecting data from completed engineering projects or literature.

[0053] Step S1-2: Preprocess the data collected in the previous step. Specifically: For outliers, use statistical methods for identification and processing. For example, for data with compressive strength significantly deviating from the reasonable range, analyze whether it is an abnormal value caused by test errors or other reasons, and decide whether to eliminate or correct it; for missing values, use interpolation methods or methods of deleting missing data to handle the null values in the data. At the same time, for the convenience of causal inference and machine learning model processing, normalize the data to ensure the consistency of different feature dimensions, specifically as follows:

[0054] Specifically as follows:

[0055]

[0056] Among them, s is the original feature value, s′ is the normalized feature value, min(s) and max(s) are the minimum and maximum values of this feature respectively.

[0057] Step S1-3: Combining laboratory tests and engineering practices, select the variables related to the strength performance of grouting materials from the preprocessed data, including feature factors such as water-cement ratio, types and dosages of admixtures, types of cement, and material curing conditions, and assume the causal relationships between the variables. On this basis, draw a causal structure diagram, connect the directly affected variables to represent the potential causal relationships, and initially construct a causal structure diagram, as shown in Figure 2 (a) in the figure.

[0058] For the relevant variables in the causal diagram, use the causal inference algorithm MRCI to evaluate the causal effect of each variable on the strength performance of the grouting material. The specific steps are as follows:

[0059] Step S1-4: Given the dataset X = {x 1 , x 2 , …, x n} that affects the strength performance of the grouting material, initialize the factor candidate set M(x) = {} and the main factor set M(y) = {} for the target node y respectively. Where the target is the compressive strength of the grouting material, and the factor candidate set M(x) includes all variables that may affect the compressive strength, such as water-cement ratio, type and dosage of admixture, type of cement, curing conditions, etc. Select the factors that have a direct causal relationship with the target node y (compressive strength) from the factor candidate set M(x), and finally construct the main factor set M(y).

[0060] Step S1-5: According to the causal strength, select the factors that have a direct causal relationship with the strength performance of the grouting material from the causal structure diagram as the main factors M(y). Use the maximum dependence - minimum redundancy criterion mRMR algorithm to incrementally search for the degree of dependence between data among the factors, seek the factor set with the maximum dependence on the survival time, then combine the ICCI algorithm to identify the direction of the causal variables, and finally obtain the complete causal network diagram. Specifically, incrementally search for the node x i that has the maximum dependence on the y node, set the threshold m, and include them in M(x) in descending order of the degree of dependence; remove most of the nodes that have no direct causal relationship with the target node y.

[0061] The specific form of the maximum dependence criterion is:

[0062]

[0063] In the formula, I(y, M) represents the dependence between the target variable y (compressive strength) and the candidate feature set M; I(x i , y) represents the mutual information between each feature x i (data feature affecting the strength performance of the grouting material) and the target variable y; I(x i , x j ) represents the mutual information between the feature x i and x j .

[0064] Step S1-6: Use the conditional independence test method to remove the non-causal nodes in the candidate node set M(x). For each variable x i in the candidate node set M(x), test its conditional independence from the target node through the above method. If x i is conditionally independent of the target node y, then x iIt has no causal effect on y and should be removed. Repeat the test until all nodes are iterated to ensure that the selected feature set M has the maximum dependence on the target variable y while minimizing the redundant information between features.

[0065] Step S1-7: Use the ICCI algorithm to determine the causal direction between the candidate factor set and the target node y. Iterate all nodes to obtain a complete causal network diagram, as shown in Figure 2 (b).

[0066] The specific form of the ICCI algorithm criterion is:

[0067] C x→y = P(y|x) - P(y)

[0068] In the formula, C x→y represents the causal effect strength from the feature factor x (the data feature affecting the strength performance of the grouting material) to the target variable y (compressive strength). If C x→y > 0, it is considered that the feature factor x has a causal relationship with the target variable y. On the contrary, if C x→y < 0, then there is a causal relationship between y and x; P(y|x) represents the probability of y given x; P(y) represents the marginal probability of the target variable y.

[0069] Causal feature screening: Determine the screening result according to the causal effect strength evaluated by causal inference and the direct influence relationship of the variable on the target feature (compressive strength). Output the screened variables and their corresponding causal effect strengths as the factor data set for the subsequent steps.

[0070] Step S1-8: Output the final result. The present disclosure finally obtains n main influencing factors: water-cement ratio, type of admixture, admixture dosage, type of cement, material curing conditions, etc. Output these n factors and their corresponding causal effect strengths, and visually display the possible direct or indirect causal relationships between variables through a causal diagram.

[0071] In this embodiment, regarding step two: machine learning prediction of the strength performance of the grouting material, it specifically includes:

[0072] Step S2-1: Based on the obtained influencing factors, screen and construct a new data set, which is the data set screened by causal inference, including n main influencing factors: water-cement ratio, type of admixture, admixture dosage, type of cement, material curing conditions, etc. Randomly shuffle the data set and divide it into a training set, a validation set, and a test set according to the ratio of 70%, 15%, and 15%. After the division, perform statistical analysis on the feature distributions of each data set to ensure that the difference in the mean μ and standard deviation σ between different data sets is less than 5%, meeting the similarity requirements.

[0073] The mean calculation formula is as follows:

[0074]

[0075] Among them, μ is the mean of the divided dataset, M is the number of samples in the divided dataset, and x i is the value of the i-th feature in the dataset.

[0076] The standard deviation calculation formula is as follows:

[0077]

[0078] Among them, σ is the standard deviation of the divided dataset, μ is the mean of the divided dataset, M is the number of samples in the divided dataset, and x i is the value of the i-th feature in the dataset.

[0079] Step S2-2: The present disclosure selects a deep neural network (DNN+Self-Attention) prediction model, embeds a self-attention mechanism in the hidden layer of the DNN model, enabling the model to automatically focus on the most important features, and uses the n main factors obtained in the first aspect as inputs. The number of neurons in the input layer is n, and the corresponding features are: water-cement ratio, type of admixture, admixture dosage, type of cement, material curing conditions, etc., which improves the accuracy of the model output results. The number of neurons in the output layer is 1, representing the prediction target of the compressive strength of the grouting material. See the attached Figure 3 shown.

[0080] Predict the compressive strength of the grouting material by training a deep neural network (DNN+Self-Attention). DNN has a multi-layer structure and high parameter flexibility. Its multi-layer non-linear mapping can effectively learn and express the high-order relationships between complex features. Based on the influencing factors screened by the causal analysis in the previous step, DNN can focus on the deep interactions of important features during the modeling process, thereby improving the prediction accuracy of the model. The self-attention mechanism dynamically adjusts the weight of each feature by calculating the correlation between each input feature and other features, so that the model pays more attention to the features that have a greater impact on the target variable (such as compressive strength).

[0081] Specifically, assume the input features are x 1 ,x 2 ,…,x n , and each x i represents a feature related to the grouting material (such as water-cement ratio, type and dosage of admixture, type of cement, etc.). The input data X is an n×d matrix, where: n is the number of samples, and d is the feature dimension (such as water-cement ratio, type of admixture, etc.).

[0082] In the self-attention mechanism, three vectors need to be generated from the input feature X first: query Q = XW q ; key K = XW k ; value V = XW v . Among them, W q , W k , W v are learned weight matrices, which represent mapping the input feature matrix X to the spaces of query, key, and value respectively. Q, K, and V represent the query, key, and value of each input feature respectively. Then, calculate the correlation between the query and the key to obtain the attention score of each feature:

[0083]

[0084] Among them, score(Q, K) is the correlation between the query and the key calculated by dot product, representing the importance of the feature; QK T is the dot product between the query and the key, and the result is an n×n matrix, representing the attention degree of each feature to other features; is a scaling factor to prevent the value of the dot product from being too large in the high-dimensional space, resulting in vanishing or exploding gradients.

[0085] Then, normalize the scores through the softmax function to obtain the attention weight of each feature:

[0086]

[0087] Softmax ensures that the weight of each feature is between 0 and 1, and the sum of all weights is 1, converting the scores into an importance matrix representing weights, so that the attention degree of each feature is quantified.

[0088] Then, perform weighted summation on the value vector V through the attention scores to obtain the final output of the self-attention mechanism:

[0089] Output = Attention(Q, K)V

[0090] Among them, Output is the final feature representation, which is the input feature after being weighted by the self-attention mechanism, representing the weighted expression of the mutual relationship between input features. This output will be fed into the subsequent layers of the DNN to help the model identify and weight the important input features that affect the target variable (compressive strength), and further used for predicting the compressive strength.

[0091] Step S2-3: By conducting a large number of network depth tests and K-fold cross-validation, and comprehensively considering factors such as training accuracy and time, in this disclosure, the hidden layer is set to m layers, namely hidden layers 1, 2, …, m, and the number of neurons in each layer is set to i. The dataset is divided into k subsets. Each time, k - 1 subsets are selected as the training set, and the remaining one subset is used as the validation set. By repeating training and validation, the stability and generalization ability of the model are ensured.

[0092] Step S2-4: The ReLU function is used as the activation function for each layer, introducing non-linearity to the network so that the model can learn complex non-linear relationships in the data.

[0093] Among them, the ReLU function is defined as:

[0094] ReLU(x) = max(0, x)

[0095] Among them, x is the input of the neuron. The ReLU function makes the output of all negative inputs 0, while the positive part is output as it is. This helps to avoid the problem of gradient disappearance and improve the training speed of the network.

[0096] Step S2-5: Dropout is set to prevent the model from overfitting. In this disclosure, the Dropout parameter is set to 0.5. In each training, each neuron has a 50% probability of being temporarily discarded, enabling the model to learn more robust feature representations during the training process, reducing the dependence on specific neurons or connections, and reducing the risk of overfitting.

[0097] Step S2-6: In this disclosure, the initial learning rate is set to 0.1, and a learning rate decay strategy of decaying by a fixed ratio is adopted. The learning rate is reduced to 0.8 times the original every 50 training epochs. The number of training times is set to 400 times. During the training process, every 10 training epochs, the model is tested on the validation set. By continuously adjusting the Dropout parameter, it is ensured that the loss function value on the validation set does not show an upward trend, and the performance of the model on the validation set gradually stabilizes.

[0098] After training is completed, the prepared test set data is input into the trained and optimized deep neural network (DNN+Self-Attention) model for predicting the compressive strength, and then the predicted value corresponding to each test sample is generated. Calculate the average of the squared differences between the predicted value and the actual value to obtain the mean squared error (MSE), and generate a scatter plot of the predicted value and the actual value to observe the overall distribution of the data and the fitting effect of the model. Check whether the scatter points are closely distributed near the ideal 45-degree diagonal line. If satisfied, it indicates that the model prediction is relatively accurate; otherwise, there is a prediction deviation, and retraining should be carried out.

[0099] The calculation formula for the mean squared error is as follows:

[0100]

[0101] Among them, y i is the actual value of the i-th sample, is the predicted value of the i-th sample, and M is the total number of data. MSE reflects the degree to which the prediction result deviates from the true value. The lower the value, the higher the prediction accuracy of the model.

[0102] According to the characteristics of the model, first determine the list of hyperparameters to be optimized (such as learning rate, number of hidden layers, number of neurons in each layer, regularization coefficient, activation function type, etc.). And based on the evaluation results of the validation set, adjust the hyperparameters, retrain and validate the model. If it is observed that the model is overfitting, the regularization term can be increased and the network complexity can be reduced; if it is observed that the model is underfitting, the number of nodes in the hidden layer can be increased and the learning rate can be increased, etc. Ensure that the model after hyperparameter optimization can achieve ideal results in practical applications.

[0103] Step S2-8: Apply the trained and optimized deep neural network model to the prediction of the strength performance of the grouting material. First, preprocess the characteristic data of the material to be tested to ensure that the data format is consistent with the model, and load the optimized model for prediction to generate the predicted value of the compressive strength. Then, based on the prediction results, generate a detailed performance prediction report to provide a reliable decision-making basis for practical engineering.

[0104] Since traditional machine learning models are often difficult to interpret due to their "black box" characteristics, and the solution described in the present invention combines causal inference analysis to screen and clarify direct causal effects, making the prediction results have a clear causal logic chain. In this way, it is possible to better understand which factors have an important effect on the strength, so as to more precisely control and optimize the material ratio in practical engineering, enhance the controllability of the solution, fully explore and utilize the potential information in historical data, and thus improve the accuracy and reliability of the prediction.

[0105] The solution described in the present invention screens out key influencing factors through causal network analysis in the early stage, effectively reducing the input dimension of the machine learning model and reducing the complexity of the model. This method reduces the unnecessary consumption of computing resources of the machine learning, improves the training speed and prediction efficiency of the model. When facing a large-scale data set, the optimized model is easier to process, thus improving the operating efficiency of the entire system.

[0106] Traditional prediction of the performance of grouting materials relies on a large number of experimental measurements, which is not only time-consuming and laborious, but also requires a large investment in cost. The present invention combines causal inference with a machine learning model to provide a fast and relatively accurate strength prediction in the material design and preparation stage, greatly reducing the need to rely on laboratory data verification.

[0107] Example Two

[0108] The purpose of this embodiment is to provide a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, the steps of the above method are implemented.

[0109] Example Three

[0110] The purpose of this embodiment is to provide a computer-readable storage medium.

[0111] A computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, the steps of the above method are executed.

[0112] Example Four

[0113] The purpose of this embodiment is to provide a grouting material strength prediction system based on causal inference and machine learning, including:

[0114] A data acquisition and preprocessing module, configured to: acquire key factor data that potentially affects the compressive strength of the grouting material;

[0115] Preprocess the acquired data;

[0116] A causal network diagram construction module, configured to: select variables related to the strength performance of the grouting material from the preprocessed data and preliminarily construct a causal structure diagram;

[0117] Process the preliminarily constructed causal structure diagram to obtain a complete causal network diagram;

[0118] A data set construction module, configured to: obtain the main influencing factors based on the complete causal network diagram; construct a data set based on the main influencing factors;

[0119] A deep neural network prediction model training module, configured to: train a deep neural network prediction model based on the data to obtain a trained deep neural network prediction model;

[0120] A prediction module, configured to: preprocess the characteristic data of the material to be measured to ensure that the data format is consistent with the model, and load the trained deep neural network prediction model for prediction to generate a predicted value of the compressive strength.

[0121] Example Five

[0122] The purpose of this embodiment is to provide a computer program product containing instructions, which, when running on a computer, enables the computer to execute the methods and functions involved in any one of the above embodiments.

[0123] The steps involved in the devices of the above embodiments correspond to those of Method Embodiment 1. For specific implementation details, please refer to the relevant description in Embodiment 1. The term "computer-readable storage medium" should be understood to include a single medium or multiple media containing one or more instruction sets; it should also be understood to include any medium that can store, encode, or carry an instruction set for execution by a processor and cause the processor to execute any method in the present invention.

[0124] Those skilled in the art should understand that the above-mentioned modules or steps of the present invention can be implemented using a general-purpose computer device. Optionally, they can be implemented using program codes executable by a computing device, so that they can be stored in a storage device for execution by the computing device, or they can be separately fabricated into individual integrated circuit modules, or multiple of them can be fabricated into a single integrated circuit module. The present invention is not limited to any specific combination of hardware and software.

[0125] Although the specific implementation of the present invention has been described above in conjunction with the accompanying drawings, it is not a limitation on the protection scope of the present invention. Those skilled in the art should understand that, based on the technical solutions of the present invention, various modifications or deformations that can be made without creative efforts by those skilled in the art are still within the protection scope of the present invention.

Claims

1. A grouting material strength prediction method based on causal inference and machine learning, characterized by: include: Collect data on key factors that have potential impact on the compressive strength of grouting materials; Preprocess the collected data; Select variables related to the strength performance of grouting materials from the preprocessed data and preliminarily construct a cause-effect structure diagram; The preliminary constructed causal structure diagram is processed to obtain a complete causal network diagram, and the main influencing factors are obtained based on the complete causal network diagram; Construct a data set based on the main influencing factors; Training a deep neural network prediction model based on the data to obtain a trained deep neural network prediction model; The characteristic data of the material to be tested is preprocessed to ensure that the data format is consistent with the model, and the trained deep neural network prediction model is loaded for prediction to generate the predicted value of compressive strength.

2. The grouting material strength prediction method based on causal inference and machine learning as claimed in claim 1 is characterized in that: When constructing the causal structure diagram, variables related to the strength performance of grouting materials are selected, including water-cement ratio, type and dosage of admixtures, type of cement, and characteristic factors of material curing conditions. The causal relationship between the variables is assumed to preliminarily construct the causal structure diagram.

3. The grouting material strength prediction method based on causal inference and machine learning as claimed in claim 1 is characterized in that: The preliminary constructed causal structure diagram is processed to obtain a complete causal network diagram, specifically: Given a data set X that affects the strength properties of grouting materials, n }, respectively initialize the factor candidate set M(x) = {} and the main factor set M(y) = {} of the target node y; Search for the node x that has the largest dependency on node y i , set the threshold m, and include M(x) in order from high to low according to the degree of dependence; remove most of the nodes that have no direct causal relationship with the target node y; Remove non-causal nodes from the candidate node set M(x) until all nodes are iterated; Determine the causal direction between the candidate factor set and the target node y, iterate all nodes, and obtain a complete causal network diagram.

4. The method for predicting the strength of grouting materials based on causal inference and machine learning as claimed in claim 1, characterized in that: Based on the complete causal network diagram, the main influencing factors are obtained, including water-cement ratio, admixture type, admixture dosage, cement type, and material curing conditions. These n types of factors and their corresponding causal effect intensity are output, and the relationship between the variables is displayed in the form of an intuitive causal network diagram.

5. The grouting material strength prediction method based on causal inference and machine learning as claimed in claim 1 is characterized in that: The deep neural network prediction model takes the obtained n main factors as input, where the number of neurons in the input layer is n, and the corresponding features are: water-cement ratio, admixture type, admixture dosage, cement type, and material curing conditions. The number of neurons in the output layer is 1, representing the prediction target of the compressive strength of the grouting material.

6. The grouting material strength prediction method based on causal inference and machine learning as claimed in claim 1, characterized in that: When training a deep neural network prediction model based on data, it includes: Set the hidden layer to m layers and the number of neurons to i; Each layer uses the ReLU function as the activation function to introduce nonlinearity into the network, so that the model can learn the complex nonlinear relationships in the data; Set Dropout to prevent the model from overfitting.

7. A grouting material strength prediction system based on causal inference and machine learning, characterized by: include: The data acquisition and preprocessing module is configured to: collect data of key factors that have a potential impact on the compressive strength of the grouting material; Preprocess the collected data; The causal network diagram construction module is configured to: select variables related to the strength performance of grouting materials from the preprocessed data and preliminarily construct a causal structure diagram; The preliminary constructed causal structure diagram is processed to obtain a complete causal network diagram; The data set construction module is configured to: obtain the main influencing factors based on the complete causal network diagram; construct the data set based on the main influencing factors; The deep neural network prediction model training module is configured to: train the deep neural network prediction model DNN+Self-Attention based on the data to obtain the trained deep neural network prediction model DNN+Self-Attention; The prediction module is configured to: pre-process the characteristic data of the material to be tested to ensure that the data format is consistent with the model, and load the trained deep neural network prediction model for prediction to generate a predicted value of compressive strength.

8. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of the method described in any one of claims 1 to 6 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method described in any one of claims 1 to 6 are performed.

Images