Intelligent prediction method and system for strength of jet grouting pile based on machine learning

Abstract

The invention relates to the technical field of engineering intelligent algorithms, in particular to an intelligent prediction method and system for the strength of a jet grouting pile based on machine learning. The method comprises the following steps: acquiring multi-dimensional data of design and construction of the jet grouting pile as an original sample data set; preprocessing the original sample data set to obtain a standardized data set; constructing a jet grouting pile strength prediction model of the jet grouting pile based on machine learning; optimizing the jet grouting pile strength prediction model according to the standardized data set to obtain an intelligent jet grouting pile strength prediction model; and predicting the strength of the jet grouting pile according to the intelligent prediction model of the strength of the jet grouting pile to obtain a prediction result of the strength of the jet grouting pile. According to the method, the standardized data set is constructed by obtaining the multi-dimensional data of design and construction of the jet grouting pile, and the intelligent prediction model of the strength of the jet grouting pile is constructed in combination with machine learning, so that the intelligent prediction of the strength of the jet grouting pile is realized, the quality of the jet grouting pile is improved, and technical reference is provided for engineering design and construction of the jet grouting pile.

Description

Technical Field

[0001] The present invention relates to the technical field of engineering intelligent algorithms, and in particular to a method and system for intelligent prediction of jet grouting pile strength based on machine learning. Background Art

[0002] High-pressure jet grouting pile technology is an efficient foundation reinforcement method, which is widely used in soft soil foundation treatment, building foundation reinforcement and tunnel construction control. High-pressure jet grouting pile technology injects cement slurry into the underground soil layer through high-pressure jet grouting, so that the soil and cement slurry are fully mixed to form a high-strength cement soil pile body. The main advantage is that the construction is flexible and can adapt to complex geological conditions. At the same time, the jet grouting pile has high compressive strength and shear strength, which effectively improves the bearing capacity and stability of the foundation. In addition, the jet grouting pile has little impact on the environment during construction, which is particularly suitable for densely urbanized areas and areas with dense underground facilities.

[0003] However, the quality and strength of jet grouting piles are affected by many factors, resulting in a large degree of discreteness and uncertainty in the strength and performance of jet grouting piles. The influencing factors include the jet grouting process itself, the setting of construction parameters, and on-site geological conditions. In previous technical solutions, the assessment of pile strength relied on on-site sampling and laboratory testing, which was not only time-consuming and labor-intensive, but also had high engineering costs. It was also difficult to achieve real-time prediction and feedback during the construction process. In addition, the strength prediction methods of jet grouting piles mostly use empirical formulas and simple statistical models, which are difficult to accurately describe the complex nonlinear relationship between construction parameters and geological conditions, thus affecting the accuracy and reliability of the prediction of the strength of jet grouting piles.

[0004] In view of this, there is an urgent need for a data-driven intelligent prediction method that can quickly and efficiently capture the influencing factors of jet grouting pile strength and provide engineering personnel with real-time and accurate jet grouting pile strength prediction results so as to optimize the construction plan in time and improve the project quality. Summary of the invention

[0005] In view of the defects in the prior art, the present invention provides a method and system for intelligent prediction of the strength of jet grouting piles based on machine learning.

[0006] In order to achieve the above-mentioned purpose, in the first aspect, the present invention provides a method for intelligent prediction of the strength of jet grouting piles based on machine learning, the method comprising the following steps: obtaining multidimensional data of the design and construction of jet grouting piles as an original sample data set; preprocessing the original sample data set to obtain a standardized data set; constructing a jet grouting pile strength prediction model of the jet grouting pile based on machine learning; optimizing the jet grouting pile strength prediction model according to the standardized data set to obtain an intelligent prediction model for the strength of the jet grouting pile; predicting the strength of the jet grouting pile according to the intelligent prediction model for the strength of the jet grouting pile to obtain a prediction result of the strength of the jet grouting pile. The present invention constructs a standardized data set by obtaining multidimensional data of the design and construction of jet grouting piles, effectively improves the data quality, constructs a jet grouting pile strength prediction model based on machine learning, and optimizes the standardized data set to obtain an intelligent prediction model for the strength of the jet grouting pile, thereby accurately predicting the strength of the jet grouting pile, which not only improves the prediction accuracy, but also realizes the real-time and intelligent nature of the prediction process, provides a scientific basis for the design and construction of the jet grouting pile project, and helps to improve the quality of the project.

[0007] Optionally, the multidimensional data for the design and construction of jet grouting piles is obtained as the original sample data set, including: obtaining the stratigraphic characteristics and geological parameters of the construction area where the jet grouting piles are located; obtaining the process scheme of the jet grouting piles based on the stratigraphic characteristics; collecting dynamic process parameters during the execution of the process scheme; and determining the strength of the jet grouting piles after the process scheme is completed. The present invention comprehensively considers the stratigraphic characteristics and geological parameters of the construction area of ​​the jet grouting piles, formulates a targeted process scheme, collects dynamic process parameters during the execution of the process scheme, and determines the strength of the jet grouting piles after the process scheme is completed, thereby ensuring the controllability of the construction quality and effect; it helps to optimize the construction design of the jet grouting piles, improve construction efficiency, and at the same time ensure the quality of the project, providing reliable data support for the prediction of the strength of the jet grouting piles.

[0008] Optionally, the preprocessing of the original sample data set to obtain a standardized data set includes: grouping the original sample data to obtain a data feature group; analyzing and processing the data of the data feature group to obtain a characteristic distribution of the data; and preprocessing the data according to the characteristic distribution of the data to obtain a standardized data set. The present invention clarifies the characteristic distribution of the data by grouping the original sample data and then analyzing and processing it, thereby performing targeted preprocessing on the data to obtain a standardized data set; this not only improves the data quality, but also enhances the comparability and usability of the data, provides data support for subsequent data analysis and model building, and helps to improve the accuracy and reliability of prediction results.

[0009] Optionally, preprocessing the data according to the characteristic distribution of the data to obtain a standardized data set includes:

[0010]

[0011] in, To normalize the data, is the original sample data, is the minimum value of the original sample data, The present invention preprocesses the data according to the characteristic distribution of the data to obtain a standardized data set, which significantly improves the consistency and comparability of the data by eliminating the dimensional differences and distribution differences between different features, and not only improves the training efficiency and prediction performance of the model, but also enhances the stability and reliability of the prediction results.

[0012] Optionally, the jet grouting pile strength prediction model based on machine learning includes: obtaining key non-temporal characteristics of the jet grouting pile strength; constructing a temporal characteristic variation relationship of the jet grouting pile strength; and combining the key non-temporal characteristics and the temporal characteristic variation relationship to perform modeling to obtain the jet grouting pile strength prediction model. The present invention obtains key non-temporal characteristics of the jet grouting pile strength, constructs a temporal characteristic variation relationship, and performs modeling by combining key non-temporal characteristics and temporal characteristic variation relationship, which significantly improves the accuracy and comprehensiveness of the jet grouting pile strength prediction model, fully considers the various factors that affect the strength and their variation patterns over time, helps to improve the prediction accuracy of the jet grouting pile strength prediction model, and provides scientific technical support for the control and optimization of the jet grouting pile construction quality.

[0013] Optionally, the key non-time-sequential features of the jet grouting pile strength are obtained, including:

[0014]

[0015] in, is the feature importance index, is the number of decision trees, is the traversal count of the decision tree, For the The out-of-bag error of the decision tree before adding random interference, For the The out-of-bag error of the decision tree after adding random interference. The present invention obtains the key non-time series features of the strength of the jet grouting pile through the random forest model, and can efficiently screen out the factors that have the greatest impact on the strength of the jet grouting pile from many features. The ensemble learning characteristics of the random forest enhance the stability and accuracy of the model, which helps to more accurately understand the main influencing factors of the strength of the jet grouting pile, and provides strong support for subsequent modeling and prediction.

[0016] Optionally, the jet grouting pile strength prediction model is optimized according to the standardized data set to obtain the jet grouting pile strength intelligent prediction model, including: splitting the standardized data set into a training data set and a validation data set; cross-validating the jet grouting pile strength prediction model according to the training data set and the validation data set to obtain a model evaluation result; adjusting the hyperparameters of the jet grouting pile strength prediction model based on the model evaluation result to obtain the jet grouting pile strength intelligent prediction model. The present invention splits the standardized data set into a training data set and a validation data set, and cross-validates the jet grouting pile strength prediction model, comprehensively evaluates the performance of the jet grouting pile strength prediction model, adjusts the hyperparameters based on the evaluation results to optimize the model, and obtains a more accurate jet grouting pile strength intelligent prediction model, thereby ensuring the stability and reliability of the model in practical applications and providing a powerful technical reference for the construction of jet grouting piles.

[0017] Optionally, cross-validating the jet grouting pile strength prediction model according to the training data set and the validation data set to obtain a model evaluation result includes:

[0018]

[0019] in, is the coefficient of determination, is the total number of recorded data, To record the traversal count of data, For the The predicted value of the recorded data, For the The measured value of the recorded data, for The present invention cross-validates the jet grouting pile strength prediction model through training data sets and validation data sets, and comprehensively and objectively evaluates the prediction performance and generalization ability of the jet grouting pile strength prediction model, which not only helps to discover potential overfitting or underfitting problems of the model, but also provides reliable data support for the optimization of model parameters, thereby ensuring that the ultimately constructed jet grouting pile strength intelligent prediction model has higher accuracy and reliability.

[0020] Optionally, the prediction of the strength of the jet grouting pile according to the jet grouting pile strength intelligent prediction model to obtain the jet grouting pile strength prediction result includes: combining the original sample data set and the jet grouting pile strength intelligent prediction model to construct a jet grouting construction plan; using the jet grouting pile strength intelligent prediction model to obtain the dynamic construction parameters in the execution process of the jet grouting construction plan in real time; and using the jet grouting pile strength intelligent prediction model to obtain the jet grouting pile strength prediction result based on the dynamic construction parameters. The present invention combines the original sample data set and the jet grouting pile strength intelligent prediction model to construct a jet grouting construction plan, which can accurately guide the construction, obtain the dynamic construction parameters in real time and use the jet grouting pile strength intelligent prediction model to predict the jet grouting pile strength, timely adjust the construction strategy, improve the construction efficiency and prediction accuracy, provide a scientific decision-making basis for the jet grouting pile construction, and ensure the engineering quality of the jet grouting pile.

[0021] In the second aspect, the present invention provides a system for intelligent prediction of the strength of jet grouting piles based on machine learning, the system executes the intelligent prediction method for the strength of jet grouting piles based on machine learning provided by the present invention, the system includes an input device, an output device, a processor and a memory, and its gain is: the hardware facilities integrated by the present invention have excellent performance, the input device, the output device, the processor and the memory are interconnected, the information transmission between the various components is smooth, and an efficient information processing system is constructed through the interaction of multiple hardware facilities. The present invention groups the original data set to ensure the targeted processing of non-sequential features and sequential features, based on the improvement of the long short-term memory network unit structure, the key non-sequential features obtained by the random forest model are embedded in the improved long short-term memory network unit structure for joint modeling, the correlation between the non-sequential features and the sequential features is strengthened, and the prediction accuracy of the strength of the jet grouting piles is significantly improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 A flow chart of a method for intelligent prediction of jet grouting pile strength based on machine learning according to an embodiment of the present invention;

[0023] Figure 2 A schematic diagram of random selection features of a random forest model according to an embodiment of the present invention;

[0024] Figure 3 This is a schematic diagram of the structure of a traditional LSTM unit in an embodiment of the present invention;

[0025] Figure 4 This is a schematic diagram of the improved LSTM unit structure of an embodiment of the present invention;

[0026] Figure 5 A schematic diagram of the overall structure of a bidirectional long short-term memory network model according to an embodiment of the present invention;

[0027] Figure 6This is a framework diagram of a system for intelligent prediction of jet grouting pile strength based on machine learning according to an embodiment of the present invention. DETAILED DESCRIPTION

[0028] The specific embodiments of the present invention will be described in detail below. It should be noted that the embodiments described herein are only for illustration and are not intended to limit the present invention. In the following description, a large number of specific details are set forth in order to provide a thorough understanding of the present invention. However, it is obvious to those of ordinary skill in the art that these specific details do not need to be adopted to implement the present invention. In other examples, in order to avoid confusing the present invention, known circuits, software or methods are not specifically described.

[0029] Throughout the specification, references to "one embodiment," "an embodiment," "an example," or "an example" mean that a particular feature, structure, or characteristic described in conjunction with the embodiment or example is included in at least one embodiment of the present invention. Therefore, the phrases "in one embodiment," "in an embodiment," "an example," or "an example" appearing in various places throughout the specification do not necessarily all refer to the same embodiment or example. In addition, particular features, structures, or characteristics may be combined in one or more embodiments or examples in any suitable combination and / or subcombination. In addition, it should be understood by those of ordinary skill in the art that the figures provided herein are for illustrative purposes and that the figures are not necessarily drawn to scale.

[0030] See also Figure 1 An embodiment of the present invention provides a method for intelligently predicting the strength of jet grouting piles based on machine learning, the method comprising the following steps:

[0031] S1. Obtain multi-dimensional data of jet grouting pile design and construction as the original sample data set.

[0032] Among them, S1 specifically includes the following steps:

[0033] S11, obtaining the stratigraphic characteristics and geological parameters of the construction area where the jet grouting pile is located.

[0034] Specifically, in this embodiment, before the design and construction of the jet grouting piles, a geological drill is used to obtain underground soil samples in the construction area where the jet grouting piles are located, and the underground soil samples are analyzed based on geological exploration technology to obtain the stratigraphic characteristics and geological parameters of the soil in the construction area.

[0035] Among them, the stratigraphic characteristics include the stratigraphic distribution and soil characteristics of the soil, the geological parameters include the physical indicators and mechanical properties of the soil, the physical indicators include natural gravity, water content, porosity, dry density, liquid limit and plasticity index, and the mechanical properties include standard penetration number, compression modulus and shear strength parameters.

[0036] S12. Obtaining a process scheme for the jet grouting piles based on the stratum characteristics.

[0037] Specifically, in this embodiment, in order to ensure the rationality of the process scheme, the process scheme of the rotary jet grouting piles is determined according to the formation characteristics and engineering requirements in combination with previous similar engineering schemes, and the preliminary process parameters of the process scheme are set.

[0038] Among them, the process scheme includes a rotary jet grouting pile forming process method, a rotary jet grouting pile design scheme, a mud ratio scheme and a process parameter scheme; the rotary jet grouting pile forming process method includes the selection of a single tube method, a double tube method or a triple tube method, and the setting of a nozzle diameter and a guide hole diameter; the rotary jet grouting pile design scheme includes a plan layout, the design of a pile number and a pile length, and whether there is repeated spraying or biting; the mud ratio scheme includes the design of a cement grade, a cement content and a water-cement ratio; the process parameter scheme includes the design of a current pile number, a current depth, a current time, grouting pressure, grouting flow, air pressure, air flow, water pressure, water flow, lifting step distance, a power head rotation speed and a lifting speed.

[0039] It should be noted that the process parameters of the process scheme directly affect the pile body forming process and the strength of the rotary jet pile. Therefore, the process parameters need to be set to obtain the preliminary process parameters, and the preliminary process parameters are adjusted in real time according to the actual construction situation of the process scheme.

[0040] S13, collecting dynamic process parameters during the execution of the process plan.

[0041] Specifically, in this embodiment, during the execution of the process scheme, the process parameters in the process parameter scheme are monitored and automatically collected in real time by sensors and monitoring equipment to obtain dynamic process parameters, and the dynamic process parameters are stored in a data recording system.

[0042] S14, measuring the strength of the jet grouting pile after the process scheme is completed.

[0043] Specifically, in this embodiment, after the execution of the process scheme is completed, the jet grouting piles are subjected to a maintenance treatment; after the maintenance treatment is completed, a cement soil core sample is obtained from the jet grouting piles using a geological drill.

[0044] The curing treatment time is 28 days, and the size, shape and continuity of the cement soil core sample must be representative.

[0045] Furthermore, the cement soil core sample is prepared into a cylindrical specimen, and the unconfined compressive strength of the cement soil core sample is tested by a loading device, and the unconfined compressive strength is used as the strength of the jet grouting pile.

[0046] Wherein, the diameter of the cylindrical sample is 50 mm and the height is 100 mm.

[0047] S2. Preprocess the original sample data set to obtain a standardized data set.

[0048] Among them, S2 specifically includes the following steps:

[0049] S21. Group the original sample data to obtain data feature groups.

[0050] In this embodiment, the original sample data is used to construct a complete data set based on the sampling frequency of the unconfined compressive strength along the burial depth.

[0051] Specifically, the original sample data set is divided into data feature groups according to the sampling frequency; the data feature groups include a fixed input feature group, a formation feature group and a dynamic time series feature group.

[0052] Among them, the fixed input feature group includes pile number, number of nozzles, hole diameter, cement grade, water-cement ratio and cement content; the formation feature group includes depth, porosity, dry density, plasticity index, liquid limit, standard penetration number and compression modulus; the dynamic timing feature group includes shotcrete pressure, shotcrete flow, air pressure, air flow, water pressure, water flow, lifting step, power head speed and lifting speed.

[0053] S22: Analyze and process the data of the data feature group to obtain characteristic distribution of the data.

[0054] Specifically, in this embodiment, first, the mean, standard deviation, maximum value, minimum value and quartile of the data of the data feature group are calculated as statistical indicators, and the data are statistically described and analyzed based on the statistical indicators; then, by drawing a histogram, a box plot and a density plot of the data, the distribution pattern and outliers of the data are visually observed using charts; then, a variable correlation analysis is performed on the data of the data feature group, and the input variables that have a greater impact on the output variable are identified by calculating the Pearson correlation coefficient, and the skewness and kurtosis of the data are calculated to evaluate whether the data has the characteristics of skewed distribution and peak distribution; finally, based on the Shapiro-Wilk test method, it is judged whether the data conforms to the normal distribution.

[0055] S23. Preprocess the data according to the characteristic distribution of the data to obtain a standardized data set.

[0056] Specifically, the data is cleaned by first using the box plot method to identify and eliminate potential outliers outside the quartile range; then, the Z-value method is used to further detect outliers, and the detected outliers are marked as outliers; in addition, in order to maintain the continuity of the data, the moving average method is used to reduce the impact of short-term fluctuations and make the data smoother and more coherent.

[0057] Furthermore, since the variable ranges of different data are quite different, the data are normalized to satisfy the following relationship:

[0058]

[0059] in, To normalize the data, is the original sample data, is the minimum value of the original sample data, is the maximum value of the original sample data.

[0060] S3. Constructing a jet grouting pile strength prediction model for the jet grouting pile based on machine learning.

[0061] In this embodiment, a rotary jet grouting pile strength prediction model is obtained based on joint modeling based on a random forest model and a bidirectional long short-term memory network model. The random forest model is used to analyze the non-time series characteristics of the original sample data, and the key non-time series characteristics that significantly affect the strength of the rotary jet grouting piles are screened out. The key non-time series characteristics are embedded as fixed inputs into the input layer of the bidirectional long short-term memory network model to provide additional contextual information for each time step to improve the prediction accuracy of the rotary jet grouting pile strength prediction model.

[0062] Among them, S3 specifically includes the following steps:

[0063] S31, obtaining key non-temporal characteristics of the strength of the jet grouting pile.

[0064] In this embodiment, a random forest model is used to process the non-time series features of the original sample, wherein the non-time series features include the fixed input feature group and the formation feature group, and the key non-time series features that have a significant impact on the strength of the jet grouting piles are screened out through feature importance analysis.

[0065] It should be noted that the main parameters of the random forest model include the number of trees, the maximum tree depth and the minimum number of sample splits. The main parameters are reasonably set to ensure accurate analysis of the non-time series features. The number of trees is set to 1000 to ensure the high stability and predictive ability of the random forest model; the maximum tree depth is set to 20 to prevent overfitting; the minimum number of sample splits is set to 5 to ensure that each node has enough samples.

[0066] Specifically, the random forest model uses a bootstrap method to generate multiple different training sets, and randomly selects non-time series features when each tree splits through a feature bagging method to capture the diversity of the non-time series features and improve the generalization ability of the random forest model.

[0067] For more details, see Figure 2 , the figure is a schematic diagram of the random selection feature of the random forest model, and the random selection feature includes the following steps: first, generate n samples through the random resampling method and train n decision trees, each tree is trained by randomly selecting features to form a forest structure; then, use the out-of-bag data to calculate the out-of-bag error of each feature; finally, calculate the feature importance index through the formula, and screen out the key non-time series features based on the feature importance index.

[0068] Furthermore, the feature importance index satisfies the following relationship:

[0069]

[0070] in, is the feature importance index, is the number of decision trees, is the traversal count of the decision tree, For the The out-of-bag error of the decision tree before adding random interference, For the Out-of-bag error of a decision tree after adding random interference.

[0071] S32, constructing a temporal characteristic variation relationship of the strength of the jet grouting pile.

[0072] In this embodiment, a bidirectional long short-term memory network model is used to process the dynamic time series data in the dynamic time series feature group; the bidirectional long short-term memory network model takes into account the previous and next information of the dynamic time series data, and can solve the gradient disappearance problem in long sequences through a gating mechanism.

[0073] Among them, the gating mechanism includes a forget gate, an input gate and an output gate; the output value range interval of the forget gate, input gate and output gate is [0,1], which determines the proportion of data passing through at each moment: the input gate controls the weight of information input, the forget gate controls the degree of information forgetting, and the output gate determines the proportion of output information; the gating mechanism is used to achieve precise control of input and output and effective protection of memory.

[0074] For details, see Figure 3 , which is a schematic diagram of a traditional LSTM unit structure. The main calculation process of the traditional LSTM unit structure includes:

[0075]

[0076]

[0077]

[0078]

[0079]

[0080]

[0081] in, is the activation value of the input gate, is the sigmoid activation function, is the weight matrix of the input gate, is the previous time step The hidden state of is the input of the memory unit, is the bias term of the input gate, is the candidate state value of the memory unit, is the hyperbolic tangent activation function, is the weight matrix of the candidate state values ​​of the memory unit, is the bias term of the candidate state value of the memory unit, is the activation value of the forget gate, is the weight matrix of the forget gate, is the bias term of the forget gate, is the current time step The output gate outputs, is the weight matrix of the output gate, is the bias term of the output gate, Current time step The state of the memory cell, is the previous time step The state of the memory cell, is the current time step The hidden state of is the activation value of the output gate.

[0082] More specifically, the sigmoid activation function and the hyperbolic tangent activation function satisfy the following relationship:

[0083]

[0084]

[0085] in, is the sigmoid activation function, is the hyperbolic tangent activation function, is the input of the activation function, is the base of natural logarithms.

[0086] For further information, see Figure 4 , the figure is a schematic diagram of the improved LSTM unit structure. By comparing with the traditional LSTM unit structure, the improved LSTM unit structure improves the activation value of the forget gate, the activation value of the output gate and the memory unit state, satisfying the following relationship:

[0087]

[0088]

[0089]

[0090] in, is the activation value of the forget gate, is the sigmoid activation function, is the weight matrix of the forget gate, is the previous time step The state of the memory cell, is the previous time step The hidden state of is the input of the memory unit, is the bias term of the forget gate, is the current time step The output gate outputs, is the weight matrix of the output gate, is the bias term of the output gate, Current time step The state of the memory cell, is the candidate state value of the memory unit.

[0091] In this embodiment, by Figure 3 and Figure 4 Compared with the traditional LSTM unit structure, two improvements are made.

[0092] First, hole connections are added to the forget gate and output gate of the traditional LSTM unit structure, enabling it to obtain the memory unit of the previous moment as input. This improvement enables the model to more flexibly select historical information that needs to be retained or forgotten.

[0093] Secondly, the traditional LSTM unit structure is improved to realize the linkage between the input gate and the forget gate. Since the collected data presents a time series characteristic, the forgotten and retained information are interdependent. The forgotten information is determined, and the remaining information is the part to be retained. Through this change, the forgotten state is only affected by the new input value, which optimizes the intuitiveness of the model logic and improves the overall performance.

[0094] The forgotten and retained information are interdependent. The forgotten information is determined, and the remaining information is the part to be retained, satisfying the following relationship:

[0095]

[0096] in, is the activation value of the input gate, is the activation value of the forget gate.

[0097] In this example, see Figure 5 , the figure is a schematic diagram of the overall structure of the bidirectional long short-term memory network model, and the bidirectional long short-term memory network model is constructed based on the improved LSTM unit structure.

[0098] Specifically, the overall structure of the bidirectional long short-term memory network model includes two bidirectional LSTM layers, each layer has 128 neurons, and the output of the first layer LSTM is used as the input of the second layer LSTM to gradually refine the time-dependent characteristics of the dynamic time series data to enhance the learning effect.

[0099] Among them, the bidirectional long short-term memory network model adopts a hyperbolic tangent activation function to capture complex relationships, and uses a regularization technique to prevent overfitting. The probability of the regularization technique is set to 0.3, and an Adam optimizer is used, and the learning rate of the optimizer is set to 0.001.

[0100] S33, combining the key non-time series characteristics with the time series characteristics change relationship to build a model to obtain the jet grouting pile strength prediction model.

[0101] In this embodiment, the jet grouting pile strength prediction model embeds non-time series features as fixed inputs, combines them with the time series feature processing results, and effectively integrates non-time series and time series features; the key non-time series features obtained based on the random forest model are embedded as fixed inputs into the bidirectional long short-term memory network model; so that in each time step, the input layer of the bidirectional long short-term memory network model contains the key non-time series features, providing stable global background information, so that the key non-time series features can be referenced in each time step of the jet grouting pile strength prediction model.

[0102] Specifically, in the bidirectional long short-term memory network model, dynamic time series features are gradually input into the bidirectional LSTM unit to capture the short-term and long-term time dependencies; at the same time, the key non-time series features are embedded as global features and interact with the dynamic time series features, so that the jet grouting pile strength prediction model can more effectively identify the multidimensional correlation between non-time series and time series data to obtain fusion features; the fusion features are dynamically weighted in the gating mechanism of the bidirectional long short-term memory network model, and the key non-time series features are propagated within the time step through the weight matrix, so that the jet grouting pile strength prediction model can comprehensively analyze the compound influence of different construction parameters, and based on the time series characteristics and non-time series characteristics of the jet grouting pile strength, the random forest model and the bidirectional long short-term memory network model are used for joint modeling to obtain the jet grouting pile strength prediction model.

[0103] S4. Optimize the jet grouting pile strength prediction model based on the standardized data set to obtain a jet grouting pile strength intelligent prediction model.

[0104] Wherein, S4 specifically includes the following steps:

[0105] S41, splitting the standardized data set into a training data set and a validation data set.

[0106] Specifically, in this embodiment, the cross-validation method is used to process the standardized data set; the standardized data set is randomly divided into five subsets of similar size to ensure that the data distribution in each subset remains similar; four of the subsets are used as training data sets for training, and the remaining subset is used as a verification data set for verification; in order to meet the needs of the rotary jet grouting pile strength prediction model, it is necessary to ensure that each subset contains a complete fixed input feature group, a formation feature group and a dynamic time series feature group during division to ensure the consistency and integrity of these three types of features in each subset, thereby supporting the effective training of the rotary jet grouting pile strength prediction model.

[0107] S42. Cross-validate the jet grouting pile strength prediction model according to the training data set and the validation data set to obtain a model evaluation result.

[0108] Specifically, the jet grouting pile strength prediction model is trained according to the training data set, and the jet grouting pile strength prediction model is verified according to the verification data set. This process is repeated five times. After each training, the mean absolute error, mean square error and determination coefficient on the verification set are calculated to evaluate the performance of the jet grouting pile strength prediction model on different data sets; finally, the average of the mean absolute error, mean square error and determination coefficient calculated five times is taken as the model evaluation result.

[0109] Furthermore, the mean absolute error, mean square error and determination coefficient satisfy the following relationship:

[0110]

[0111]

[0112]

[0113] in, is the mean absolute error, is the mean square error, is the coefficient of determination, is the total number of recorded data, To record the traversal count of data, For the The predicted value of the recorded data, For the The measured value of the recorded data, for The average value of the recorded data.

[0114] S43. Based on the model evaluation result, the hyperparameters of the jet grouting pile strength prediction model are adjusted to obtain an intelligent jet grouting pile strength prediction model.

[0115] Specifically, in this embodiment, the hyperparameters of the jet grouting pile strength prediction model are tuned by Bayesian optimization to minimize the error in the model evaluation result; Bayesian optimization predicts the optimal hyperparameter combination based on the constructed proxy function to gradually reach the optimal solution; Bayesian optimization is used to optimize and adjust the main parameters of the random forest model, the number of layers of the bidirectional long short-term memory network model, and the probability rate of regularization within a specified range. Bayesian optimization finds the optimal hyperparameter combination with fewer iterations, thereby constructing the jet grouting pile strength intelligent prediction model.

[0116] S5. Predicting the strength of the jet grouting piles according to the intelligent prediction model for the strength of the jet grouting piles obtains a prediction result for the strength of the jet grouting piles.

[0117] Among them, S5 specifically includes the following steps:

[0118] In this embodiment, the intelligent prediction model for jet grouting pile strength is applied to actual construction. By inputting jet grouting construction parameters and soil layer parameters, the unconfined compressive strength of the jet grouting pile body is obtained as the prediction result of the jet grouting pile strength.

[0119] S51. Construct a jet grouting construction plan by combining the original sample data set and the jet grouting pile strength intelligent prediction model.

[0120] Specifically, by utilizing the training data of previous projects and the geological survey data of the current construction area, combined with the prediction capability of the intelligent prediction model for the strength of jet grouting piles, a jet grouting construction plan for the new project is designed to provide a basis and guidance for on-site construction.

[0121] S52, using the intelligent prediction model for jet grouting pile strength to obtain in real time the dynamic construction parameters during the execution of the jet grouting construction scheme.

[0122] Specifically, during the execution of the jet grouting construction plan, the jet grouting pile strength intelligent prediction model is used to monitor and obtain the dynamic construction parameters in real time, and the dynamic construction parameters are stored in a memory.

[0123] S53, obtaining a prediction result of the strength of the jet grouting piles using the intelligent prediction model of the strength of the jet grouting piles based on the dynamic construction parameters.

[0124] Specifically, based on the dynamic construction parameters, the jet grouting pile strength intelligent prediction model is used to predict the jet grouting pile strength. According to the matching of the prediction results with the design indicators, the process parameters of the dynamic timing characteristics are adjusted in time to ensure that the construction quality meets the expected standards.

[0125] See also Figure 6 In an optional embodiment, in order to be able to efficiently execute a method for intelligent prediction of the strength of jet grouting piles based on machine learning provided by the present invention, the present invention provides a system for intelligent prediction of the strength of jet grouting piles based on machine learning, the system comprising an input device, an output device, a processor and a memory, the hardware facilities being interconnected, wherein the memory is used to store a computer program, the computer program comprising program instructions, the processor being configured to call the program instructions and execute the specific steps of the relevant embodiments of the method for intelligent prediction of the strength of jet grouting piles based on machine learning provided by the present invention. The intelligent prediction system for the strength of jet grouting piles based on machine learning provided by the present invention has a complete structure, is objective and stable, and improves the overall applicability and practical application capability of the present invention.

[0126] In summary, the method of the present invention provides a method and system for intelligent prediction of the strength of jet grouting piles based on machine learning, by obtaining the original data set of the jet grouting piles and preprocessing them, ensuring the targeted processing of non-sequential features and sequential features, based on the improvement of the long short-term memory network unit structure, the key non-sequential features obtained by the random forest model are embedded in the improved long short-term memory network unit structure and then jointly modeled, strengthening the correlation between non-sequential features and sequential data, realizing the intelligent prediction of the strength of the jet grouting piles, significantly improving the prediction accuracy of the strength of the jet grouting piles, providing a technical reference for the engineering design and construction of the jet grouting piles, and helping to improve the quality of the jet grouting piles. The method of the present invention is easy to understand, simple to calculate, small in workload, convenient for engineering application, and provides a theoretical basis and technical support for further development in the field of engineering intelligent algorithm technology.

[0127] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. These modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be included in the scope of the claims and specification of the present invention.

Claims

1. A method for intelligent prediction of jet grouting pile strength based on machine learning, characterized in that: The steps include: The multi-dimensional data of jet grouting pile design and construction are obtained as the original sample data set; Preprocessing the original sample data set to obtain a standardized data set; Constructing a jet grouting pile strength prediction model for the jet grouting pile based on machine learning; Optimizing the jet grouting pile strength prediction model according to the standardized data set to obtain a jet grouting pile strength intelligent prediction model; The strength of the jet grouting piles is predicted according to the intelligent prediction model for the strength of the jet grouting piles to obtain a prediction result for the strength of the jet grouting piles.

2. The method for intelligent prediction of jet grouting pile strength based on machine learning according to claim 1, characterized in that: The multi-dimensional data of jet grouting pile design and construction is obtained as the original sample data set, including: Obtaining the stratigraphic characteristics and geological parameters of the construction area where the jet grouting pile is located; Obtaining a process scheme for the jet grouting pile based on the stratum characteristics; Collecting dynamic process parameters during the execution of the process scheme; The strength of the jet grouting piles is measured after the process scheme is completed.

3. The intelligent prediction method for jet grouting pile strength based on machine learning according to claim 1 is characterized in that: The preprocessing of the original sample data set to obtain a standardized data set includes: Grouping the original sample data to obtain data feature groups; Analyzing and processing the data of the data feature group to obtain characteristic distribution of the data; The data are preprocessed according to the characteristic distribution of the data to obtain a standardized data set.

4. The intelligent prediction method for jet grouting pile strength based on machine learning according to claim 3 is characterized in that: The preprocessing of the data according to the characteristic distribution of the data to obtain a standardized data set includes: ； in, To normalize the data, is the original sample data, is the minimum value of the original sample data, is the maximum value of the original sample data.

5. The intelligent prediction method for jet grouting pile strength based on machine learning according to claim 1 is characterized in that: The jet grouting pile strength prediction model based on machine learning comprises: Obtaining key non-time-sequential characteristics of the jet grouting pile strength; Constructing a temporal characteristic variation relationship of the jet grouting pile strength; The jet grouting pile strength prediction model is obtained by combining the key non-time series characteristics with the time series characteristics change relationship to build a model.

6. The intelligent prediction method for jet grouting pile strength based on machine learning according to claim 5 is characterized in that: The key non-sequential features for obtaining the strength of the jet grouting pile include: ； in, is the feature importance index, is the number of decision trees, is the traversal count of the decision tree, For the The out-of-bag error of the decision tree before adding random interference, For the Out-of-bag error of a decision tree after adding random interference.

7. The intelligent prediction method for jet grouting pile strength based on machine learning according to claim 1 is characterized in that: The method of optimizing the jet grouting pile strength prediction model according to the standardized data set to obtain the jet grouting pile strength intelligent prediction model includes: Splitting the standardized data set into a training data set and a validation data set; Cross-validating the jet grouting pile strength prediction model according to the training data set and the validation data set to obtain a model evaluation result; Based on the model evaluation results, the hyperparameters of the jet grouting pile strength prediction model are adjusted to obtain an intelligent jet grouting pile strength prediction model.

8. The intelligent prediction method for jet grouting pile strength based on machine learning according to claim 7 is characterized in that: The cross-validation of the jet grouting pile strength prediction model according to the training data set and the validation data set to obtain a model evaluation result includes: ； in, is the coefficient of determination, is the total number of recorded data, To record the traversal count of data, For the The predicted value of the recorded data, For the The measured value of the recorded data, for The average value of the recorded data.

9. The method for intelligent prediction of jet grouting pile strength based on machine learning according to claim 1, characterized in that: The method of predicting the strength of the jet grouting pile according to the jet grouting pile strength intelligent prediction model to obtain a jet grouting pile strength prediction result includes: Combining the original sample data set with the intelligent prediction model for jet grouting pile strength to construct a jet grouting construction plan; The intelligent prediction model for jet grouting pile strength is used to obtain dynamic construction parameters in real time during the execution of the jet grouting construction scheme; The jet grouting pile strength prediction result is obtained based on the dynamic construction parameters using the jet grouting pile strength intelligent prediction model.

10. An intelligent prediction system for jet grouting pile strength based on machine learning, characterized in that: The system includes an input device, an output device, a processor and a memory, wherein the input device, the output device, the processor and the memory are interconnected, wherein the memory is used to store a computer program, the computer program includes program instructions, and the processor is configured to call the program instructions to execute the intelligent prediction method for the strength of jet grouting piles based on machine learning as described in any one of claims 1 to 9.

Images