Construction control method and machine tool of pile forming machine tool for spirally replacing cement reinforced soil pile

Abstract

The invention provides a pile forming machine tool construction control method and machine tool for a spiral replacement cement reinforced soil pile, and relates to the technical field of long spiral replacement cement reinforced soil piles, the method comprises the following steps: adding a flow state cement reinforced soil composite material into a slurry stirring machine according to a preset proportion for mixing to obtain flow state cement reinforced soil; based on the pile position determined through surveying and setting out, drilling of a preformed hole is conducted through a spiral drill rod; flow-state cement reinforcing soil is poured into the preformed hole, and a spiral drill rod is lifted through a lifting system; the automatic control system dynamically adjusts the lifting speed of the spiral drill rod and the flow of the poured flow-state cement reinforced soil according to the monitored pressure numerical value of the pressure sensor; and after the spiral drill rod is lifted to the preset pile bottom elevation, pouring of the flow-state cement reinforcing soil is stopped, the drill rod is cleaned, and pile forming construction of the next pile position is conducted. According to the method, the lifting speed of the spiral drill rod and the flow of the poured flow-state cement reinforced soil can be better controlled in practical application.

Description

Technical Field

[0001] The invention relates to the technical field of long spiral replacement cement reinforced soil piles, and in particular to a construction control method and a machine for pile forming a spiral replacement cement reinforced soil pile. Background Art

[0002] Cement mixing piles are an effective form of soft foundation treatment. Cement is used as a curing agent. A mixer is used to spray cement into the soil and mix it thoroughly, so that a series of physical and chemical reactions occur between the cement and the soil, which hardens the soft soil and improves the foundation strength. Cement soil mixing piles are widely used in engineering construction due to their low cost, simple construction, and flexible reinforcement forms. In some highway projects, cement soil mixing piles account for more than 90% of the total engineering volume of various soft soil foundation reinforcement treatment measures.

[0003] However, traditional cement-soil mixing piles also have their own problems, such as uneven mixing of cement-soil, which leads to poor pile quality and low pile bearing capacity. It is found in engineering practice that the pile quality of cement-soil mixing piles is affected by the properties of the soil layer. For soft clay layers, the cement-soil mixing is uneven and the pile quality is poor. It is also found that single or combined measures such as increasing the mixing intensity, increasing the amount of curing agent material, and increasing the shotcrete pressure cannot effectively solve the problem of poor pile quality of cement-soil in soft clay layers and high-pressure water areas.

[0004] Currently, no effective solution has been proposed for the problems in the related technologies. Summary of the invention

[0005] In view of this, the present invention provides a construction control method and a pile forming machine for spiral replacement cement reinforced soil piles to solve the above-mentioned problems.

[0006] In order to solve the above problems, the specific technical solutions adopted by the present invention are as follows:

[0007] According to one aspect of the present invention, a method for controlling the construction of a pile forming machine for spiral replacement cement reinforced soil piles is provided, the method comprising the following steps:

[0008] S 1. Add the fluidized cement reinforced soil composite material into a slurry mixer according to a preset ratio for mixing to obtain fluidized cement reinforced soil;

[0009] S2, based on the pile position determined by surveying and setting out, and according to the pre-set pile bottom elevation, use the auger to drill the pre-formed hole;

[0010] S3, pouring fluid cement reinforcement soil into the pre-formed hole, and lifting the auger rod through the lifting system; the automatic control system dynamically adjusts the lifting speed of the auger rod and the flow rate of pouring fluid cement reinforcement soil according to the monitored pressure value of the pressure sensor;

[0011] S4. When the auger rod is lifted to the preset pile bottom elevation, stop pouring the fluid cement reinforcement soil, clean the drill rod and proceed with the next pile construction.

[0012] Preferably, the automatic control system dynamically adjusts the lifting speed of the auger rod and the flow rate of pouring fluid cement reinforced soil according to the monitored pressure value of the pressure sensor, including the following steps:

[0013] S31, constructing a data set using the acquired historical pressure values ​​of the pressure sensor, the historical lifting speed of the auger rod, the historical flow rate of the poured fluid cement reinforced soil, and the historical forming quality of the cement reinforced soil pile, and performing feature processing on the data set to obtain a feature set;

[0014] S32, using the t-SNE algorithm to reduce the dimension of the feature set, and obtain a low-dimensional data set through iterative solution;

[0015] S33, constructing and training a parameter control strategy model based on the obtained low-dimensional data set;

[0016] S34, using the monitored pressure value of the pressure sensor as the input value of the parameter control strategy model, and outputting the optimal values ​​of the lifting speed of the auger rod and the flow rate of pouring fluid cement reinforced soil through model analysis;

[0017] S35. Based on the optimal values ​​of the lifting speed of the auger rod and the flow rate of pouring the fluid cement reinforced soil, the automatic control system adjusts the lifting speed of the auger rod and the flow rate of pouring the fluid cement reinforced soil.

[0018] Preferably, using the t-SNE algorithm to reduce the dimension of the feature set and obtaining a low-dimensional data set by iterative solution includes the following steps:

[0019] S321, for each feature point in the feature set, calculating the joint probability between the feature point and other feature points, and constructing a joint probability distribution matrix in a high-dimensional space;

[0020] S322, randomly initialize a low-dimensional data set, and calculate a joint probability distribution matrix in the low-dimensional space;

[0021] S323. Calculate the KL divergence of the joint probability distribution in the high-dimensional space and the joint probability distribution in the low-dimensional space, take minimizing the KL divergence as the objective function, and iteratively solve the low-dimensional data set using the gradient descent method.

[0022] Preferably, for each feature point in the feature set, calculating the joint probability between the feature point and other feature points, and constructing a joint probability distribution matrix in a high-dimensional space comprises the following steps:

[0023] S3211. For each feature point in the feature set, calculate the Euclidean distance between the feature point and other feature points;

[0024] S3212, clustering the Euclidean distance using a K-means clustering algorithm, and assigning weights to the Euclidean distance according to the clustering result to obtain a weighted distance;

[0025] S3213. Using the weighted distance, calculate the joint probability between each feature point and other feature points in the high-dimensional space, and construct a joint probability distribution matrix in the high-dimensional space based on the joint probability.

[0026] Preferably, clustering the Euclidean distance using a K-means clustering algorithm, and assigning weights to the Euclidean distance according to the clustering results to obtain a weighted distance includes the following steps:

[0027] S32121. Randomly select K feature points as initial cluster centers;

[0028] S32122. According to the Euclidean distance between each feature point and other feature points, each feature point is assigned to the cluster where the centroid with the closest distance is located;

[0029] S32123. For each cluster, calculate the average Euclidean distance of all feature points belonging to the cluster, and use this average value as the new cluster center;

[0030] S32124, repeat steps S32122 and S32123 until the centroid no longer changes, and obtain a clustering result;

[0031] S32125. Based on the clustering results, the table parameter optimization method is used to assign weights to the Euclidean distance to obtain the weighted distance.

[0032] Preferably, the weighted distance is used to calculate the joint probability between each feature point and other feature points in the high-dimensional space:

[0033]

[0034] In the formula, p ij Represents the feature point x i and feature point x j The joint probability between j|i Represents the feature point x j As feature point x i Neighbor’s probability, p i|j Represents the feature point x i As feature point x j The probability of neighbors, N represents the total number of feature points, Represents the feature point x i and feature point x jThe weighted distance between i Represents the feature point x i The Gaussian variance centered at Represents the feature point x i and feature point x n The weighted distance between i|j Represents the feature point x i As feature point x j The probability of neighbors, Represents the feature point x j and feature point x i The weighted distance between Represents the feature point x j and feature point x n The weighted distance between j Represents the feature point x j The variance of a Gaussian centered at .

[0035] Preferably, constructing and training a parameter control strategy model according to the obtained low-dimensional data set includes the following steps:

[0036] S331, dividing the obtained low-dimensional data set into a training set and a test set;

[0037] S332, using the monitored pressure value of the pressure sensor as input, the lifting speed of the auger rod and the flow rate of the poured fluid cement reinforced soil as output, and using the training set to train the deep convolutional neural network to obtain a primary parameter control strategy model;

[0038] S333. Evaluate the parameter control strategy model using the test set. Based on the evaluation results, optimize and adjust the primary parameter control strategy model using the phased weighted ideal point method to obtain the final parameter control strategy model.

[0039] Preferably, the parameter control strategy model is evaluated using a test set, and based on the evaluation result, the primary parameter control strategy model is optimized and adjusted using the phased weighted ideal point method to obtain the final parameter control strategy model, which includes the following steps:

[0040] S3331. Based on the construction requirements of the pile-driving machine, a multi-objective function is established, wherein the multi-objective function includes the smoothness of the lifting speed, the accuracy of the flow control and the minimization of the construction cost;

[0041] S3332. Evaluate the performance of the primary parameter control strategy model using the test set and calculate the value of each objective function;

[0042] S3333, dividing the optimization process of the primary parameter control strategy model into several phases according to different phases of the pile driving machine construction; and setting a weight value for each objective function of each phase;

[0043] S3334. Using the staged weighted ideal point method, the multi-objective optimization problem is transformed into a single-objective optimization problem, and the single-objective optimization problem is solved by the particle swarm optimization algorithm to obtain the optimal model parameters for each stage;

[0044] S3335. Integrate the optimal model parameters of each phase to obtain the final parameter control strategy model.

[0045] Preferably, the multi-objective optimization problem is converted into a single-objective optimization problem by using the staged weighted ideal point method, and the single-objective optimization problem is solved by the particle swarm optimization algorithm to obtain the optimal model parameters for each stage, including the following steps:

[0046] S33341. For each objective function of each period, determine the ideal value based on historical data;

[0047] S33342. Construct a single objective optimization function according to the weight value of each period and the deviation between each objective function and the ideal value;

[0048] The expression of the single objective optimization function is:

[0049]

[0050] Where L(θ) represents a single objective optimization function, H represents the number of objective functions, and ω t represents the weight value of the tth objective function, F t (θ) represents the value of the tth objective function, F t * represents the ideal value of the tth objective function.

[0051] S33343. Randomly generate an initial solution for each target optimization function, where the initial solution represents model parameters;

[0052] S33344. Use the particle swarm optimization algorithm to iteratively solve the single objective optimization problem. When the preset number of iterations is reached, stop the iteration and output the optimal model parameters for each period.

[0053] According to another aspect of the present invention, a pile forming machine for spiral replacement cement reinforced soil piles is provided, the machine comprising an automatic control system arranged at the top of one side of a lifting system, a slurry mixer is arranged at one side of the automatic control system, an auger rod is arranged at the top of the other side of the lifting system, a grouting pipe connected to the slurry mixer is arranged inside the auger rod, and a pressure sensor is arranged at the bottom end of the auger rod;

[0054] A drum is arranged inside the slurry mixer, and gear teeth are arranged on the inner wall of the drum. The cross section of the gear teeth is a regular body structure, and a slurry spraying port is arranged between adjacent gear teeth.

[0055] The spiral drill rod is composed of a first drill rod and a second drill rod, and the first drill rod and the second drill rod are connected by bolts, the first drill rod is located at the top of the second drill rod, and the top of the first drill rod is connected to the lifting system;

[0056] The top ends of the first drill rod and the second drill rod are both provided with a plurality of fixing screw rods matched with the bolts, and the bottom ends of the first drill rod and the second drill rod are both provided with a plurality of screw holes matched with the fixing screw rods.

[0057] The beneficial effects of the present invention are:

[0058] 1. The present invention uses the t-SNE algorithm to perform dimensionality reduction processing on the feature set, reduces the dimension of the data, thereby reducing the computational complexity, and improving the training speed and response speed of the model. By calculating the joint probability between feature points and constructing a joint probability distribution matrix in a high-dimensional space, the relationship between feature points can be considered more comprehensively, thereby improving the robustness and generalization ability of the model. The K-means clustering algorithm is used to cluster the Euclidean distance, and the Euclidean distance is weighted according to the clustering result to obtain a weighted distance, which helps to highlight important features and suppress noise, thereby enhancing the accuracy of the model. By dividing the low-dimensional data set into a training set and a test set, the model is evaluated and optimized, ensuring the performance and stability of the model, and helping to better control the lifting speed of the auger rod and the flow rate of pouring fluid cement reinforced soil in practical applications.

[0059] 2. The present invention can comprehensively consider multiple key factors in the construction process of the pile-driving machine by establishing a multi-objective function including the smoothness of the lifting speed, the accuracy of the flow control and the minimization of the construction cost, thereby obtaining a more comprehensive and practical parameter control strategy. According to the different stages of the pile-driving machine construction, the optimization process is divided into several stages, and a weight value is set for each objective function of each stage. The dynamic changes in the construction process can be fully considered, so that the optimization strategy is closer to the actual construction needs. The complex multi-objective optimization problem is converted into a single-objective optimization problem by using the staged weighted ideal point method, which greatly simplifies the solution process. At the same time, the single-objective optimization problem is solved by the particle swarm optimization algorithm, and the optimal model parameters of each stage can be found efficiently and accurately.

[0060] 3. The present invention realizes precise control of the drilling rod lifting speed and the flow rate of fluid cement-solidified soil injection through an automatic control system, and the overall construction conditions of the pile body are kept consistent, thereby improving the uniformity and integrity of the spiral replacement cement-reinforced soil pile body.

[0061] 4. The present invention pre-mixes the fluidized cement-solidified soil with a slurry mixer, thereby ensuring the mixing uniformity of the solidified soil and solving the problems of uneven mixing and poor pile quality of traditional reinforced soil piles. The cement-solidified soil is brought into the best working state by pre-mixing, thereby improving the pile quality and the bearing capacity of a single pile.

[0062] 5. The present invention uses a long spiral drill rod to pre-drill holes, which can accurately control the pile diameter. The long spiral drill rod plays a wall protection role during the drilling process to prevent hole collapse. Compared with traditional cement soil mixing piles, the construction time can be reduced from the original 40min-50min to 10min-15min, the construction efficiency is improved by 2-3 times, and the construction progress is significantly accelerated.

[0063] 6. The present invention pre-forms a hole by drilling with a spiral drill rod. During the drilling process of the spiral drill rod, the original soil can be drilled out of the hole. The drilled soil can be used for the preparation of fluidized cement reinforced soil. The original soil can be used on-site without the need to transport the soil outside, which can effectively reduce the project cost and save costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. In the drawings:

[0065] Figure 1 is a flow chart of a method for controlling the construction of a pile forming machine for spiral replacement cement reinforced soil piles according to an embodiment of the present invention;

[0066] Figure 2 is a front view of a pile forming machine for spiral replacement cement reinforced soil piles according to an embodiment of the present invention;

[0067] Figure 3 is a front view of a first drill rod and a second drill rod in a spiral drill rod in a pile forming machine for spiral replacement cement reinforced soil piles according to an embodiment of the present invention;

[0068] Figure 4 1. It is a top view of a grouting machine in a pile forming machine for spiral replacement cement reinforced soil piles according to an embodiment of the present invention;

[0069] Figure 5 Schematic diagram of the structure of a cement reinforced soil pile body according to an embodiment of the present invention.

[0070] In the figure:

[0071] 1. Automatic control system; 2. Mud mixer; 21. Roller; 22. Gear teeth; 23. Spraying port; 3. Auger rod; 31. Fixed screw; 32. Screw hole; 4. Grouting pipe; 5. Pressure sensor. DETAILED DESCRIPTION

[0072] In order to enable those skilled in the art to better understand the technical solutions in this application, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this application.

[0073] According to an embodiment of the present invention, a construction control method and a pile forming machine for spiral replacement cement reinforced soil piles are provided.

[0074] The present invention is further described with reference to the accompanying drawings and specific embodiments. Figure 1 As shown, according to one embodiment of the present invention, a method for controlling the construction of a pile forming machine for spiral replacement cement reinforced soil piles is provided, the method comprising the following steps:

[0075] S 1. Add the fluidized cement reinforced soil composite material into the slurry mixer 2 according to a preset ratio for mixing to obtain fluidized cement reinforced soil;

[0076] Specifically, the fluidized cement reinforced soil composite material includes cement, slag powder, water, and natural foundation soil. In terms of parts, the fluidized cement reinforced soil includes 15-25 parts of cement, 5-10 parts of slag powder, 25-45 parts of water, and 100-140 parts of soil. Add the composite material to the slurry mixer according to the proportion, and control the mixing time to 3min-5min. Mix until the maximum particle size of the fluidized cement reinforced soil particles does not exceed 10mm, and the slump is 130-200mm.

[0077] S2, based on the pile position determined by measuring and setting out, and according to the pre-set pile bottom elevation, use the auger rod 3 to drill the pre-formed hole;

[0078] Specifically, after measuring and laying out the lines and determining the pile position, the auger is started to drill the hole. The drilling is stopped when the designed pile bottom elevation is reached to complete the pre-drilling.

[0079] S3, pouring fluid cement reinforcement soil into the pre-formed hole, and lifting the auger rod 3 through the lifting system; the automatic control system 1 dynamically adjusts the lifting speed of the auger rod 4 and the flow rate of pouring fluid cement reinforcement soil according to the monitoring pressure value of the pressure sensor 5;

[0080] As a preferred embodiment, the automatic control system 1 dynamically adjusts the lifting speed of the auger rod 3 and the flow rate of pouring fluid cement reinforced soil according to the monitored pressure value of the pressure sensor 5, including the following steps:

[0081] S31, constructing a data set using the obtained historical pressure values ​​of the pressure sensor 5, the historical lifting speed of the auger rod 3, the historical flow rate of the poured fluid cement reinforced soil, and the historical cement reinforced soil pile forming quality, and performing feature processing on the data set to obtain a feature set;

[0082] Specifically, the pressure value of the pressure sensor is an important indicator reflecting soil resistance, drill rod stress and injection pressure.

[0083] The lifting speed of the auger rod: reflects the progress of the construction process and the movement of the drill rod.

[0084] The flow rate of pouring fluidized cement reinforced soil is the key factor in controlling the quality and strength of reinforced soil.

[0085] Forming quality of cement-reinforced soil piles: This is a direct reflection of the final construction quality and may include parameters such as pile diameter, length, and strength.

[0086] The collected raw data contains noise, outliers or missing values, so it needs to be preprocessed. The preprocessing steps may include:

[0087] Cleaning data: removing or correcting erroneous, anomalous, or duplicate data points.

[0088] Fill missing values: Use interpolation, averaging, or other methods to fill in missing data.

[0089] Data normalization or standardization: converting data to the same scale for subsequent analysis and comparison.

[0090] After preprocessing, useful features need to be extracted from the data. The purpose of feature extraction is to convert raw data into numerical values ​​or vectors that can reflect the essential characteristics of the data. In this scenario, features may include:

[0091] Statistical characteristics of pressure values: such as mean, standard deviation, maximum value, minimum value, etc.

[0092] The changing characteristics of the lifting speed: such as the rate of change of speed, acceleration, etc.

[0093] Stability characteristics of flow: such as flow fluctuation range, average flow, etc.

[0094] Related characteristics of molding quality: such as the ratio of pile diameter to length, the relationship between strength and injection flow rate, etc.

[0095] S32, using the t-SNE algorithm to reduce the dimension of the feature set, and obtain a low-dimensional data set through iterative solution;

[0096] As a preferred implementation, using the t-SNE algorithm to reduce the dimension of the feature set, and obtaining a low-dimensional data set through iterative solution includes the following steps:

[0097] S321, for each feature point in the feature set, calculating the joint probability between the feature point and other feature points, and constructing a joint probability distribution matrix in a high-dimensional space;

[0098] As a preferred implementation, for each feature point in the feature set, calculating the joint probability between the feature point and other feature points, and constructing a joint probability distribution matrix in a high-dimensional space includes the following steps:

[0099] S3211. For each feature point in the feature set, calculate the Euclidean distance between the feature point and other feature points;

[0100] It should be noted that when calculating the Euclidean distance between each feature point in the feature set and other feature points, it is necessary to first clarify the data structure of the feature set. Usually, the feature set can be represented as a two-dimensional array (or matrix), where each row represents a feature point and each column represents a specific attribute or dimension of the feature point.

[0101] Assume that the feature set is X, which contains n feature points and each feature point has m dimensions. Then X can be represented as an n×m matrix.

[0102] For any two feature points x in the feature set i and x j (where i, j = 1, 2, ..., n and i ≠ j), the Euclidean distance d(x i ,x j ) can be calculated by the following formula:

[0103]

[0104] Among them, d ij Represents the feature point x i and x j The Euclidean distance between ik and x jk Represent the feature points x i and x j The value in the kth dimension.

[0105] S3212, clustering the Euclidean distance using a K-means clustering algorithm, and assigning weights to the Euclidean distance according to the clustering result to obtain a weighted distance;

[0106] As a preferred embodiment, the Euclidean distance is clustered using the K-means clustering algorithm, and weights are assigned to the Euclidean distance according to the clustering results to obtain the weighted distance, which includes the following steps:

[0107] S32121. Randomly select K feature points as the initial clustering centers;

[0108] It should be noted that in the K-means clustering algorithm, randomly selecting K feature points as the initial clustering centers is a key step. The purpose is to provide a starting point for the algorithm to start the iterative process.

[0109] S32122. According to the Euclidean distance between each feature point and other feature points, assign each feature point to the cluster where the nearest centroid is located;

[0110] Specifically, for each feature point x i , it is necessary to compare its Euclidean distance with all centroids and assign it to the cluster where the centroid with the smallest distance is located. That is, find the j that makes d ij the smallest, and then add x i to cluster j.

[0111] S32123. For each cluster, calculate the average Euclidean distance of all feature points belonging to the cluster and use this average value as the new clustering center;

[0112] S32124. Repeat steps S32122 and S32123 until the centroids no longer change to obtain the clustering result;

[0113] S32125. Based on the clustering result, use the table parameter optimization method to assign weights to the Euclidean distance to obtain the weighted distance.

[0114] For example, according to the clustering result, the Euclidean distance is divided into several categories, such as short distance, medium distance, and long distance. First, it is necessary to define two or more distance thresholds L1 and L2 to distinguish data points at different distance levels.

[0115] According to different distance categories, set different weights. Usually, the following weights can be set:

[0116] For short distance (d ij <L1), assign a lower weight w1<1;

[0117] For medium distance (L1≤d ij <L2), assign the standard weight w2 = 1;

[0118] For long distance (d ij≥L2), and assign a higher weight w3>1.

[0119] Next, the table parameter optimization method is used to determine the weight of each distance category. The specific steps are as follows:

[0120] Create a parameter table containing different weight combinations based on experiments or prior knowledge. For example, you can set up a table that lists the weight values ​​corresponding to different distance categories.

[0121] For each set of weight combinations, evaluate its performance indicators (such as classification accuracy, clustering quality, etc.) on the reduced-dimensional dataset. You can use methods such as cross-validation to evaluate performance.

[0122] According to the performance evaluation results, the best set of weight values ​​is selected as the final weight parameters.

[0123] For example, suppose that K-means clustering has been completed and the clustering results have been obtained. Now we need to assign weights to the Euclidean distances based on the clustering results. Specifically:

[0124] Determine the distance category and define three distance levels; among them, close distance: d ij <5, medium distance: 5≤d ij <10, long distance: d ij ≥10;

[0125] Set the weights for different distance levels; among them, close distance: w1 = 0.5, medium distance: w2 = 1.0, long distance: w3 = 1.5;

[0126] Create a parameter table listing different weight combinations;

[0127] Close distance weight w1: [0.5, 0.6, 0.7];

[0128] Medium distance weight w2: [1.0];

[0129] Long distance weight w3: [1.2, 1.3, 1.5];

[0130] For each weight combination, evaluate its performance indicators on the reduced-dimensional dataset. For example, the following performance indicators can be used for evaluation:

[0131] Classification accuracy: If the dataset has labels, the classification accuracy can be calculated.

[0132] Clustering quality: Clustering quality indicators such as the Silhouette coefficient can be used to evaluate the effect of clustering.

[0133] Suppose the following set of weights are chosen for evaluation:

[0134] The weight for close distance is w1=0.5; the weight for medium distance is w2=1.0; the weight for long distance is w3=1.5;

[0135] After evaluating the reduced-dimensionality dataset, it was found that this set of weights resulted in the best clustering effect and the highest Silhouette coefficient.

[0136] According to the evaluation results, the best performing set of weight values ​​is selected as the final weight parameter. In this example, w1=0.5, w2=1.0, w3=1.5 are selected as the optimal weight combination.

[0137] Once the optimal weight combination is determined, the weighted distance can be calculated. For example, for two data points x i and x j , and the weighted distance between them is calculated by multiplying them with the weight.

[0138] S3213. Using the weighted distance, calculate the joint probability between each feature point and other feature points in the high-dimensional space, and construct a joint probability distribution matrix in the high-dimensional space based on the joint probability.

[0139] It should be noted that the calculated joint probability values ​​are organized into an N×N matrix, where N is the total number of feature points. Each element p of the matrix ij Represents the feature point x i and feature point x j The joint probability between them. This matrix is ​​the joint probability distribution matrix in the high-dimensional space.

[0140] As a preferred implementation, the weighted distance is used to calculate the joint probability between each feature point and other feature points in the high-dimensional space:

[0141]

[0142] In the formula, p ij Represents the feature point x i and feature point x j The joint probability between j|i Represents the feature point x j As feature point x i Neighbor’s probability, p i|j Represents the feature point x i As feature point x j The probability of neighbors, N represents the total number of feature points, Represents the feature point x i and feature point x j The weighted distance between i Represents the feature point x i The Gaussian variance centered at Represents the feature point x i and feature point x n The weighted distance between i|j Represents the feature point x i As feature point x j The probability of neighbors, Represents the feature point x j and feature point x i The weighted distance between Represents the feature point x j and feature point x n The weighted distance between j Represents the feature point x j The variance of a Gaussian centered at .

[0143] S322, randomly initialize a low-dimensional data set, and calculate a joint probability distribution matrix in the low-dimensional space;

[0144] It should be noted that the dimension and size of the low-dimensional data set are first set, and a random number generator (such as uniform distribution or normal distribution) is used to generate a number of data points in each dimension.

[0145] For the joint probability distribution matrix in the low-dimensional space, it can be based on the calculation principle of the joint probability distribution matrix in the high-dimensional space mentioned above.

[0146] S323. Calculate the KL divergence of the joint probability distribution in the high-dimensional space and the joint probability distribution in the low-dimensional space, take minimizing the KL divergence as the objective function, and iteratively solve the low-dimensional data set using the gradient descent method.

[0147] It should be understood that KL divergence, full name Kullback-Leibler Divergence, is an important indicator to measure the difference between two probability distributions. It originated from information theory and was jointly proposed by Solomon Kullback and Richard Leibler in 1951. KL divergence is mainly used to describe the difference or distance between two probability distributions P and Q, although it is not a strict distance metric (because it does not satisfy all the properties of distance metrics, such as symmetry and triangle inequality).

[0148] It should be noted that the KL divergence formula is used to calculate the difference between the joint probability distribution in the high-dimensional space and the joint probability distribution in the low-dimensional space. The formula for KL divergence is:

[0149]

[0150] Where testKL(p, q) represents the KL divergence between the joint probability distribution in the high-dimensional space and the joint probability distribution in the low-dimensional space, pij represents the joint probability between feature points in the high-dimensional space, qij represents the joint probability between data points in the low-dimensional space, and i and j both represent indexes.

[0151] In addition, taking KL divergence minimization as the objective function, the gradient descent method is used to iteratively solve the low-dimensional data set, including the following steps:

[0152] Take the partial derivative of the KL divergence with respect to the coordinates of each data point in the low-dimensional data set to obtain a gradient vector, which indicates how to change the position of the low-dimensional data point to reduce the KL divergence;

[0153] Use the gradient descent method (or similar optimization algorithm) to update the coordinates of each data point in the low-dimensional dataset according to the calculated gradient vector, and repeat this process until a predetermined stopping condition is reached (such as a limit on the number of iterations, a change in KL divergence less than a threshold, etc.);

[0154] Output the optimized low-dimensional data set, which should retain the structural information in the high-dimensional data set as much as possible while reducing the KL divergence.

[0155] S33, constructing and training a parameter control strategy model based on the obtained low-dimensional data set;

[0156] As a preferred implementation, constructing and training a parameter control strategy model based on the obtained low-dimensional data set includes the following steps:

[0157] S331, dividing the obtained low-dimensional data set into a training set and a test set;

[0158] It should be noted that the obtained low-dimensional data set is divided into a training set and a test set. This is to evaluate the performance of the model on unknown data and prevent overfitting. Usually, the ratio can be adjusted according to specific needs. The common ratio is 70% for training and 30% for testing.

[0159] S332, using the monitored pressure value of the pressure sensor as input, the lifting speed of the auger rod and the flow rate of the poured fluid cement reinforced soil as output, and using the training set to train the deep convolutional neural network to obtain a primary parameter control strategy model;

[0160] Specifically, a deep convolutional neural network (CNN) is used to train the parameter control strategy model. In this step, the monitored pressure value of the pressure sensor 5 is used as the input feature, and the lifting speed of the auger rod 3 and the flow rate of the poured fluid cement reinforced soil are used as the output targets.

[0161] Use a deep learning framework (such as TensorFlow or PyTorch) to build a deep convolutional neural network. The specific architecture of the network (such as the number of layers, convolution kernel size, activation function, etc.) needs to be designed according to the specific problem and data characteristics.

[0162] Use the training set data to train the model. During the training process, you need to select a suitable loss function (such as mean square error) and optimization algorithm (such as Adam), and set an appropriate learning rate and number of training rounds.

[0163] S333. Evaluate the parameter control strategy model using the test set. Based on the evaluation results, optimize and adjust the primary parameter control strategy model using the phased weighted ideal point method to obtain the final parameter control strategy model.

[0164] The parameter control strategy model is evaluated using the test set. Based on the evaluation results, the primary parameter control strategy model is optimized and adjusted using the phased weighted ideal point method. The final parameter control strategy model includes the following steps:

[0165] S3331. Based on the construction requirements of the pile-driving machine, a multi-objective function is established, wherein the multi-objective function includes the smoothness of the lifting speed, the accuracy of the flow control and the minimization of the construction cost;

[0166] It should be noted that when establishing a multi-objective function based on the construction requirements of pile driving equipment, the main focus is on three core objectives: smoothness of speed improvement, accuracy of flow control, and minimization of construction costs. The following is a detailed description of the construction of these three objective functions:

[0167] The smoothness of the speed increase is crucial for the construction of pile driving equipment because it directly affects the construction quality and efficiency. In order to quantify this indicator, the sum of the squares of the speed change rate can be used as the objective function.

[0168] The accuracy of flow control is another key indicator in the construction of pile driving machinery. In order to quantify the accuracy of flow control, the square sum of the deviation between the actual flow and the set flow can be used as the objective function.

[0169] Construction cost is an important factor that needs to be considered in any engineering project. In the construction of pile-driving equipment, the cost mainly includes equipment depreciation, energy consumption, labor costs, etc. In order to simplify the problem, it can be assumed that the cost is proportional to the weight of the objective function, that is:

[0170] f3 = α·equipment depreciation + β·energy consumption + γ·labor cost;

[0171] Among them, α, β and γ are the weight coefficients of equipment depreciation, energy consumption and labor costs respectively. These coefficients can be adjusted according to actual conditions to reflect the importance of different cost factors.

[0172] S3332. Evaluate the performance of the primary parameter control strategy model using the test set and calculate the value of each objective function;

[0173] Specifically, the primary parameter control strategy model is applied to each scenario in the test set. In the process of running the model, the following data needs to be collected:

[0174] Improvement speed: record the improvement speed value given by the model at each time point.

[0175] Actual flow: records the actual flow value under model control.

[0176] Construction cost: Calculate the construction cost of each scenario based on the model's operation and the preset cost calculation formula.

[0177] After collecting the data, the value of each objective function can be calculated:

[0178] S3333, dividing the optimization process of the primary parameter control strategy model into several phases according to different phases of the pile driving machine construction; and setting a weight value for each objective function of each phase;

[0179] Specifically, according to the different stages of pile driving machine construction, the optimization process of the primary parameter control strategy model can be divided into several phases. Each phase may correspond to different links or key nodes in the construction process, such as the start phase, the intermediate stable phase and the end phase.

[0180] Next, it is necessary to set weights for each objective function of each phase, and these weights should reflect the importance and priority of each objective function at different construction stages. For example, at the beginning of construction, the smoothness of the lifting speed may be more important because it is necessary to ensure that the machine can start smoothly and avoid excessive impact; in the intermediate stable stage, the accuracy of flow control may become the main focus to ensure construction quality and efficiency; at the end of construction, minimizing construction costs may be more important because the overall economic benefits need to be considered.

[0181] Specifically, the process of setting weight values ​​can be based on experience or determined through data analysis or expert consultation. The following is a simple example of how to set weight values ​​for each objective function for each period:

[0182] (1) Initial stage

[0183] Smoothness of lifting speed: 0.6 (because smooth start is crucial for both machinery and construction);

[0184] Flow control accuracy: 0.3 (In the beginning, flow control is important but may not be the primary concern);

[0185] Minimization of construction costs: 0.1 (in the initial stage, cost may not be the main consideration);

[0186] (2) Intermediate stable stage

[0187] Improve speed smoothness: 0.3 (in the intermediate stage, speed smoothness is still important, but may not be the most important);

[0188] Flow control accuracy: 0.6 (in the intermediate stable stage, ensuring accurate flow control is crucial to construction quality);

[0189] Minimization of construction costs: 0.1 (although cost is important, it should be considered under the premise of ensuring quality and efficiency);

[0190] (3) Ending stage

[0191] Increase the smoothness of the speed: 0.2 (speed smoothness still needs attention in the end stage to avoid damage to the machine);

[0192] Flow control accuracy: 0.3 (flow control is still important at the end stage, but the requirements may be gradually relaxed);

[0193] Minimization of construction costs: 0.5 (In the final stage, cost considerations gradually become important to ensure overall economic benefits).

[0194] S3334. Using the staged weighted ideal point method, the multi-objective optimization problem is transformed into a single-objective optimization problem, and the single-objective optimization problem is solved by the particle swarm optimization algorithm to obtain the optimal model parameters for each stage;

[0195] As a preferred implementation, the multi-objective optimization problem is transformed into a single-objective optimization problem by using the staged weighted ideal point method, and the single-objective optimization problem is solved by the particle swarm optimization algorithm to obtain the optimal model parameters for each stage, including the following steps:

[0196] S33341. For each objective function of each period, determine the ideal value based on historical data;

[0197] It should be noted that for each objective function of each phase, an ideal value needs to be determined based on historical data or expert experience. This ideal value represents the optimal objective function value that is expected to be achieved under this phase. For example, for the smoothness of the speed increase, the ideal value may be zero (indicating that the speed is completely smooth); for the accuracy of flow control, the ideal value may be the set flow value (indicating that the actual flow is completely consistent with the set flow); for the construction cost, the ideal value may be the lowest historical cost or the budget limit.

[0198] S33342. Construct a single objective optimization function according to the weight value of each period and the deviation between each objective function and the ideal value;

[0199] The expression of the single objective optimization function is:

[0200]

[0201] Where L(θ) represents a single objective optimization function, H represents the number of objective functions, and ω t represents the weight value of the tth objective function, F t (θ) represents the value of the tth objective function, F t * represents the ideal value of the tth objective function.

[0202] S33343. Randomly generate an initial solution for each target optimization function, where the initial solution represents model parameters;

[0203] It should be noted that when constructing the multi-objective optimization problem, the problem has been transformed into a single-objective optimization problem through the phased weighted ideal point method. Next, in order to solve this single-objective optimization problem, it is necessary to randomly generate initial solutions for each objective optimization function (in fact, in this transformed framework, it has been integrated into a single-objective optimization function). These initial solutions represent the starting point of the model parameters, and the optimization algorithm will start searching for the optimal solution from this starting point.

[0204] S33344. Use the particle swarm optimization algorithm to iteratively solve the single objective optimization problem. When the preset number of iterations is reached, stop the iteration and output the optimal model parameters for each period.

[0205] It should be understood that Particle Swarm Optimization (PSO) is an optimization algorithm based on swarm intelligence, which simulates the behavior of individuals in a group of birds, fish, etc., collaborating with each other to find food. When PSO is applied to a single-objective optimization problem, each possible solution is regarded as a "particle", and the positions and velocities of these particles are updated through an iterative process to find the optimal solution.

[0206] S3335. Integrate the optimal model parameters of each phase to obtain the final parameter control strategy model.

[0207] It should be noted that integrating the optimal model parameters of each phase to obtain the final parameter control strategy model includes the following steps:

[0208] The optimal parameters for each phase are organized into a table or array, including parameter name, phase number, optimal value and other information.

[0209] Use charts (such as line charts and scatter plots) to show the changes in each parameter in different stages and observe whether there are obvious trends or correlations between the parameters.

[0210] If there are many parameters, calculate the correlation coefficients between the parameters to quantify the relationship between them, and determine which parameters may need to be adjusted in coordination based on the size of the correlation coefficients.

[0211] The optimal model parameters of each stage are integrated using linear interpolation method to obtain the final parameter control strategy model.

[0212] S34, using the monitored pressure value of the pressure sensor 5 as the input value of the parameter control strategy model, and outputting the optimal values ​​of the lifting speed of the auger rod 3 and the flow rate of pouring fluid cement reinforced soil through model analysis;

[0213] It should be noted that the pressure sensor 5 monitors and provides pressure values ​​in real time. These values ​​are the input values ​​of the parameter control strategy model, which are used for the analysis and prediction of the model. The parameter control strategy model is a trained deep learning model, which can analyze and output the optimal values ​​of the lifting speed of the auger rod 3 and the flow rate of the poured fluid cement reinforced soil according to the input pressure values. These optimal values ​​are generated by the model based on the learned data characteristics and rules, aiming to achieve the optimal control effect.

[0214] S35. Based on the optimal values ​​of the lifting speed of the auger rod 3 and the flow rate of pouring the fluidized cement reinforced soil, the automatic control system 1 adjusts the lifting speed of the auger rod 3 and the flow rate of pouring the fluidized cement reinforced soil.

[0215] It should be noted that the automatic control system 1 is the key part responsible for executing the optimal value. When the parameter control strategy model outputs the optimal value, the automatic control system 1 will receive these optimal values ​​and adjust the lifting speed of the auger rod 3 and the flow rate of the fluid cement reinforced soil according to the optimal value. In this way, the control parameters can be dynamically adjusted according to the real-time monitored pressure value to achieve the best working effect.

[0216] S4. When the auger rod is lifted to the preset pile bottom elevation, stop pouring the fluid cement reinforcement soil, clean the drill rod and proceed with the next pile construction.

[0217] Specifically, when the auger rod 3 is lifted to 0.5 m above the preset pile bottom elevation, the pouring of fluid cement reinforcement soil is stopped.

[0218] like Figure 2-4 As shown, according to another embodiment of the present invention, a pile forming machine for spiral replacement cement reinforced soil piles is provided, the machine comprising an automatic control system 1 arranged at the top of one side of a lifting system, a slurry mixer 2 is arranged on one side of the automatic control system 1, an auger rod 3 is arranged at the top of the other side of the lifting system, a grouting pipe 4 connected to the slurry mixer 2 is arranged inside the auger rod 3, and a pressure sensor 5 is arranged at the bottom end of the auger rod 3;

[0219] It should be noted that the lifting system is used for lifting and driving the spiral drilling, and includes a mast column, a hydraulic power unit, an electrical control system, etc. The composition of the lifting system is existing technology and will not be elaborated in detail here.

[0220] The automatic control system 1 can control the drilling and lifting speed of the spiral drill rod and the flow rate of pouring fluid cement reinforced soil in real time according to the pressure data monitored by the pressure sensor 5.

[0221] The grouting pipe 4 is located inside the auger rod 3 and is connected to the slurry mixer 2. The material of the grouting pipe 4 can be steel or a reinforced hose with higher strength, and the strength requirement is not less than 10 MPa.

[0222] A pressure sensor 5 is installed at the bottom of the auger rod. The pressure sensor 5 can monitor the pressure between the bottom of the auger rod 3 and the poured cement soil in real time, and transmit the pressure value back to the automatic control system 1. The automatic control system 1 automatically controls the speed of drilling rod lifting and the flow rate of poured solidified soil according to the pressure value returned by the pressure sensor 5.

[0223] A drum 21 is provided in the slurry mixer 2, and gear teeth 22 are provided on the inner wall of the drum 21. The cross section of the gear teeth 22 is a regular body structure, and a slurry spraying port 23 is provided between adjacent gear teeth 22.

[0224] Specifically, the gear teeth 22 are evenly arranged on the roller 21, and the evenly distributed gear teeth 22 are staggered with each other. The soil can be quickly crushed by the cutting and biting action of the gear teeth 22. The size of the grouting port 23 is 2mm-5mm, and the grouting pressure of the grouting port 23 is 2MPa-7MPa. The soil is further crushed by the punching action of the high-pressure grouting, making the cement soil mixing more uniform.

[0225] The spiral drill rod 3 is composed of a first drill rod and a second drill rod, and the first drill rod and the second drill rod are connected by bolts. The first drill rod is located at the top of the second drill rod, and the top of the first drill rod is connected to the lifting system; the diameter of the first drill rod is 700mm-1000mm, and the diameter of the second drill rod is 500mm-700mm (excluding 700mm); the tops of the first drill rod and the second drill rod are both provided with a plurality of fixing screws 31 that cooperate with the bolts, and the bottoms of the first drill rod and the second drill rod are both provided with a plurality of screw holes 32 that cooperate with the fixing screws 31.

[0226] Specifically, the spiral drill rod 3 can be assembled in sections, and the sections are connected by bolts. There are screw holes 32 and fixed screws 31 between the stages. After docking, the bolts can be tightened and fixed. The spiral drill rod 3 is divided into two types according to different sizes, a first drill rod (coarse drill rod) with a diameter of 700mm-1000mm, and a second drill rod (thin drill rod) with a diameter of 500mm-700mm (excluding 700mm). The segment length of the drill rod is 0.5m or 1m. Through the segment assembly of drill rods of different sizes, cement reinforced soil piles of different sizes and "upper thick and lower thin" variable cross-section cement reinforced soil piles can be formed. Thick and thin rods of different diameters and lengths can be arbitrarily combined, and spiral drill rods 3 of different cross-sectional sizes and lengths are assembled through segments, so that piles of different sizes can be formed.

[0227] like Figure 5 As shown, according to another embodiment of the present invention, there is provided a cast-in-place reinforced soil pile, including a pile body, which is prepared by using a construction method of a pre-cast-in-place reinforced soil pile forming machine according to the above embodiment.

[0228] In summary, with the help of the above-mentioned technical scheme of the present invention, the present invention uses the t-SNE algorithm to perform dimensionality reduction processing on the feature set, reduces the dimension of the data, thereby reducing the calculation complexity, and improving the training speed and response speed of the model. By calculating the joint probability between the feature points and constructing a joint probability distribution matrix in a high-dimensional space, it is possible to more comprehensively consider the relationship between the feature points, thereby improving the robustness and generalization ability of the model. The K-means clustering algorithm is used to cluster the Euclidean distance, and the Euclidean distance is weighted according to the clustering results to obtain a weighted distance, which helps to highlight important features and suppress noise, thereby enhancing the accuracy of the model. By dividing the low-dimensional data set into a training set and a test set, the model is evaluated and optimized, ensuring the performance and stability of the model, which helps to better control the lifting speed of the auger rod and the flow rate of the poured fluid cement reinforced soil in practical applications. The present invention can comprehensively consider multiple key factors in the construction process of pile-forming machinery by establishing a multi-objective function including the smoothness of the lifting speed, the accuracy of flow control and the minimization of construction cost, so as to obtain a more comprehensive and practical parameter control strategy. According to the different stages of the construction of the pile-forming machinery, the optimization process is divided into several stages, and a weight value is set for each objective function of each stage, which can fully consider the dynamic changes in the construction process, so that the optimization strategy is closer to the actual construction needs. The method of setting weighted ideal points in stages is used to convert complex multi-objective optimization problems into single-objective optimization problems, which greatly simplifies the solution process. At the same time, the particle swarm optimization algorithm is used to solve the single-objective optimization problem, and the optimal model parameters of each stage can be found efficiently and accurately. The present invention realizes the precise control of the lifting speed of the drill rod and the pouring flow rate of the fluid cement-cured soil through an automatic control system, and the overall construction conditions of the pile body are kept consistent, which improves the uniformity and integrity of the spiral replacement cement-reinforced soil pile. The present invention pre-mixes the fluidized cement solidified soil by a slurry mixer, thereby ensuring the uniformity of the solidified soil mixing, solving the problems of uneven mixing and poor pile quality of traditional reinforced soil piles, and achieving the best working state of the cement solidified soil by pre-mixing, thereby improving the pile quality and the bearing capacity of the single pile. The present invention pre-drills holes by using a long spiral drill rod, which can accurately control the pile diameter. The long spiral drill rod plays a wall protection role during the drilling process to prevent the hole from collapsing. Compared with traditional cement soil mixing piles, the construction time can be reduced from the original 40min-50min to 10min-15min, the construction efficiency is improved by 2-3 times, and the construction progress is significantly accelerated. The present invention pre-drills holes by using a spiral drill rod. During the drilling process of the spiral drill rod, the in-situ soil body can be drilled out of the hole, and the drilled soil body can be used for the preparation of fluidized cement reinforced soil. The in-situ soil body can be used on-site without the need to transport the soil outside, which can effectively reduce the project cost and save costs.

[0229] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, optical storage, etc.) containing computer-usable program code.

[0230] The specific embodiments described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for controlling the construction of a pile-forming machine for spiral replacement cement-reinforced soil piles, characterized in that: The method comprises the following steps: S1. Adding the fluidized cement reinforced soil composite material into a slurry mixer according to a preset ratio for mixing to obtain fluidized cement reinforced soil; S2, based on the pile position determined by surveying and setting out, and according to the pre-set pile bottom elevation, use the auger to drill the pre-formed hole; S3, pouring fluid cement reinforcement soil into the pre-formed hole, and lifting the auger rod through the lifting system; the automatic control system dynamically adjusts the lifting speed of the auger rod and the flow rate of pouring fluid cement reinforcement soil according to the monitored pressure value of the pressure sensor; S4. When the auger rod is lifted to the preset pile bottom elevation, stop pouring the fluid cement reinforcement soil, clean the drill rod and proceed with the next pile construction.

2. The method for controlling the construction of a pile-forming machine for spiral replacement cement-reinforced soil piles according to claim 1, characterized in that: The automatic control system dynamically adjusts the lifting speed of the auger rod and the flow rate of pouring fluid cement reinforced soil according to the monitored pressure value of the pressure sensor, including the following steps: S31, constructing a data set using the acquired historical pressure values ​​of the pressure sensor, the historical lifting speed of the auger rod, the historical flow rate of the poured fluid cement reinforced soil, and the historical forming quality of the cement reinforced soil pile, and performing feature processing on the data set to obtain a feature set; S32, using the t-SNE algorithm to reduce the dimension of the feature set, and obtain a low-dimensional data set through iterative solution; S33, constructing and training a parameter control strategy model based on the obtained low-dimensional data set; S34, using the monitored pressure value of the pressure sensor as the input value of the parameter control strategy model, and outputting the optimal values ​​of the lifting speed of the auger rod and the flow rate of pouring fluid cement reinforced soil through model analysis; S35. Based on the optimal values ​​of the lifting speed of the auger rod and the flow rate of pouring the fluid cement reinforced soil, the automatic control system adjusts the lifting speed of the auger rod and the flow rate of pouring the fluid cement reinforced soil.

3. The method for controlling the construction of a pile-forming machine for spiral replacement cement-reinforced soil piles according to claim 2, characterized in that: The method of using the t-SNE algorithm to reduce the dimension of the feature set and obtaining a low-dimensional data set by iterative solution includes the following steps: S321, for each feature point in the feature set, calculating the joint probability between the feature point and other feature points, and constructing a joint probability distribution matrix in a high-dimensional space; S322, randomly initialize a low-dimensional data set, and calculate a joint probability distribution matrix in the low-dimensional space; S323. Calculate the KL divergence of the joint probability distribution in the high-dimensional space and the joint probability distribution in the low-dimensional space, take minimizing the KL divergence as the objective function, and iteratively solve the low-dimensional data set using the gradient descent method.

4. The method for controlling the construction of a pile-forming machine for spiral replacement cement-reinforced soil piles according to claim 3 is characterized in that: For each feature point in the feature set, calculating the joint probability between the feature point and other feature points, and constructing a joint probability distribution matrix in a high-dimensional space includes the following steps: S3211. For each feature point in the feature set, calculate the Euclidean distance between the feature point and other feature points; S3212, clustering the Euclidean distance using a K-means clustering algorithm, and assigning weights to the Euclidean distance according to the clustering result to obtain a weighted distance; S3213. Using the weighted distance, calculate the joint probability between each feature point and other feature points in the high-dimensional space, and construct a joint probability distribution matrix in the high-dimensional space based on the joint probability.

5. The method for controlling the construction of a pile-forming machine for spiral replacement cement-reinforced soil piles according to claim 4 is characterized in that: The method of clustering the Euclidean distance using the K-means clustering algorithm and assigning weights to the Euclidean distance according to the clustering results to obtain the weighted distance includes the following steps: S32121. Randomly select K feature points as initial cluster centers; S32122. According to the Euclidean distance between each feature point and other feature points, each feature point is assigned to the cluster where the centroid with the closest distance is located; S32123. For each cluster, calculate the average Euclidean distance of all feature points belonging to the cluster, and use this average value as the new cluster center; S32124, repeat steps S32122 and S32123 until the centroid no longer changes, and obtain a clustering result; S32125. Based on the clustering results, the table parameter optimization method is used to assign weights to the Euclidean distance to obtain the weighted distance.

6. The method for controlling the construction of a pile-forming machine for spiral replacement cement-reinforced soil piles according to claim 5, characterized in that: The formula for calculating the joint probability between each feature point and other feature points in the high-dimensional space by using weighted distance is: In the formula, p ij Represents the feature point x i and feature point x j The joint probability between j|i Represents the feature point x j As feature point x i Neighbor’s probability, p i|j Represents the feature point x i As feature point x j The probability of neighbors, N represents the total number of feature points, Represents the feature point x i and feature point x j The weighted distance between i Represents the feature point x i The Gaussian variance centered at Represents the feature point x i and feature point x n The weighted distance between i|j Represents the feature point x i As feature point x j The probability of neighbors, Represents the feature point x j and feature point x i The weighted distance between Represents the feature point x j and feature point x n The weighted distance between j Represents the feature point x j The variance of a Gaussian centered at .

7. The method for controlling the construction of a pile-forming machine for spiral replacement cement-reinforced soil piles according to claim 2, characterized in that: The step of constructing and training the parameter control strategy model based on the obtained low-dimensional data set includes the following steps: S331, dividing the obtained low-dimensional data set into a training set and a test set; S332, using the monitored pressure value of the pressure sensor as input, the lifting speed of the auger rod and the flow rate of the poured fluid cement reinforced soil as output, and using the training set to train the deep convolutional neural network to obtain a primary parameter control strategy model; S333. Evaluate the parameter control strategy model using the test set. Based on the evaluation results, optimize and adjust the primary parameter control strategy model using the phased weighted ideal point method to obtain the final parameter control strategy model.

8. The method for controlling the construction of a pile-forming machine for spiral replacement cement-reinforced soil piles according to claim 7, characterized in that: The method of evaluating the parameter control strategy model using the test set, optimizing and adjusting the primary parameter control strategy model using the phased weighted ideal point method based on the evaluation result, and obtaining the final parameter control strategy model includes the following steps: S3331. Based on the construction requirements of the pile-driving machine, a multi-objective function is established, wherein the multi-objective function includes the smoothness of the lifting speed, the accuracy of the flow control, and the minimization of the construction cost; S3332. Evaluate the performance of the primary parameter control strategy model using the test set and calculate the value of each objective function; S3333, dividing the optimization process of the primary parameter control strategy model into several phases according to different phases of the pile driving machine construction; and setting a weight value for each objective function of each phase; S3334. Using the staged weighted ideal point method, the multi-objective optimization problem is transformed into a single-objective optimization problem, and the single-objective optimization problem is solved by the particle swarm optimization algorithm to obtain the optimal model parameters for each stage; S3335. Integrate the optimal model parameters of each phase to obtain the final parameter control strategy model.

9. The method for controlling the construction of a pile-forming machine for spiral replacement cement-reinforced soil piles according to claim 7, characterized in that: The method of using the staged weighted ideal point method to transform the multi-objective optimization problem into a single-objective optimization problem, and solving the single-objective optimization problem by the particle swarm optimization algorithm to obtain the optimal model parameters for each stage includes the following steps: S33341. For each objective function of each period, determine the ideal value based on historical data; S33342. Construct a single objective optimization function according to the weight value of each period and the deviation between each objective function and the ideal value; The expression of the single objective optimization function is: Where L(θ) represents a single objective optimization function, H represents the number of objective functions, and ω t represents the weight value of the tth objective function, F t (θ) represents the value of the tth objective function, F t * represents the ideal value of the tth objective function; S33343. Randomly generate an initial solution for each target optimization function, wherein the initial solution represents a model parameter; S33344. Use the particle swarm optimization algorithm to iteratively solve the single objective optimization problem. When the preset number of iterations is reached, stop the iteration and output the optimal model parameters for each period.

10. A pile forming machine for spiral replacement cement reinforced soil piles, used to implement the construction control method for the pile forming machine for spiral replacement cement reinforced soil piles according to any one of claims 1 to 9, characterized in that: The machine tool comprises an automatic control system arranged at the top of one side of the lifting system, a slurry mixer is arranged at one side of the automatic control system, an auger rod is arranged at the top of the other side of the lifting system, a grouting pipe connected to the slurry mixer is arranged inside the auger rod, and a pressure sensor is arranged at the bottom end of the auger rod; A drum is arranged inside the slurry mixer, and gear teeth are arranged on the inner wall of the drum, and the cross section of the gear teeth is a regular body structure, and a grouting port is arranged between adjacent gear teeth; The spiral drill rod is composed of a first drill rod and a second drill rod, and the first drill rod and the second drill rod are connected by bolts, the first drill rod is located at the top of the second drill rod, and the top of the first drill rod is connected to the lifting system; The top ends of the first drill rod and the second drill rod are both provided with a plurality of fixing screws matched with the bolts, and the bottom ends of the first drill rod and the second drill rod are both provided with a plurality of screw holes matched with the fixing screws.

Images