Cement mixing pile forming quality real-time prediction system

Abstract

The invention discloses a real-time prediction system for the pile-forming quality of a cement mixing pile, and relates to the technical field of pile-forming quality prediction. The real-time prediction system for the pile-forming quality of the cement mixing pile comprises a pile periphery disturbance evolution analysis module, a construction and soil analysis module and a pile-forming quality prediction analysis module which are respectively used for acquiring geological three-dimensional point cloud data, soil property data and a construction parameter set of each time period; a pile periphery disturbance evolution driving index, a pile foundation response adaptability index and a construction disturbance implantation index are extracted, and a pile forming quality prediction analysis module synthesizes the three indexes to calculate and obtain a pile body pile forming quality driving index in the next time period; quality feedback processing is conducted on the set cement mixing pile through the pile body pile forming quality driving index of the next time period in the pile forming quality feedback module, so that procedural early warning and response type intervention on the pile body quality are achieved, and then the continuous stability of pile body construction and the consolidation quality reliability are remarkably improved.

Description

Technical Field

[0001] The invention relates to the technical field of pile quality prediction, in particular to a real-time prediction system for the quality of cement mixing piles. Background Art

[0002] The Dongting Lake area, as a typical area of silt and soft soil foundation in my country, has long faced problems such as large settlement of embankment structures, strong fluctuations in pile quality, and poor construction adaptability. In particular, in key engineering links such as embankment reinforcement and soft foundation treatment, traditional cement mixing pile construction technology faces multiple technical challenges. On the one hand, the soil structure in the Dongting Lake area is complex, with a large number of buried boulders, soft and hard interlayers, and soft soil structures with large differences in water content, resulting in uneven pile resistance and strong pile disturbance during construction, affecting the continuity and structural stability of the pile body; on the other hand, traditional pile quality evaluation methods mostly rely on post-completion sampling inspection, lack the ability to continuously perceive and provide real-time feedback on the pile process, and are unable to meet the technical requirements of high-level embankment projects for construction accuracy and process controllability.

[0003] Existing quality control methods for cement mixing piles are mostly based on experience-based design and post-testing. They lack dynamic disturbance monitoring, pile foundation soil response identification, and construction behavior modeling and analysis mechanisms during the construction process. As a result, construction quality problems are easily identified only after the pile is formed, delaying the treatment opportunity, increasing repair costs, and affecting the stability and durability of the project.

[0004] Prior art, such as the patent application with publication number CN117371168A, discloses a real-time prediction method, system, device, and medium for cement mixing pile quality. The method includes: determining soil layer information corresponding to different depth segments in the construction area where the mixing pile is located based on survey information of the construction area where the mixing pile is located, the soil layer information including soil layer type and soil layer parameters corresponding to the soil layer type; obtaining construction data recorded at preset intervals during the construction of the mixing pile, and based on the obtained construction data, calculating the construction parameters of the mixing pile at different depth segments; presetting an age, and for each depth segment, combining the soil layer information, construction parameters, and the set age corresponding to the depth segment as input information to obtain input information corresponding to each depth segment; inputting the input information corresponding to each depth segment into a constructed pile quality prediction model to obtain the pile quality corresponding to each depth segment. The present invention can quickly predict the pile quality of the mixing pile.

[0005] Based on the above solution, it was found that the limitations of existing technologies include at least the following problems. Existing technologies have difficulty in constructing a dynamic coupling modeling mechanism between pile-surrounding disturbance behavior, pile foundation response capability, and construction disturbance input. As a result, pile quality prediction lacks a unified expression of the interaction relationship between the three key factors at the structural level. This easily leads to the quality prediction process remaining at the level of static regression or empirical fitting based on soil parameters and construction data. This makes it difficult to characterize the nonlinear dynamic evolutionary behaviors such as disturbance diffusion, structural deformation, and plastic evolution caused by construction disturbance during actual construction. Furthermore, it is difficult to accurately express the driving effect of construction behavior on pile-surrounding disturbance, the adaptive evolution path of disturbance on soil structure, and the reverse impact of pile foundation response on construction control strategies. Consequently, the prediction results lack the ability to explain the physical mechanism of the quality evolution process, lack the ability to dynamically track the safety status of the pile foundation structure, and are difficult to make real-time control decisions based on the prediction results. This limits the quality perception, risk warning, and active control capabilities of intelligent cement mixing pile construction. Furthermore, it is difficult to support high-precision quality assessment and active optimization of construction behavior in complex geological environments, which restricts the further improvement of intelligent engineering quality control capabilities. Summary of the Invention

[0006] In response to the shortcomings of the existing technology, the present invention provides a real-time prediction system for the quality of cement mixing piles, which solves the problem that the existing technology lacks a dynamic coupling modeling mechanism for disturbance, response and construction, which limits the improvement of the prediction accuracy and control ability of pile quality.

[0007] To achieve the above objectives, the present invention is implemented through the following technical solutions: a real-time prediction system for the quality of cement mixing piles, comprising: a pile-circumference disturbance evolution analysis module, used to continuously obtain geological three-dimensional point cloud data of each time period of a set cement mixing pile, and perform comprehensive analysis in combination with a pre-trained disturbance recognition model to obtain the pile-circumference disturbance evolution driving index of the corresponding time period; a construction and soil analysis module, used to obtain the construction parameter set of each time period of the set cement mixing pile and the soil property data of each time period of several soil layers, and respectively analyze the construction disturbance implantation index and pile foundation response adaptability index of the corresponding time period; a pile quality prediction and analysis module, used to predict and analyze the pile-circumference disturbance evolution driving index, pile foundation response adaptability index and construction disturbance implantation index of each time period of the set cement mixing pile, and obtain the pile body pile quality driving index of the next time period of the set cement mixing pile; a pile quality feedback module, used to perform quality feedback processing on the set cement mixing pile based on the pile body pile quality driving index of the next time period.

[0008] Furthermore, the geological three-dimensional point cloud data includes the voxel value, three-dimensional coordinates and corresponding reflection intensity value of each voxel point. The specific steps for obtaining the geological disturbance index of each time period of the set cement mixing pile are as follows: the geological three-dimensional point cloud data of each time period of the set cement mixing pile is input into the pre-trained disturbance recognition model for interference feature analysis to obtain the interference evaluation set of the corresponding time period, including the disturbance directional potential energy index, the stress transfer tensor index, the shear aggregation index and the plastic boundary penetration index; based on the interference evaluation set of each time period of the set cement mixing pile, the geological disturbance index of the corresponding time period is analyzed.

[0009] Furthermore, the disturbance recognition model is specifically a graph structure neural network, including a point cloud map construction layer, a neighborhood disturbance coding layer, a time coding layer, a physical constraint embedding layer, and a mapping output layer. The specific steps of obtaining the interference evaluation set for each time period of the set cement mixing pile are as follows: in the point cloud map construction layer of the graph structure neural network, the geological three-dimensional point cloud data for each time period of the set cement mixing pile is received and graph construction processing is performed; in the neighborhood disturbance coding layer of the graph structure neural network, the geological three-dimensional point cloud data for each time period of the set cement mixing pile after graph construction processing is performed neighborhood feature coding processing, and the disturbance feature vector set of the corresponding time period is analyzed; in the graph structure neural network, the geological three-dimensional point cloud data for each time period of the set cement mixing pile is received and graph construction processing is performed; in the neighborhood disturbance coding layer of the graph structure neural network, the geological three-dimensional point cloud data for each time period of the set cement mixing pile is performed after graph construction processing, and the disturbance feature vector set of the corresponding time period is analyzed; in the graph structure neural network, the geological three-dimensional point cloud data for each time period of the set cement mixing pile is received and graph construction processing is performed; in the neighborhood disturbance coding layer of ... In the time encoding layer of the structural neural network, the disturbance feature vector set of each time period of the set cement mixing pile is subjected to time evolution processing; in the physical constraint embedding layer of the graph structure neural network, the disturbance feature vector set of each time period of the set cement mixing pile after time evolution processing is subjected to structural constraint mapping processing to obtain the physical consistency disturbance representation vector set of the corresponding time period; in the mapping output layer of the graph structure neural network, the physical consistency disturbance representation vector set of each time period of the set cement mixing pile is subjected to feature mapping processing to obtain the disturbance directional potential energy index, stress transfer tensor index, shear aggregation index, and plastic boundary penetration index of the corresponding time period.

[0010] Furthermore, the specific formula for calculating the driving index of the disturbance evolution around the cement mixing pile in a certain period of time is as follows: ;in, To set the driving index of the disturbance evolution around the cement mixing pile in a certain period of time, To set the disturbance convergence potential energy index of cement mixing pile in a certain period of time, is the clustering coefficient stored in the database, To set the stress transfer tensor index of a cement mixing pile at a certain period, is the transfer coefficient stored in the database, To set the shear polymerization index of cement mixing pile at a certain period of time, is the aggregation coefficient stored in the database, To set the plastic boundary penetration index of cement mixing pile at a certain period, is the penetration coefficient stored in the database, is the synergy coefficient stored in the database.

[0011] Furthermore, the soil property data includes a water content change value, a pore fluctuation ratio, a medium resistance change rate value, a microstrain rate disturbance amplitude, and a seepage disturbance index. The specific steps for analyzing the pile foundation response adaptability index of each soil layer and each time period for setting the cement mixing pile are as follows: evaluating and analyzing the soil property data of each soil layer and each time period for setting the cement mixing pile to obtain a soil disturbance evaluation set for the corresponding time period, including a pile foundation structure disturbance accommodation index and a pile foundation seepage stress buffer index; based on the soil disturbance evaluation set of each soil layer and each time period for setting the cement mixing pile, analyzing the pile foundation response adaptability index for the corresponding time period.

[0012] Furthermore, the specific steps for obtaining the soil disturbance assessment set for each soil layer and each time period of the set cement mixing pile are as follows: based on the Z-score standardization, the pore fluctuation ratio and micro-strain rate disturbance amplitude of each soil layer and each time period of the set cement mixing pile are weighted analyzed to obtain the pile foundation structure disturbance accommodation index of the corresponding time period; based on the Z-score standardization, the water content change value, medium resistance change rate value and seepage disturbance index of each soil layer and each time period of the set cement mixing pile are weighted analyzed to obtain the pile foundation seepage stress buffer index of the corresponding time period.

[0013] Furthermore, the construction parameter set includes a mixing ratio value, a drilling speed value, a mixing shaft rotation value, a grouting pressure value, a grouting return pressure value, a rod torque fluctuation value, and a grouting fluctuation frequency value. The specific steps for analyzing and setting the construction disturbance implantation index for each time period of the cement mixing pile are as follows: a comprehensive analysis is performed on the construction parameter set for each time period of the cement mixing pile to obtain the construction behavior evaluation set for the corresponding time period, including the grouting dynamic disturbance index and the drilling and mixing motion disturbance index.

[0014] Furthermore, the specific steps for obtaining the construction behavior evaluation set for each time period of the set cement mixing pile are as follows: obtaining the environmental correction factor for each time period of the set cement mixing pile, and performing weighted analysis with the mixing ratio value, shotcrete pressure value, shotcrete return pressure value, and grouting fluctuation frequency value of the corresponding time period to obtain the grouting dynamic disturbance index of the corresponding time period; performing weighted analysis on the drilling speed value, stirring shaft rotation value, and rod torque fluctuation value of each time period of the set cement mixing pile based on Z-score standardization to obtain the drilling and mixing motion disturbance index of the corresponding time period.

[0015] Furthermore, the specific steps for obtaining the pile quality index of the next time period of the set cement mixing pile are as follows: based on the pile-circumference disturbance evolution driving index, the pile foundation response adaptability index, and the construction disturbance implantation index of each time period of the set cement mixing pile, the pile body pile quality driving index of the corresponding time period is analyzed; and based on the pile body pile quality driving index of each time period of the set cement mixing pile, a trend evolution analysis is performed to obtain the pile body pile quality driving index of the next time period of the set cement mixing pile.

[0016] Furthermore, the specific steps for calculating the pile quality driving index of a set cement mixing pile in a certain period of time are as follows: ;in, To set the pile quality driving index of cement mixing pile in a certain period of time, To set the driving index of the disturbance evolution around the cement mixing pile in a certain period of time, To set the pile foundation response adaptability index of cement mixing pile in a certain period of time, is the superposition coefficient stored in the database, is the coefficient of variation stored in the database, To set the construction disturbance implantation index of cement mixing piles in a certain period of time, is the construction coefficient stored in the database.

[0017] The present invention has the following beneficial effects: (1) The real-time prediction system for the quality of cement mixing piles forms a complete dynamic prediction closed-loop mechanism by constructing the pile-surrounding disturbance evolution driving index, the pile foundation response adaptability index and the construction disturbance implantation index, and performs continuous time period data acquisition and structured modeling. Then, through the trend evolution analysis method, the pile quality driving index is predicted and judged in each time period, thereby effectively supporting the execution of subsequent control strategies. The system can not only identify risks such as increased disturbance, decreased soil bearing capacity or severe construction disturbance in real time during the pile construction process, but also adjust the construction parameters according to the prediction results, realize process-based early warning and responsive intervention of pile quality, thereby significantly improving the continuous stability of pile construction and the reliability of consolidation quality.

[0018] (2) The real-time prediction system for the quality of cement mixing piles realizes multi-level physical perception and structural mapping of disturbance evolution characteristics by constructing a graph structure neural network model based on geological three-dimensional point cloud. The model includes point cloud map construction, neighborhood disturbance encoding, time evolution processing, physical constraint embedding and index mapping output, so that the disturbance behavior of each voxel point in the pile formation process can be spatially modeled, time-series tracked and physically constrained, and the corresponding structural indicators can be output, providing a high-dimensional and physically consistent expression of the disturbance characteristics around the pile. In particular, the rules such as disturbance conservation, shear limit and plastic boundary penetration are introduced in the physical constraint layer to ensure that the output characteristics meet the inherent mechanical laws of geological disturbance evolution, thereby significantly enhancing the characteristic adaptability and prediction accuracy of the model under variable geological conditions.

[0019] (3) The real-time prediction system for the quality of cement mixing piles has established a complete closed-loop mechanism of prediction, judgment and control by constructing a pile quality prediction analysis module and a pile quality feedback module. The system continuously calculates the pile quality driving index in each period and introduces a sliding window trend evolution analysis method to form a basic prediction value, a disturbance amplification factor, a proportional consistency factor and a disturbance negative adjustment factor, and then analyzes the prediction value for the next period. The quality feedback module uses the prediction value as a judgment basis to determine in real time whether there is a quality downside risk in the construction status and jointly executes parameter adjustment measures, so that the system has the ability to continuously adapt to the adjustment. In addition, the mechanism can quickly respond to quality fluctuations during the pile construction process, realize the closed-loop operation of continuous prediction, active judgment and real-time control, and thus enhance the quality assurance capability of the pile process.

[0020] Of course, any product implementing the present invention does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 This is a block diagram of a real-time prediction system for cement mixing pile quality according to the present invention.

[0022] Figure 2 The present invention provides a flow chart of the specific steps for obtaining the geological disturbance index of each time period of a set cement mixing pile in a real-time prediction system for the quality of cement mixing piles.

[0023] Figure 3 The present invention provides a schematic diagram of a time sequence of setting a grouting dynamic disturbance index of a cement mixing pile in a real-time prediction system for the quality of cement mixing piles.

[0024] Figure 4 The present invention provides a schematic diagram of a time sequence of setting a disturbance index of the drilling motion of a cement mixing pile in a real-time prediction system for the quality of cement mixing piles.

[0025] Figure 5This is a flowchart of the specific steps for obtaining the pile quality index of a set cement mixing pile in the next time period in a real-time prediction system for the pile quality of a cement mixing pile of the present invention. DETAILED DESCRIPTION

[0026] See also Figure 1 The embodiment of the present invention provides a technical solution: a real-time prediction system for the quality of cement mixing piles, comprising: a pile-surrounding disturbance evolution analysis module, for obtaining geological three-dimensional point cloud data of each time period (e.g., 10 seconds as a time period) of the set cement mixing piles in a set area (such as the Dongting Lake area) during the set cement mixing pile construction process, and performing comprehensive analysis in combination with a pre-trained disturbance recognition model to obtain the pile-surrounding disturbance evolution driving index of the corresponding time period (reflecting the response evolution intensity of the soil structure around the pile under the action of pile construction disturbance, mainly characterizing the changing trend of disturbance phenomena such as shear aggregation, plastic deformation, and crack conduction. The higher the index value, the more severe the disturbance effect on the pile environment, the lower the structural stability, and the worse the pile construction quality); a construction and soil analysis module, for obtaining a construction parameter set for each time period of the set cement mixing piles and soil property data for each time period of several soil layers, and respectively analyzing the construction disturbance implantation index (the disturbance amplitude introduced by the construction behavior and its influence on the structural implantation of the pile body during the pile construction process) and the pile foundation response adaptability index (the influence of the soil layer on the soil structure). The module comprises a pile quality prediction and analysis module, which is used to predict and analyze the pile periphery disturbance evolution driving index, pile foundation response adaptability index, and construction disturbance implantation index of each time period of the set cement mixing pile, and obtain the pile quality driving index of the corresponding time period. The module comprises a pile quality feedback module, which is used to provide quality feedback to the set cement mixing pile based on the pile quality driving index of the next time period. Specifically, the module determines whether the pile quality driving index of the set cement mixing pile in the next time period is lower than a preset pile quality driving index threshold. If the pile quality driving index is lower than the preset pile quality driving index threshold, the current construction status is judged to have a quality downgrade risk, and control adjustment measures are implemented, including but not limited to resetting construction parameters, switching mixing modes, or correcting slurry ratios, to enhance pile stability and consolidation quality. If the pile quality driving index is not lower than the preset pile quality driving index threshold, the current construction status is judged to be within an acceptable quality range, the current construction parameters are maintained, and the data collection and analysis process for the next time period begins.

[0027] Specifically, if Figure 2As shown, the geological three-dimensional point cloud data includes the voxel value, three-dimensional coordinates and corresponding reflection intensity value of each voxel point. The specific steps for obtaining the geological disturbance index of each time period of the set cement mixing pile are as follows: the geological three-dimensional point cloud data of each time period of the set cement mixing pile is input into the pre-trained disturbance recognition model for disturbance feature analysis to obtain the disturbance evaluation set of the corresponding time period, including the disturbance directional potential energy index, stress transfer tensor index, shear aggregation index and plastic boundary penetration index; based on the disturbance evaluation set of each time period of the set cement mixing pile, the geological disturbance index of the corresponding time period is analyzed.

[0028] The specific formula for calculating the driving index of the disturbance evolution around a cement mixing pile in a certain period of time is as follows: ;in, To set the driving index of the disturbance evolution around the cement mixing pile in a certain period of time, To set the disturbance convergence potential energy index of cement mixing pile in a certain period of time, is the clustering coefficient stored in the database, To set the stress transfer tensor index of a cement mixing pile at a certain period, is the transfer coefficient stored in the database, To set the shear polymerization index of cement mixing pile at a certain period of time, is the aggregation coefficient stored in the database, To set the plastic boundary penetration index of cement mixing pile at a certain period, is the penetration coefficient stored in the database, is the synergy coefficient stored in the database.

[0029] It needs to be explained that the formula This term is used to regulate the amplification effect between the perturbation convergence trend and the plastic boundary penetration behavior to prevent the model from experiencing exponential deviation or physical imbalance in local high-coupling areas.

[0030] The specific expression of the tanh function is: ,in, is a natural constant and can be taken as 2.71 in this embodiment, with a domain of (−∞, +∞) and a range of (−1, +1).

[0031] 、 、 、 、 It can be obtained through the following steps: based on historical data, the initial influence weights of each variable (disturbance convergence potential energy index, stress transfer tensor index, shear aggregation index, plastic boundary penetration index) on the pile disturbance evolution driving index are determined through statistical regression analysis. Then, the disturbance response analysis method is used (multiple value intervals are set for each of the above variables respectively. While keeping other parameters unchanged, the test values are input in sequence and the output pile disturbance evolution driving index is calculated to observe the changing trend and response amplitude of the output results, thereby verifying the influence of each coefficient on the model stability) to adjust the value range of the coefficient to evaluate the stability and applicability of these parameters to the formula output. Next, the weights are further fitted through model optimization (such as machine learning algorithm) to ensure that the formula can accurately reflect the evolution state of the actual pile disturbance.

[0032] In this implementation scheme, by introducing the interference assessment set and weight fitting mechanism, the structural stability enhancement and parameter adaptability optimization of the pile disturbance evolution driving index calculation process are achieved. Based on the geological three-dimensional point cloud data of each time period, the system extracts the disturbance convergence potential energy, stress transfer tensor, shear aggregation and plastic boundary penetration index, and constitutes an interference assessment set, so that the disturbance state around the pile has a clear structural characterization basis. Secondly, the initial influence weight of each disturbance index is obtained by the statistical regression method, and multiple groups of variable combinations are set for simulation in combination with the disturbance response analysis to quantify the response sensitivity of each coefficient to the model output fluctuation, thereby screening out the abnormal amplification behavior of the high-coupling section, and providing a reliable and robust disturbance-driven basic indicator for subsequent pile quality prediction.

[0033] Specifically, the disturbance recognition model is a graph structure neural network, including a point cloud map construction layer, a neighborhood disturbance coding layer, a time coding layer, a physical constraint embedding layer, and a mapping output layer. The specific steps of obtaining the interference evaluation set for each time period of the set cement mixing pile are as follows: in the point cloud map construction layer of the graph structure neural network, the geological three-dimensional point cloud data of each time period of the set cement mixing pile (the voxel value, three-dimensional coordinate and corresponding reflection intensity value of each voxel point in each time period) are received and the graph construction processing is performed; in the neighborhood disturbance coding layer of the graph structure neural network, the geological three-dimensional point cloud data of each time period of the set cement mixing pile after the graph construction processing are subjected to neighborhood feature coding processing, and the disturbance of the corresponding time period is analyzed. dynamic feature vector set; in the time encoding layer of the graph structure neural network, the disturbance feature vector set of each time period of the set cement mixing pile is subjected to time evolution processing; in the physical constraint embedding layer of the graph structure neural network, the disturbance feature vector set of each time period of the set cement mixing pile after time evolution processing is subjected to structural constraint mapping processing (based on physical rules) to obtain the physical consistency disturbance representation vector set of its corresponding time period; in the mapping output layer of the graph structure neural network, the physical consistency disturbance representation vector set of each time period of the set cement mixing pile is subjected to feature mapping processing to obtain the disturbance directional potential energy index, stress transfer tensor index, shear aggregation index, and plastic boundary penetration index of its corresponding time period.

[0034] The specific steps of graph construction processing are as follows: first, the point cloud data collected in each period are preliminarily screened to remove voxel points with missing coordinates, reflection intensity or voxel value lower than the corresponding threshold, so as to ensure the stability and reliability of the node data; then, the three-dimensional coordinates of all retained voxel points are uniformly scaled so that they fall within a unified numerical range; then, with each voxel point as the center, several voxel points with the closest spatial distance are searched to form a set of adjacent points, and undirected edges are established between each group of center points and adjacent points to form the initial edge set of the graph structure, so as to ensure that the graph construction has time series distribution. The algorithm takes into account the cutness and spatial locality of each edge, and then performs a weighted calculation on each edge based on the three-dimensional distance, voxel value difference and reflection intensity difference between its connection points, and assigns a weight value representing the intensity of disturbance propagation to the edge. This weight will be used as the neighborhood aggregation weight in the graph neural network to strengthen the directionality and intensity characteristics in the disturbance path. Finally, the complete graph structure data of the current period is output, including a node set (voxel points after screening and normalization), an edge set (undirected edges between each node and its adjacent points) and a node feature vector set (including fields such as three-dimensional coordinates, reflection intensity, disturbance voxel value and timestamp).

[0035] The specific steps of the neighborhood feature encoding process are as follows: for each voxel point node, the set of adjacent nodes connected to it in the graph structure is traversed, and the feature vectors of all its adjacent nodes in the current time period are collected, including the normalized coordinate information, reflection intensity value, and other original physical properties of the adjacent voxel points. On this basis, the feature information of the adjacent nodes is aggregated to the central node in a weighted manner to generate the perturbation response feature vector of the central node in the current graph state. During the aggregation process, not only the spatial geometric distance between the adjacent points and the central point is considered, but also the connection edge weights recorded in the edge set are introduced as aggregation weight factors to strengthen the influence of the neighboring points with larger perturbation intensity on the representation of the current node. Subsequently, a feature transformation operation is performed on the aggregated local perturbation vector, that is, multi-scale perturbation structure features are extracted through multi-layer graph convolution. After multiple rounds of graph convolution processing, the perturbation structure representation vector of each voxel point node in the current time period can be obtained. This vector represents multiple perturbation behavior characteristics such as the perturbation aggregation effect, perturbation direction trend, and perturbation intensity unevenness of the voxel point in its spatial neighborhood.

[0036] The specific steps of time evolution processing are as follows: each continuous pile formation period of the cement mixing pile is numbered to form an ordered time series, and the disturbance feature vector of each period is ensured to correspond one-to-one with the voxel point position to form a time series sample. For each voxel point, the disturbance feature vectors of multiple periods are combined in chronological order to form a disturbance evolution sequence. A fixed-length sliding time window is used to segmentally encode the disturbance sequence. Each segment is input into a one-dimensional temporal convolutional network to extract local time features such as disturbance intensity increase, direction drift, and frequency change. Then, multi-layer convolution and activation functions are used to model the short-term and medium- to long-term change trends. Subsequently, a gated recurrent network or a temporal attention mechanism is used to perform segmented encoding on the disturbance sequence. The disturbance sequence is globally modeled to capture rhythmic evolution, mutation and high-frequency aggregation behavior, and differentiated attention weights are assigned to each time slice to highlight key disturbance moments. Finally, a dynamic disturbance representation vector of fixed dimension is output to characterize the speed, rhythm and cumulative characteristics of the disturbance, which is used to identify the plastic area around the pile. The basis for identifying the plastic area is: the disturbance intensity continues to increase, the direction consistency is enhanced and the spatial aggregation is improved. The voxel subsets that meet the above conditions are clustered. If the disturbance enhancement, direction concentration and high density aggregation characteristics are met simultaneously in multiple time periods, the area is marked as a plastic zone, and the boundary is recorded to form a plastic area boundary set for subsequent physical constraint analysis and structural index modeling.

[0037] The specific steps of the structural constraint mapping process are as follows: for each voxel node in its corresponding time period, extract its perturbation intensity amplitude, perturbation direction vector, perturbation evolution rate and its connection relationship characteristics in the local graph structure as the original input field for subsequent constraint embedding; then, perform perturbation conservation constraint processing: in each local spatial neighborhood, with the voxel node as the center, calculate the sum of the perturbation intensities of all its adjacent nodes in the time period and compare it with the perturbation intensity of the central node. If the difference exceeds the set perturbation conservation tolerance threshold (for example, ±5%), this threshold is determined by The experimental data and the soil physical properties are analyzed and the disturbance is determined to be a non-conservative area; the system constructs a disturbance conservation residual term, and uses the residual term as part of the loss function to participate in the weight update of the feature mapping process to ensure that the disturbance feature meets the condition of relatively stable total disturbance within the neighborhood; shear limit constraint processing: for each voxel point, a field representing the rate of change of the shear direction is separated from the disturbance feature, and the rate of change of the angle between the point and its adjacent point is calculated to determine whether it exceeds the preset soil shear response limit (for example, set to 30° / s, this value is based on The shear limit obtained by field soil layer experiments); if the limit is exceeded, the point is marked as a potential shear instability point, and the shear tension regularization term is introduced to limit the gradient growth rate of its disturbance direction to prevent the model from outputting physically unreasonable shear transition behavior; Plastic boundary penetration constraint processing: According to the plastic boundary voxel set identified in the previous stage, it is judged whether the disturbance direction vector of the current node points to the outside of the plastic boundary, and whether the disturbance intensity is higher than the set plastic response threshold (such as 3 times the disturbance mean, a parameter verified by experiments); If both conditions are met, it is regarded as a penetrating disturbance behavior, and this type of disturbance will be A penetration penalty term is added to the voxel points, and the propagation of their disturbance energy outside the plastic boundary is restricted by embedding the boundary adsorption function. The above three physical constraint rules are all constructed as regularization functions in a differentiable form, and work together with the main task loss function during model training or real-time inference. The expression parameters of the disturbance features are updated in the back-propagation stage. After this processing, the output disturbance feature vector not only retains its original temporal and spatial distribution information, but also meets the requirements of local disturbance energy conservation, controllable shear behavior and boundary disturbance stability under the laws of geophysics, and finally forms a physically consistent disturbance representation vector set.

[0038] The specific steps of feature mapping processing are as follows: The specific steps of feature mapping processing are as follows: the physical consistency disturbance representation vector set of each voxel point in its corresponding time period is used as input. The representation vector contains the multi-dimensional disturbance feature field after disturbance conservation, shear constraint and boundary response regularization, including disturbance amplitude field, disturbance direction field, disturbance evolution rate field, direction consistency field and stress response gradient field, etc. Then, the following four index mapping extraction operations are performed in sequence: Extraction steps of the disturbance convergence potential energy index: Calculate the angle between the disturbance direction field of each voxel point and the disturbance direction field of its adjacent voxel point, and evaluate the aggregation trend of the neighboring disturbance directions based on the direction cosine similarity index; if more than 70% of the adjacent disturbance directions (the angle between the adjacent point disturbance direction and the pile center is less than 30°) tend to the pile center, it is considered to have high convergence. The system maps this convergence trend into a potential energy value, and combines it with the disturbance intensity field to form a weighted convergence potential energy value, which is converted into a normalized index value between 0 and 1 through a nonlinear mapping function (such as Sigmoid or Tanh); Stress transfer tensor index extraction steps: Based on the stress gradient response field in the disturbance feature, a tensor description matrix is constructed. By calculating the diffusivity of the disturbance energy along different directions in each voxel point in its neighborhood, a disturbance propagation anisotropy tensor is formed. Then, the angle between the tensor principal axis direction and the pile axis direction is calculated, and combined with the principal stress axis ratio to form a comprehensive tensor strength index. Finally, it is mapped into an index value describing the disturbance stress conduction efficiency, which is used to characterize the effective aggregation degree of the disturbance stress toward the pile body. Shear aggregation index extraction steps: Extract the disturbance direction change rate field and the shear response field, construct a local shear streamline feature map based on the continuous deviation trend of the disturbance direction within the sliding window, and calculate the coherence of the disturbance direction change. If multiple adjacent voxels show high coupling in disturbance direction and shear trend, the system regards this area as a shear coupling zone, further performs statistical analysis on its disturbance intensity mean and directional concentration, and forms a shear coupling aggregation index, which is then mapped to the shear aggregation index through a multi-layer perceptron. The steps for extracting the plastic boundary penetration index are as follows: the disturbance direction vector of the voxel point in the current period is compared with the normal direction of the plastic boundary. If the direction has strong penetration (the angle is close to 180°) and the disturbance intensity is higher than the preset boundary response threshold, it is identified as a penetration candidate point. Such points are further clustered to form a penetration area distribution map, and the ratio of the total disturbance energy in the area to the total area of the boundary is calculated as the penetration intensity index. Finally, the index is normalized and mapped to the plastic boundary penetration index, which reflects the risk degree of the disturbance to the stability of the boundary structure.

[0039] The pre-training process of the graph structure neural network is as follows: A disturbance identification training dataset was constructed. The dataset consists of geological 3D point cloud data and construction cycle segmented data for several cement mixing piles in each time period, and is annotated by geological engineering experts based on on-site monitoring records and construction feedback. Each set of samples in the dataset contains the voxel coordinates, voxel value, reflection intensity, actual plastic zone boundary, and the true value labels of four types of structural indices (disturbance convergence potential energy index, stress transfer tensor index, shear aggregation index, and plastic boundary penetration index) of the set cement mixing piles in multiple consecutive time periods. Each sample in the dataset is fully annotated and divided into a training set and a validation set after consistency verification.

[0040] The parameters of each processing layer of the graph structure neural network are initialized. The point cloud map construction layer is initialized using mean normalization and adjacent edge constraints; the neighborhood perturbation coding layer is initialized using Xavier to ensure local aggregation stability; the temporal coding layer is initialized using Kaiming; the physical constraint embedding layer and the mapping output layer are initialized using He-Uniform to ensure stable convergence of the regularization path.

[0041] Training is performed based on the training set, and the number of training rounds (such as 100 rounds) is set. In each round of training, the following steps are performed in sequence: Forward propagation stage: each set of input data (including the geological three-dimensional point cloud data of each period of multiple continuous periods of cement mixing piles and the corresponding disturbance index true value) is input into the graph structure neural network, and the predicted values of the four types of indexes are output respectively through each layer. Loss function construction stage: loss functions are designed for the corresponding indexes respectively, and the disturbance directional potential energy index and the shear aggregation index use L2 regression loss to fit their normalized numerical trends; the stress transfer tensor index is constructed by combining the angle loss function based on the difference in the direction of the tensor principal axis and the mean square error loss function; the plastic boundary penetration index is constructed by the edge An exponential deviation penalty loss function composed of the bounded energy penetration is used; the above loss functions are weightedly combined according to the influence weight of each disturbance index in quality prediction to form a unified total loss function for model optimization. In the back propagation and parameter optimization stage: full-graph back propagation is performed based on the total loss, all trainable parameters are updated, and the AdamW optimizer is used for adaptive adjustment. The gradient clipping mechanism is enabled during training to avoid the problem of gradient explosion caused by local disturbance propagation, and batch normalization (BatchNorm) and Dropout mechanisms are combined to enhance the generalization ability of the model. The learning rate is dynamically adjusted during training, and the cosine annealing scheduling strategy is used to enable the model to slowly approach the optimal convergence state in the later stage of training.

[0042] After each round of training, the model performance is evaluated based on the validation set. The mean absolute error (MAE), mean square error (MSE), and R² evaluation indicators of the perturbation convergence potential energy index, stress transfer tensor index, shear aggregation index, and plastic boundary penetration index are calculated, and trend curves of training loss and validation error over epochs are plotted. If the validation error of various perturbation indices does not decrease significantly over several consecutive rounds of training, the early stopping mechanism is automatically triggered to terminate training early to prevent overfitting.

[0043] After training is completed, the parameter file of the converged graph structure neural network model is saved for deployment and use in the real-time assessment of cement mixing pile quality.

[0044] In this implementation, a five-layer graph-structured neural network model is constructed to achieve high-precision, physically constrained, and consistent representation of the evolution of disturbances around cement mixing piles during construction. In the point cloud construction layer, the system constructs highly robust graph-structured data based on the spatial adjacency and disturbance differences of the geological three-dimensional point cloud. In the neighborhood disturbance encoding layer, an edge weighting mechanism and graph convolution operations are introduced to ensure that the disturbance aggregation direction and local structural response truly reflect the disturbance behavior. In the temporal encoding layer, convolution and gating mechanisms are integrated to accurately model the evolution of disturbance behavior over the entire time period. In the physical constraint embedding layer, physical rules such as disturbance conservation, shear limit, and plastic boundary are introduced to embed the disturbance representation vector in a regularized manner to ensure that the disturbance expression meets geomechanical consistency. Finally, the corresponding index is extracted in the mapping output layer, providing a stable feature basis for pile quality assessment. The overall process not only achieves spatial, temporal, and physical coordinated analysis of disturbance behavior, but also ensures high stability and verification of the training process, and supports model deployment in actual pile quality assessment scenarios.

[0045] Specifically, the soil property data include the change value of water content, the pore fluctuation ratio, the medium resistance change rate value, the micro-strain rate disturbance amplitude, and the seepage disturbance index. The specific steps for analyzing the pile foundation response adaptability index of each time period of the set cement mixing pile are as follows: the soil property data of each time period of each soil layer of the set cement mixing pile are evaluated and analyzed to obtain the soil disturbance evaluation set of the corresponding time period, including the pile foundation structure disturbance accommodation index and the pile foundation seepage stress buffer index; based on the soil disturbance evaluation set of each time period of each soil layer of the set cement mixing pile, the pile foundation response adaptability index of the corresponding time period is analyzed (that is, the pile foundation structure disturbance accommodation index and the pile foundation seepage stress buffer index of each time period of each soil layer are weighted summed, and the weighted average processing is performed based on the weighted summation processing result, and the result obtained is the pile foundation response adaptability index of each time period).

[0046] The moisture content change value is the intensity of the change in the moisture state of the soil layer during the period. A sharp fluctuation in moisture content will lead to an uneven water-cement ratio of the slurry, causing dry or wet areas, thereby forming inconsistent strength or shear-weak areas. The method for obtaining it is: pre-embed a TDR (time domain reflectometry) moisture sensor in the middle of each target soil layer, continuously obtain the moisture content value at each time point during the period, and perform variance processing. The result is the moisture content change value.

[0047] The pore fluctuation ratio is the ratio of the maximum to minimum porosity of the soil layer during the period. It is used to reveal the compactness or looseness of the soil during the disturbance process. High pore fluctuation indicates an unstable pore structure, which can easily lead to difficulty in fully consolidating the slurry, thereby affecting the density and bearing capacity of the pile. The method for obtaining it is as follows: γ-ray volume density sensors are embedded at the upper and lower boundaries and center of each soil layer, and TDR moisture sensors are deployed to obtain real-time moisture content. Through the simultaneous monitoring of density and moisture content, the dry density of the soil layer at different times is estimated, that is, density / (1+moisture content). The porosity value of the soil layer is calculated in combination with the standard soil particle density value. At each time point in the period, the maximum and minimum porosity values of the soil layer are extracted, and the ratio between the two is calculated as the pore fluctuation ratio for that period.

[0048] The dielectric resistivity change rate value is the rate of change of the soil layer resistivity during the period, reflecting the slurry solidification state and the dynamic characteristics of ion migration at the soil-water-slurry interface. Drastic changes in resistivity are often related to changes in water content or uneven slurry permeability, affecting solidification uniformity and early strength. Its acquisition method is as follows: a four-electrode resistivity measurement probe is placed in the middle of each set soil layer, a standard pulse current is injected into the soil layer, and the resistivity values of the soil layers between the probes at each time point are recorded in real time to form a continuous time series data sequence. Time difference analysis is then performed on this sequence, and the result obtained is the dielectric resistivity change rate value.

[0049] The microstrain rate disturbance amplitude is the rate of change of the microscopic deformation of the soil layer during the period when the disturbance occurs. It is used to reveal the internal shear or stress redistribution. The abnormal increase of microstrain indicates that the soil layer may have initial shear rupture or expansion of the damage zone, affecting the overall stability of the pile. The method for obtaining it is as follows: distributed optical fiber strain gauges or high-sensitivity resistive strain gauges are buried in the upper and lower boundary areas of each target soil layer, and the microstrain values in the axial and radial directions are continuously recorded. During the period, the microstrain time series data obtained are processed by first-order variation to obtain the disturbance rate, and then the disturbance rate is subjected to statistical fluctuation analysis to form the microstrain rate disturbance amplitude.

[0050] The seepage disturbance index is the degree of spatial variation and directional consistency of seepage parameters in the soil layer during the period, which is used to reflect the transient disturbance intensity of seepage. Severe seepage disturbance can easily lead to abnormal pore connectivity, uneven slurry infiltration and pore water pressure imbalance, affecting the solidification quality. The acquisition method is as follows: in each soil layer, micro pore water pressure sensors and micro seepage velocity sensors are pre-buried at the upper boundary, middle and lower boundary of the soil layer, respectively. The pore pressure sensor is used to obtain the pore water pressure value of the position at each time point within the set period, and the seepage velocity sensor is used to synchronously record the seepage velocity direction data of the corresponding position. The direction data represents the angle between the seepage velocity vector and the vertical direction in the form of polar angles. The standard deviation of the pore water pressure value sequence at each time point in the period of each position is calculated, and the pore pressure fluctuation amplitude of the position is calculated; then the pore pressure fluctuation amplitude of all sensor positions is calculated by the difference between the maximum and minimum values. The result obtained is the pore pressure disturbance amplitude index. The variance of the seepage velocity direction data recorded by all seepage velocity sensors is calculated at each time point to evaluate the consistency of the seepage direction inside the soil layer at that time point. The variance values of the seepage velocity direction at all time points are further averaged, and the obtained value is the seepage direction disturbance index. At each sampling time point, the vector field consistency analysis of the seepage velocity direction change between the sensor points inside the soil layer is performed. The specific method is to calculate the average cosine value of the angle between the seepage velocity vectors at different positions. This value is used to indicate the consistency degree of the seepage path, and the value range is 0 to 1. The larger the value, the more coherent the seepage path. The consistency index is averaged over the entire time period, and its complement (i.e., 1 minus the average value) is used as the seepage path consistency destruction index. Finally, the above three indicators are integrated into one by weighted summation to form the seepage disturbance index within the set time period.

[0051] The specific steps for obtaining the soil disturbance assessment set for each soil layer and each time period of the set cement mixing pile are as follows: based on the Z-score standardization, a weighted analysis is performed on the pore fluctuation ratio and microstrain rate disturbance amplitude of each soil layer and each time period of the set cement mixing pile (i.e., the pore fluctuation ratio and microstrain rate disturbance amplitude are first standardized based on the Z-score standardization, and then weighted based on the standardized processing results), and the pile foundation structure disturbance accommodation index of the corresponding time period is obtained; based on the Z-score standardization, a weighted analysis is performed on the water content change value, medium resistance change rate value, and seepage disturbance index of each soil layer and each time period of the set cement mixing pile (i.e., the water content change value, medium resistance change rate value, and seepage disturbance index are first standardized based on the Z-score standardization, and then weighted based on the standardized processing results), and the pile foundation seepage stress buffer index of the corresponding time period is obtained.

[0052] In this implementation plan, a complete pile foundation response adaptability assessment path is constructed by introducing soil property parameters with physical significance and temporal dynamic characteristics. The water content change value, pore fluctuation ratio, and dielectric resistivity change rate value can reflect the hydraulic state, structural density, and electrical property changes of the soil layer during the disturbance process in real time, thereby enhancing the ability to capture microscopic physical changes caused by construction disturbances. Secondly, the microstrain rate disturbance amplitude and seepage disturbance index reveal potential damage signs such as shear strain accumulation, pore pressure disturbance, and seepage direction dislocation, making the pile foundation response modeling not only limited to mechanical results, but also focusing on its formation mechanism. Finally, through Z-score standardization and weighted fusion operations, the assessment of the pile foundation structure's disturbance tolerance capacity and seepage stress buffering capacity is ensured to be consistent, comparable, and adjustable.

[0053] Specifically, the construction parameter set includes the mixing ratio value, drilling speed value, mixing shaft rotation value, shotcrete pressure value, shotcrete return pressure value, rod torque fluctuation value, and grouting fluctuation frequency value. The specific steps for analyzing and setting the construction disturbance implantation index of each time period of the cement mixing pile are as follows: a comprehensive analysis is performed on the construction parameter set of each time period of the cement mixing pile to obtain the construction behavior evaluation set of the corresponding time period, including the grouting dynamic disturbance index and the drilling and mixing motion disturbance index; a comprehensive analysis is performed on the construction parameter set of each time period of the cement mixing pile to obtain the construction disturbance implantation index of the corresponding time period.

[0054] Among them, the stirring ratio value is the ratio between the mass of water injected into the slurry during the period and the mass of cement powder. The method of obtaining it is: a water flow meter and a cement weighing sensor are respectively set in the slurry preparation device, and the water mass value cumulatively output by the water pump during the period is integrated and the ratio is calculated with the cement powder mass value stored in the cement weighing sensor during the period to obtain the corresponding stirring ratio value.

[0055] The drilling speed value is the axial movement speed of the drill rod when it is pulled upward from the bottom during the pile construction process. It is obtained by obtaining the position value of this period through a laser displacement sensor and performing wallpaper processing with the duration of this period. The result is the drilling speed value.

[0056] The agitator shaft rotation value is the speed at which the pile body rotates around its axis during the pile forming process. It is obtained by obtaining the rotation speed at each time point through a Hall-type speed sensor and performing average processing. The result obtained is the agitator shaft rotation value.

[0057] The shotcrete pressure value is the outlet pressure when the slurry is sprayed out from the nozzle, which is used to reflect the slurry penetration ability and grouting strength. It is obtained by placing a pressure sensor at the end of the grouting pipe near the nozzle, obtaining the instantaneous pressure value at each time point, and taking the maximum value as the shotcrete pressure value.

[0058] The shotcrete return flow pressure value is the reverse pressure value generated by the slurry returning to the nozzle due to soil impedance or pore saturation. It is used to judge the penetration resistance and pore blockage risk of the grouting process. The acquisition method is as follows: a differential pressure sensor is set in the nozzle return pipe, and the pressure values of the main nozzle and the return port at each time point are collected respectively. The difference processing (the absolute value of the difference between the pressure values of the main nozzle and the return port) is performed, and the average processing result is used as the shotcrete return flow pressure value based on the difference processing result.

[0059] The rod torque fluctuation value is the dynamic variation amplitude of the torque applied to the drill rod during rotation. It is used to reveal the uneven resistance phenomenon during drilling and mixing. The acquisition method is as follows: a strain-type torque sensor is set at the drill rod connection section to collect the instantaneous torque value of the drill rod at each time point. The rod torque fluctuation value is obtained by performing standard deviation calculation on the torque sequence.

[0060] The grouting fluctuation frequency value is the pressure fluctuation frequency caused by unstable pumping, pressure interference and other factors during the injection process of slurry. It is used to characterize the slurry flow continuity and injection stability. It is obtained by setting a high-frequency sampling pressure sensor at the outlet of the main grouting pump, collecting the complete shotcrete pressure time series signal sequence, and using the fast Fourier transform (FFT) method to perform spectrum analysis on it, extracting the main frequency and its corresponding amplitude from the frequency domain. The extracted main frequency is the grouting fluctuation frequency value.

[0061] The specific steps for obtaining the construction behavior evaluation set of each time period of the set cement mixing pile are as follows: obtaining the environmental correction factor of each time period of the set cement mixing pile, and performing weighted analysis on the mixing ratio value, shotcrete pressure value, shotcrete return pressure value, and grouting fluctuation frequency value of the corresponding time period (i.e., first standardizing the environmental correction factor, mixing ratio value, shotcrete pressure value, shotcrete return pressure value, and grouting fluctuation frequency value based on Z-score standardization, and performing weighted processing based on the standardized processing results), to obtain the grouting dynamic disturbance index of the set corresponding time period; performing weighted analysis on the drilling speed value, mixing shaft rotation value, and rod torque fluctuation value of each time period of the set cement mixing pile based on Z-score standardization (i.e., first standardizing the drilling speed value, mixing shaft rotation value, and rod torque fluctuation value based on Z-score standardization, and performing weighted processing based on the standardized processing results), to obtain the drilling and mixing motion disturbance index of the corresponding time period.

[0062] The specific steps for obtaining the environmental correction factor for each period of cement mixing piles are as follows: The ambient temperature value of each time period of the set cement mixing pile is obtained (the temperature value of each time point in the time period is obtained by the temperature sensor, and the average value is processed to obtain the ambient temperature value) and the ambient humidity value (the humidity value of each time point in the time period is obtained by the temperature sensor, and the average value is processed to obtain the ambient humidity value) are obtained, and standardized. Based on the standardized processing result, weighted processing is performed. The result obtained is the environmental correction factor of each time period of the set cement mixing pile.

[0063] The specific formula for calculating the construction disturbance implantation index of a given cement mixing pile during a certain period of time is as follows: ;in, To set the construction disturbance implantation index of cement mixing piles in a certain period of time, To set the grouting dynamic disturbance index of a cement mixing pile at a certain period of time, is the grouting disturbance coefficient stored in the database, To set the disturbance index of drilling motion of cement mixing pile in a certain period of time, is the drilling disturbance coefficient stored in the database, is the interaction disturbance coefficient stored in the database.

[0064] What needs to be explained is that 、 、 It can be obtained through the following steps: using historical data, combined with the grouting dynamic disturbance index and the drilling and mixing motion disturbance index, to conduct statistical regression analysis, quantify the specific impact of each factor on the construction disturbance implantation index, and thus fit the initial weight value; secondly, using the disturbance response analysis method (that is, setting different value intervals for the above two variables, while keeping other parameters unchanged, gradually adjusting their values and calculating the change range of the construction disturbance implantation index to test the influence trend of each parameter disturbance on the output result), adjust the value range of each coefficient, observe its impact on the construction disturbance implantation evaluation results, and ensure the stability and rationality of the model.

[0065] The specific implementation example of calculating the construction disturbance implantation index of a certain period of cement mixing pile is as follows. The existing data are as follows: including the grouting dynamic disturbance index and the drilling and mixing motion disturbance index of 5 periods (randomly selected) of cement mixing pile, as shown in Table 1 and Figure 3-4 Table 1 Example of time series data of construction disturbance implantation index of cement mixing pile

[0066] Grouting disturbance coefficients stored in the database Approximately: 0.463; Drilling disturbance coefficient stored in the database Approximately: 0.526; Interaction perturbation coefficients stored in the database Approximately: 1.238; Substituting the data in Table 1 and the above coefficients into the specific formula for calculating the construction disturbance implantation index of a given cement mixing pile during a certain period of time, we obtain: The construction disturbance implantation index of the first period of cement mixing pile is set as ln(1+((0.237^0.463+0.263^0.526) / 2))×(1+tanh(1.238×√(0.237×0.263)))≈0.529; The construction disturbance implantation index of the second period of cement mixing pile is set as ln(1+((0.184^0.463+0.241^0.526) / 2))×(1+tanh(1.238×√(0.184×0.241)))≈0.481; The construction disturbance implantation index of the third period of cement mixing pile is set as ln(1+((0.316^0.463+0.289^0.526) / 2))×(1+tanh(1.238×√(0.316×0.289)))≈0.606; The construction disturbance implantation index of the fourth period of cement mixing pile is set as ln(1+((0.264^0.463+0.186^0.526) / 2))×(1+tanh(1.238×√(0.264×0.186)))≈0.496; The construction disturbance implantation index of the fifth period of cement mixing pile is set as ln(1+((0.158^0.463+0.238^0.526) / 2))×(1+tanh(1.238×√(0.158×0.238)))≈0.458.

[0067] In this implementation plan, by introducing a structural calculation method for the construction disturbance implantation index, a quantifiable evaluation path for the impact of construction behavior on pile disturbance is constructed. The system constructs the grouting dynamic disturbance index and the drilling motion disturbance index based on grouting parameters and drilling parameters respectively, and integrates them with preset weighting coefficients to form a unified index that comprehensively reflects the input path, intensity and direction characteristics of the construction disturbance. In terms of weight acquisition, the system uses statistical regression analysis of historical data to clarify the actual influence ratio of various disturbance sources under different construction conditions, and realizes the initial quantitative allocation of disturbance factors; then, the disturbance response analysis method is introduced to systematically test the response amplitude and trend change of each disturbance variable to the final index under different values, ensuring that the control range of each parameter coefficient on the output result is controllable and physically reasonable; finally, through weight adjustment, a systematic modeling of the stability and sensitivity of the construction disturbance process is formed, so that the construction disturbance implantation index can accurately reflect the disturbance intensity, thereby providing a traceable construction behavior characteristic indicator basis for pile quality prediction and risk identification.

[0068] Specifically, the specific steps for obtaining the pile quality index of the next time period of the set cement mixing pile are as follows: based on the pile-circumference disturbance evolution driving index, the pile foundation response adaptability index, and the construction disturbance implantation index of each time period of the set cement mixing pile, the pile body pile quality driving index of each time period of the set cement mixing pile is analyzed; and based on the pile body pile quality driving index of each time period of the set cement mixing pile, a trend evolution analysis is performed to obtain the pile body pile quality driving index of the next time period of the set cement mixing pile.

[0069] Among them, the specific steps of trend evolution analysis are as follows: set a time window (such as the last three time periods), perform sliding weighted average processing on the pile body pile quality driving index of each time period of the set cement mixing pile within the time window, and obtain the foundation pile body pile quality driving prediction index of the set cement mixing pile, and perform difference (the absolute value of the difference between the pile body pile quality driving index of adjacent time periods) and standard deviation processing on the pile body pile quality driving index of each time period of the set cement mixing pile within the time window to obtain the maximum difference of the pile body pile quality driving index, the minimum difference of the pile body pile quality driving index, and the standard deviation of the pile body pile quality driving index of the set cement mixing pile within the time window, and calculate the disturbance amplification factor of the set cement mixing pile within the time window, that is, (maximum difference of the pile body pile quality driving index - minimum difference of the pile body pile quality driving index) / standard deviation of the pile body pile quality driving index, and perform pile periphery disturbance evolution driving index and pile foundation response adaptability index of the current time period of the set cement mixing pile. , and the construction disturbance implantation index are comprehensively analyzed to obtain the proportional consistency factor and the negative disturbance adjustment factor, that is, the proportional consistency factor = exp(-|pile disturbance evolution driving index - pile foundation response adaptability index| -|pile foundation response adaptability index - construction disturbance implantation index| -|pile disturbance evolution driving index - construction disturbance implantation index|), and the negative disturbance adjustment factor = (|pile disturbance evolution driving index - pile foundation response adaptability index| +|pile foundation response adaptability index - construction disturbance implantation index| +|pile disturbance evolution driving index - construction disturbance implantation index|) / 3. The foundation pile quality driving prediction index, disturbance amplification factor, proportional consistency factor, and negative disturbance adjustment factor of the cement mixing pile are comprehensively analyzed, that is, the foundation pile quality driving prediction index × (1 + disturbance amplification factor - disturbance negative adjustment factor) × proportional consistency factor, to obtain the predicted pile quality driving index of the cement mixing pile, which is marked as the pile quality driving index for the next period.

[0070] The specific steps for calculating the pile quality driving index of a given cement mixing pile at a certain time period are as follows: ;in, To set the pile quality driving index of cement mixing pile in a certain period of time, To set the driving index of the disturbance evolution around the cement mixing pile in a certain period of time, To set the pile foundation response adaptability index of cement mixing pile in a certain period of time, is the superposition coefficient stored in the database, is the coefficient of variation stored in the database, To set the construction disturbance implantation index of cement mixing piles in a certain period of time, is the construction coefficient stored in the database.

[0071] What needs to be explained is that 、 、 You can obtain it by following the steps below: Overlay coefficients stored in the database The acquisition method is as follows: select no less than 50 cement mixing piles in the set area that have been completed and have three types of data fields: disturbance evolution driving index (unit is dimensionless), pile foundation response adaptability index (unit is dimensionless) and pile quality measured grade (grade values are A, B, and C and assigned values of 1.0, 0.75, and 0.5) as the calculation objects, and perform point-by-point weighted average of the disturbance evolution driving index and pile foundation response adaptability index of each pile in each period during the construction period (for example, the disturbance evolution driving index and the pile foundation response adaptability index are multiplied by different weights and added), and then perform difference analysis with the measured grade value; the weight ratio corresponding to the minimum weighted average value is taken as and stored in the database with three decimal places.

[0072] Coefficient of variation stored in the database , which is obtained by taking the absolute value of the difference between the disturbance evolution driving index and the pile foundation response adaptability index of each pile in each period during the construction cycle, counting the difference data of all periods, calculating the maximum and minimum values, and recording the standard deviation of the full-cycle difference sequence; the difference degree of the pile is defined as = maximum value - minimum value / standard deviation, the obtained value reflects the inconsistent fluctuation amplitude between disturbance and response, and the unit is dimensionless ratio; after verification, it is recorded as Enter the database for standard fields.

[0073] Construction coefficients stored in the database The acquisition method is as follows: the construction disturbance implantation index of each pile in each period during the construction cycle is compared with the fluctuation value of the construction parameters in the corresponding period, including the standard deviation of the shotcrete pressure, the torque fluctuation value, and the water-cement ratio fluctuation frequency. The three fluctuation values are normalized by Z-score and then the weighted average is calculated as the construction fluctuation intensity value; the construction fluctuation intensity value of each period and the corresponding construction disturbance implantation index are used to form a two-dimensional data pair, and the average ratio of all data pairs in the whole cycle is calculated, that is, the construction disturbance implantation index / construction fluctuation intensity value, and the result is normalized to obtain .

[0074] In this implementation plan, by constructing a trend evolution analysis and multi-factor adjustment mechanism, the prediction accuracy and dynamic adaptability of the pile quality driving index are significantly improved. The system introduces a sliding time window strategy in the time dimension, and performs weighted mean, fluctuation difference and standard deviation analysis based on the pile quality driving index of the past three time periods to extract the disturbance amplification factor, which is used to quantify the severity of recent disturbances and the instability of the prediction results. At the same time, by calculating the difference relationship between the pile perimeter disturbance, pile foundation response and construction disturbance, a proportional consistency factor and a disturbance negative adjustment factor are constructed to reflect the response coordination and potential interference contradictions between the interference sources, respectively. Finally, the above three quantitative results are jointly corrected and calculated with the basic prediction value to realize the trend prediction of the pile quality in the next time period. This not only integrates the current disturbance status and historical trend characteristics, but also has dynamic feedback capabilities, and can effectively identify quality downside risks and disturbance structural anomalies, thereby providing a continuous and high-precision prediction basis for subsequent construction parameter adjustments and pile quality assurance.

[0075] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0076] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. A real-time prediction system for cement mixing pile quality, characterized in that: include: The pile disturbance evolution analysis module is used to continuously acquire geological 3D point cloud data for each period of a set cement mixing pile, and conduct a comprehensive analysis based on the pre-trained disturbance recognition model to obtain the pile disturbance evolution driving index for the corresponding period; The construction and soil analysis module is used to obtain the construction parameter set for each period of cement mixing piles and the soil property data for each period of several soil layers, and analyze the construction disturbance implantation index and pile foundation response adaptability index of the corresponding period respectively; The pile quality prediction and analysis module is used to predict and analyze the pile disturbance evolution driving index, pile foundation response adaptability index, and construction disturbance implantation index of each period of the set cement mixing pile, and obtain the pile quality driving index of the next period of the set cement mixing pile; The pile quality feedback module is used to perform quality feedback processing on the set cement mixing pile based on the pile quality driving index of the next period.

2. The real-time prediction system for cement mixing pile quality according to claim 1 is characterized in that: The geological three-dimensional point cloud data includes the voxel value, three-dimensional coordinates and corresponding reflection intensity value of each voxel point. The specific steps for obtaining the geological disturbance index for each time period of the cement mixing pile are as follows: The geological 3D point cloud data of each period of cement mixing piles is input into the pre-trained disturbance recognition model for disturbance feature analysis, and the disturbance evaluation set of the corresponding period is obtained, including the disturbance convergence potential energy index, stress transfer tensor index, shear aggregation index, and plastic boundary penetration index. Based on the disturbance assessment set of each period of cement mixing pile, the geological disturbance index of the corresponding period is analyzed.

3. The real-time prediction system for cement mixing pile quality according to claim 2 is characterized in that: The disturbance recognition model is specifically a graph-structured neural network, including a point cloud construction layer, a neighborhood disturbance encoding layer, a time encoding layer, a physical constraint embedding layer, and a mapping output layer. The specific steps for obtaining the disturbance assessment set for each time period of the cement mixing pile are as follows: In the point cloud graph construction layer of the graph structure neural network, the geological three-dimensional point cloud data of each period of the cement mixing pile is received and the graph construction process is performed; In the neighborhood perturbation coding layer of the graph-structured neural network, neighborhood feature coding is performed on the geological 3D point cloud data of each period of the cement mixing pile after graph construction, and the perturbation feature vector set of the corresponding period is analyzed. In the time encoding layer of the graph structure neural network, the disturbance feature vector set of each period of the cement mixing pile is processed with time evolution; In the physical constraint embedding layer of the graph structure neural network, the disturbance feature vector set of each period of the set cement mixing pile after time evolution is subjected to structural constraint mapping processing to obtain the physical consistency disturbance representation vector set of the corresponding period; In the mapping output layer of the graph structure neural network, the physical consistency disturbance representation vector set of each period of the cement mixing pile is subjected to feature mapping processing to obtain the disturbance directional potential energy index, stress transfer tensor index, shear aggregation index, and plastic boundary penetration index of the corresponding period.

4. The real-time prediction system for cement mixing pile quality according to claim 2 is characterized in that: The specific formula for calculating the driving index of the disturbance evolution around a cement mixing pile in a certain period of time is as follows: ； in, 、 、 、 、 The following are the pile disturbance evolution driving index, disturbance convergence potential energy index, stress transfer tensor index, shear aggregation index, and plastic boundary penetration index of a certain period of time of the cement mixing pile. 、 、 、 、 They are the convergence coefficient, transfer coefficient, aggregation coefficient, penetration coefficient and synergy coefficient stored in the database respectively.

5. The real-time prediction system for cement mixing pile quality according to claim 1 is characterized in that: The soil property data includes the water content change value, the pore fluctuation ratio, the medium resistance change rate value, the micro-strain rate disturbance amplitude, and the seepage disturbance index. The specific steps for analyzing and setting the pile foundation response adaptability index for each soil layer and each time period of the cement mixing pile are as follows: The soil property data of each soil layer and each time period of the cement mixing pile are evaluated and analyzed to obtain the soil disturbance assessment set of the corresponding time period, including the pile foundation structure disturbance accommodation index and the pile foundation seepage stress buffer index; Based on the soil disturbance assessment set of each soil layer and each period of the cement mixing pile, the pile foundation response adaptability index of the corresponding period is analyzed.

6. The real-time prediction system for cement mixing pile quality according to claim 5 is characterized in that: The specific steps to obtain the soil disturbance assessment set for each soil layer and each time period of the set cement mixing pile are as follows: Based on the Z-score standardization, the pore fluctuation ratio and micro-strain rate disturbance amplitude of each soil layer in each period of the cement mixing pile are weighted and analyzed to obtain the disturbance tolerance index of the pile foundation structure in the corresponding period. Based on the Z-score standardization, a weighted analysis was performed on the water content change value, medium resistance change rate value, and seepage disturbance index of each soil layer of the cement mixing pile in each period, and the pile foundation seepage stress buffer index of the corresponding period was obtained.

7. The real-time prediction system for cement mixing pile quality according to claim 1 is characterized in that: The construction parameter set includes a mixing ratio value, a drilling speed value, a mixing shaft rotation value, a shotcrete pressure value, a shotcrete return pressure value, a rod torque fluctuation value, and a grouting fluctuation frequency value. The specific steps for analyzing and setting the construction disturbance implantation index for each period of the cement mixing pile are as follows: A comprehensive analysis is conducted on the construction parameter set of each period of cement mixing pile construction to obtain the construction behavior evaluation set of the corresponding period, including the grouting dynamic disturbance index and the drilling and mixing motion disturbance index. A comprehensive analysis is conducted on the construction parameter set of each period of cement mixing pile construction to obtain the construction disturbance implantation index of the corresponding period.

8. The real-time prediction system for cement mixing pile quality according to claim 7 is characterized in that: The specific steps to obtain the construction behavior evaluation set for each period of the set cement mixing pile are as follows: Obtain the environmental correction factor for each period of the set cement mixing pile, and perform weighted analysis on the mixing ratio value, shotcrete pressure value, shotcrete return pressure value, and grouting fluctuation frequency value of the corresponding period to obtain the grouting dynamic disturbance index of the set corresponding period; Based on Z-score standardization, a weighted analysis is performed on the drilling speed value, mixing shaft rotation value, and rod torque fluctuation value of each period of the cement mixing pile to obtain the drilling and mixing motion disturbance index of the corresponding period.

9. The real-time prediction system for cement mixing pile quality according to claim 1 is characterized in that: The specific steps for obtaining the pile quality index of the next period of a set cement mixing pile are as follows: Based on the pile disturbance evolution driving index, pile foundation response adaptability index, and construction disturbance implantation index of each period of cement mixing pile, the pile quality driving index of the corresponding period is analyzed. A trend evolution analysis is performed based on the pile quality driving index of each period of the set cement mixing pile to obtain the pile quality driving index of the next period of the set cement mixing pile.

10. The real-time prediction system for cement mixing pile quality according to claim 9 is characterized in that: The specific steps for calculating the pile quality driving index of a given cement mixing pile at a certain time period are as follows: ； in, 、 、 、 The following are the pile quality driving index, pile periphery disturbance evolution driving index, pile foundation response adaptability index, and construction disturbance implantation index of the cement mixing pile in a certain period of time. 、 、 They are the superposition coefficient, difference coefficient and construction coefficient stored in the database respectively.

Images