Bearing capacity determination method based on special-shaped pile bearing character analysis

Abstract

The invention relates to the field of civil engineering, and provides a bearing capacity determination method based on special-shaped pile bearing character analysis, which comprises the following steps: firstly, obtaining stress distribution diagrams of two types of pile bodies under the same load, extracting spatial characteristics of the stress distribution diagrams, and calculating differential characteristics; and then characteristic significant enhancement is used for conducting enhanced recognition on a specific stress concentration area of the special-shaped pile in the difference graph so as to quantify nonlinear mechanical response, then space significant differences are mapped into dynamic correction coefficients of friction resistance and end resistance through characteristic decoding, and finally correction of the ultimate bearing capacity of the special-shaped pile is completed in combination with bearing capacity data of a standard pile. Therefore, the frictional resistance distribution dynamic change and end resistance multi-peak concentration phenomena caused by geometric asymmetry of the pile body can be accurately captured, meanwhile, the correction parameter optimization efficiency under the complex working condition can be remarkably improved, and the core demands of new energy pile foundation engineering for material saving and high reliability can be met.

Description

Technical Field

[0001] The present application relates to the field of civil engineering, and more specifically, to a method for determining the bearing capacity based on the analysis of the bearing characteristics of special-shaped piles. Background Art

[0002] With the promotion of new energy infrastructure construction, the demand for pile foundations in projects such as photovoltaic power stations has increased dramatically. Traditional pile foundations are often designed to meet bearing capacity requirements by increasing pile length or diameter. This not only leads to material waste and increased construction costs, but also requires the deployment of hundreds of thousands of piles in complex scenarios such as soft soil, posing a significant conflict between economic efficiency and environmental protection. In these projects, special-shaped piles (such as X-shaped, H-shaped, expanded-base piles, and rectangular piles with concave corners) demonstrate significant advantages in complex geological engineering due to their unique geometric structure. For example, they improve bearing capacity by increasing the pile-soil contact area or optimizing stress transfer paths. They are widely used in soft soil foundations, bridges with high bearing capacity requirements, and high-rise buildings.

[0003] However, accurately determining their bearing capacity remains challenging. While theoretical and experimental research on special-shaped piles has yielded some promising results, traditional methods are limited by simplified assumptions about the mechanical behavior of the pile-soil system. For example, the side friction and end resistance of special-shaped piles are simplified to linear correction coefficients for standard piles, ignoring the spatial heterogeneity of stress distribution caused by cross-sectional changes or geometric asymmetry. Specifically, when a special-shaped pile is loaded, the distribution range of its side friction may dynamically change with the pile profile, and the end resistance is more likely to form multi-peak stress concentrations in the localized expanded diameter region. Traditional models, lacking a detailed analysis of the global stress spatial characteristics of the pile, struggle to quantify this nonlinear mechanical response, leading to inaccurate correction coefficient values. Furthermore, existing numerical simulation techniques, limited by computational efficiency and size effects, cannot effectively validate bearing capacity calculation models under complex working conditions, further exacerbating the difficulty of systematically optimizing design parameters. This disconnect between theory and engineering practice often forces conservative strategies in the design of special-shaped piles, which not only fails to fully realize their material-saving potential but also poses safety risks under extreme loads.

[0004] Therefore, an optimized bearing capacity determination scheme based on the bearing behavior analysis of special-shaped piles is desired. Summary of the Invention

[0005] In order to solve the above technical problems, the present application is proposed. The embodiments of the present application provide a method for determining the bearing capacity based on the analysis of the bearing characteristics of special-shaped piles.

[0006] According to one aspect of the present application, a method for determining the bearing capacity based on the analysis of the bearing characteristics of special-shaped piles is provided, comprising: Obtaining stress distribution diagrams of standard piles and special-shaped piles under predetermined loads; Extracting stress distribution spatial characteristics from the stress distribution diagram of the standard pile and the stress distribution diagram of the special-shaped pile to obtain a stress distribution characteristic diagram of the standard pile and a stress distribution characteristic diagram of the special-shaped pile; Performing spatial differential feature enhancement on the stress distribution characteristic map of the standard pile and the stress distribution characteristic map of the special-shaped pile to obtain a stress distribution spatially significant differential feature map, including: calculating a stress distribution spatially significant differential feature map between the stress distribution characteristic map of the standard pile and the stress distribution characteristic map of the special-shaped pile; performing regional feature association receptive field saliency on the stress distribution spatially significant differential feature map to obtain the stress distribution spatially significant differential feature map; Performing feature decoding on the stress distribution spatial significant difference feature map to obtain a side friction correction coefficient and an end resistance correction coefficient; Obtaining the total pile side friction resistance and the total pile end resistance of the standard pile under the predetermined load; Based on the side friction correction coefficient and the end resistance correction coefficient, the total pile side friction and the total pile end resistance are corrected to obtain the ultimate bearing capacity of the special-shaped pile under a predetermined load.

[0007] Compared with the prior art, the bearing capacity determination method based on the bearing characteristics analysis of special-shaped piles provided by this application first obtains the stress distribution diagram of two types of piles under the same load, extracts their spatial features and calculates the differential features, then uses the feature significant enhancement to strengthen the identification of the stress concentration area unique to the special-shaped piles in the differential diagram to quantify the nonlinear mechanical response, and then maps the spatial significant differences into dynamic correction coefficients of friction resistance and end resistance through feature decoding, and finally completes the correction of the ultimate bearing capacity of the special-shaped piles in combination with the bearing capacity data of the standard piles. In this way, the dynamic changes in the friction resistance distribution caused by the geometric asymmetry of the pile body and the multi-peak concentration phenomenon of the end resistance can be accurately captured, and at the same time, the optimization efficiency of the correction parameters under complex working conditions can be significantly improved, which is conducive to meeting the core demands of new energy pile foundation projects for material saving and high reliability. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] The above and other purposes, features, and advantages of the present application will become more apparent through a more detailed description of the embodiments of the present application in conjunction with the accompanying drawings. The accompanying drawings are intended to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the drawings, the same reference numerals generally represent the same components or steps.

[0009] Figure 1 Flowchart of a method for determining bearing capacity based on bearing property analysis of special-shaped piles according to an embodiment of the present application.

[0010] Figure 2This is a flowchart of step S3 in the method for determining the bearing capacity based on the analysis of the bearing characteristics of special-shaped piles according to an embodiment of the present application.

[0011] Figure 3 This is a flowchart of step S32 in the method for determining the bearing capacity based on the analysis of the bearing characteristics of special-shaped piles according to an embodiment of the present application. DETAILED DESCRIPTION

[0012] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described herein.

[0013] With the advancement of new energy infrastructure construction, demand for pile foundations in projects such as photovoltaic power plants has increased significantly. Traditional pile foundation design typically meets bearing capacity requirements by increasing pile length or diameter, but this can lead to material waste and increased costs. This is particularly true in complex scenarios such as soft soil, where large-scale pile deployment is required, highlighting the conflict between economic and environmental considerations. Special-shaped piles (such as X-shaped, H-shaped, and expanded-base piles) demonstrate performance advantages in soft soil, high-bearing bridges, and buildings by increasing the pile-soil contact area and optimizing stress transfer paths due to their unique geometry. However, accurate calculation of their bearing capacity remains a technical bottleneck: existing theories simplify pile-soil mechanical behavior and fail to accurately characterize the complex stress characteristics of special-shaped piles. Traditional methods use linear correction coefficients to estimate lateral friction and end resistance, ignoring spatial stress heterogeneity caused by sudden cross-sectional changes—for example, the impact of dynamic changes in pile profile on friction distribution or the multi-peak stress concentration phenomenon caused by expanded diameter regions. Furthermore, numerical simulation techniques, limited by computational efficiency and size effects, struggle to effectively validate bearing capacity models under complex conditions, hindering the systematic optimization of design parameters. The disconnect between theory and practice forces engineering design to adopt a conservative strategy, which not only restricts the material-saving potential of special-shaped piles, but also may cause safety hazards under extreme loads.

[0014] In response to the above technical issues, the technical concept of this application is to dynamically generate correction coefficients to achieve accurate bearing capacity prediction by comparing the spatial characteristics of stress distribution between special-shaped piles and standard piles. The specific processing process is as follows: first, the stress distribution diagrams of the two types of piles under the same load are obtained, their spatial characteristics are extracted, and the differential characteristics are calculated; then, the stress concentration areas unique to special-shaped piles (such as the sudden change in the pile side profile and the expanded diameter end) in the differential diagram are enhanced and identified to quantify the nonlinear mechanical response; then, through feature decoding, the spatially significant differences are mapped into dynamic correction coefficients for friction and end resistance, and finally, combined with the bearing capacity data of standard piles, the ultimate bearing capacity correction of special-shaped piles is completed. This method directly addresses the defect of traditional models that ignore the spatial heterogeneity of stress in special-shaped piles. Through spatial feature differentiation and significance processing, it accurately captures the dynamic changes in friction distribution and the multi-peak concentration of end resistance caused by the geometric asymmetry of the pile body, breaking through the crudeness of traditional linear correction. At the same time, the coefficient generation mechanism based on feature decoding replaces the inefficient numerical simulation verification process, significantly improving the efficiency of correction parameter optimization under complex working conditions. It not only avoids material redundancy caused by conservative design, but also reduces safety risks under extreme loads through accurate bearing capacity prediction, which is in line with the core demands of new energy pile foundation projects for material saving and high reliability.

[0015] Figure 1 FIG. 1 is a flow chart of a method for determining bearing capacity based on analysis of bearing characteristics of special-shaped piles according to an embodiment of the present application. Figure 1 As shown, according to the embodiment of the present application, the bearing capacity determination method based on the bearing property analysis of the special-shaped pile includes: S1, obtaining the stress distribution diagram of the standard pile and the stress distribution diagram of the special-shaped pile under a predetermined load; S2, extracting the stress distribution spatial characteristics from the stress distribution diagram of the standard pile and the stress distribution diagram of the special-shaped pile to obtain the stress distribution characteristic diagram of the standard pile and the stress distribution characteristic diagram of the special-shaped pile; S3, performing spatial differential feature enhancement on the stress distribution characteristic diagram of the standard pile and the stress distribution characteristic diagram of the special-shaped pile to obtain a stress distribution spatial significant differential characteristic diagram; S4, performing feature decoding on the stress distribution spatial significant differential characteristic diagram to obtain a side friction resistance correction coefficient and an end resistance correction coefficient; S5, obtaining the total pile side friction resistance and the total pile end resistance of the standard pile under the predetermined load; S6, based on the side friction resistance correction coefficient and the end resistance correction coefficient, correcting the total pile side friction resistance and the total pile end resistance to obtain the ultimate bearing capacity of the special-shaped pile under the predetermined load.

[0016] In step S1, stress distribution diagrams for standard and special-shaped piles under a predetermined load are obtained. It should be understood that both stress distribution diagrams for standard and special-shaped piles under a predetermined load contain information on the spatial distribution of stresses between the pile and the surrounding soil under load. Specifically, they cover stress types (normal stress, shear stress, etc.), stress magnitudes, stress directions, and stress gradient variations at different locations (e.g., at different depths in the pile body and in different regions of the soil surrounding the pile). The stress distribution diagram for standard piles reflects the stress transfer patterns of conventional cross-section piles under load. Its stress distribution is generally symmetrical and regular, serving as a benchmark reference. However, the stress distribution diagram for special-shaped piles reflects the heterogeneity of stress distribution caused by cross-sectional geometric features (e.g., concave corners, expanded diameters, and multiple branches). Stress concentration may occur at locations with sudden changes in cross-sectional area (e.g., concave corner edges and expanded diameter regions), or dynamic adjustments in the stress distribution range may occur due to changes in the pile-soil contact area. For example, the side friction distribution of the pile may exhibit non-uniform characteristics as the pile body contour fluctuates, and the end resistance may exhibit multi-peak distributions in local expanded diameter regions. In summary, by obtaining the stress distribution information of the two types of piles under the same loading conditions and conducting an in-depth analysis, we can accurately capture the unique stress distribution of the special-shaped piles due to differences in their geometric structures. Then, by extracting the differences in the spatial characteristics of the stress distribution between the two, we can quantify the differences in the mechanical responses of the special-shaped piles relative to the standard piles, providing a direct mechanical basis for correcting the pile side friction and end resistance, so that the bearing capacity calculation can more realistically reflect the actual stress state of the special-shaped piles.

[0017] In step S2, stress distribution spatial features are extracted from the stress distribution diagrams of the standard pile and the special-shaped pile to obtain stress distribution characteristic diagrams of the standard pile and the special-shaped pile. Specifically, in an embodiment of the present application, step S2 includes: processing the stress distribution diagrams of the standard pile and the special-shaped pile using a stress distribution spatial feature detector based on a twin convolutional neural network to obtain the stress distribution characteristic diagrams of the standard pile and the special-shaped pile. Accordingly, considering that different types of piles (standard piles and special-shaped piles) have different stress distributions under the same load, these differences reflect their respective bearing characteristics. For example, sudden changes in the cross-section of a special-shaped pile (such as X-shaped ribs, expanded bottom ends, etc.) can lead to asymmetric expansion of the stress transfer path at the pile-soil contact interface. For example, the pile side friction resistance exhibits a gradient distribution along the special-shaped contour, and the end resistance forms a multi-peak concentration in the expanded diameter area. Therefore, in order to highlight these essential characteristics and eliminate the interference of irrelevant or secondary information, it is helpful to more clearly compare and analyze the differences in mechanical properties between the two types of piles, laying the foundation for further research on the bearing characteristics of special-shaped piles. In the technical solution of this application, stress distribution spatial features are extracted from the stress distribution maps of the standard pile and the stress distribution maps of the special-shaped pile to obtain the stress distribution feature maps of the standard pile and the stress distribution feature maps of the special-shaped pile. In particular, in one example of this application, a stress distribution spatial feature detector based on a twin convolutional neural network is used to process the stress distribution maps of the standard pile and the special-shaped pile to obtain the stress distribution feature maps of the standard pile and the special-shaped pile. It is worth mentioning that the stress distribution spatial feature detector based on the twin convolutional neural network is used to process the stress distribution maps of the standard pile and the special-shaped pile. Its core lies in the simultaneous extraction of high-dimensional spatial features of the stress distribution of the two types of piles through a dual-channel network architecture with shared weights. The dual-input characteristic of the twin network ensures that the feature maps of the standard pile and the special-shaped pile are encoded in the same latent space, thereby preserving the comparability of the two in terms of spatial scale and texture pattern. For example, for the stress distribution map of the expanded base pile, the network can retain the characteristics of the annular high stress area in the end diameter expansion area (through the large receptive field convolution layer) and the friction resistance distribution details of the pile body's gradient cross-section (through the small-scale convolution layer) in the feature map through the sliding perception of multi-layer convolution kernels, thereby providing key data for subsequent difference feature analysis.

[0018] The following is a detailed description of a specific implementation process of “processing the stress distribution map of the standard pile and the stress distribution map of the special-shaped pile using a stress distribution spatial characterizer based on a twin convolutional neural network to obtain the stress distribution characteristic map of the standard pile and the stress distribution characteristic map of the special-shaped pile”: First, data input preparation is required to construct standardized input data suitable for the twin convolutional neural network. Standardization of stress distribution maps requires converting stress information in physical space into a multi-channel numerical matrix. Specifically, for the stress distribution of standard and irregular piles under predetermined loads, information such as the stress type (normal stress, shear stress, etc.), stress magnitude, direction, and gradient variation in the pile and surrounding soil must be quantified and encoded. For example, the stress parameters at each spatial location (establishing a coordinate system with the pile top as the origin, the pile length as the vertical axis, and the cross-section as the horizontal axis) can be mapped to pixels in a matrix. Different stress types correspond to different channels in the matrix, forming a standardized data format (H×W×C) with spatial dimensions (height H, width W) and characteristic channel dimensions (C). Based on this, paired data samples are constructed: the stress distribution map of the standard pile and the stress distribution map of the irregular pile under the same loading conditions. The spatial coordinates of the two pairs must be strictly aligned, ensuring, for example, a one-to-one correspondence between the coordinate points at each depth of the pile and along the radial direction of the cross-section. This provides a basis for geometric consistency for subsequent feature comparison and analysis.

[0019] After completing data preparation, the architecture design phase of the twin convolutional neural network began. This network employs a dual-channel structure with shared weights, consisting of two completely symmetrical branches, one for processing the stress distribution maps of standard piles and the other for processing the stress distribution maps of special-shaped piles. This shared weight design ensures that both branches process the input data using the same feature extraction rules, placing the extracted features in the same semantic space, facilitating subsequent differential analysis. The network architecture is divided into multiple layers of feature extraction modules, each of which is responsible for different feature extraction tasks. The bottom convolutional layer uses a small convolution kernel (e.g., 3×3) to capture the basic spatial features of the stress distribution through a shallow network, such as the stress gradient boundary at the sudden change in the pile cross-section and the subtle texture changes at the local stress concentration point. These features reflect the edge information and local details of the stress distribution. The middle convolutional layer gradually increases the convolution kernel size (e.g., 5×5) and the network depth to extract cross-regional stress distribution patterns, such as the difference between the high-stress band at the edge of the rib of a special-shaped pile and the uniform stress area on the smooth side of a standard pile. Through the processing of the middle-layer network, regional differences and patterned features of the stress distribution can be identified. The high-level convolutional layer uses a large convolution kernel or global pooling operation to focus on capturing the global structural features of the stress distribution, such as the annular stress diffusion pattern in the expanded diameter area at the end of the expanded pile and the axisymmetric uniform distribution of the end resistance of the standard pile. These features reflect the transmission law and overall distribution of stress throughout the pile body. After each convolution operation, a nonlinear activation function (such as ReLU) is introduced to enhance the network's ability to express nonlinear stress features and avoid the limitations of feature expression caused by linear operations. At the same time, average pooling or maximum pooling operations are used to downsample the feature map, reduce the spatial dimension, and extract core features with translation invariance, thereby reducing the complexity of subsequent calculations while retaining key information.

[0020] The feature extraction process strictly adheres to a dual-channel parallel processing mechanism. The standard pile branch receives the standardized standard pile stress distribution map and processes it layer by layer through the bottom, middle, and high convolutional layers, extracting low-level features (stress edges, local gradients), intermediate features (regional stress patterns), and high-level features (global stress structure). For example, in the bottom convolutional layer, the network identifies stress edges at different depths in the pile, such as the shear stress boundary at the pile-soil interface. The middle convolutional layer combines these edge features to form regional features indicating uniform frictional resistance distribution in the middle section of the pile. The high-level convolutional layer uses global pooling to obtain the overall features of the axisymmetric distribution of the standard pile end resistance, ultimately outputting a standard pile stress distribution feature map (with dimensions H'×W'×C'). This feature map fully preserves the typical spatial pattern of the standard pile stress distribution. The special-shaped pile branch uses the same network structure to process stress distribution maps for special-shaped piles. Due to geometric differences in the input data (such as X-shaped ribs and expanded base structures), the network captures unique stress distribution features in the middle and high convolutional layers. For example, the high-gradient frictional resistance distribution caused by geometric irregularities at the rib edges and the multi-peak end resistance concentration area caused by the expanded cross-section at the expanded end. These features are extracted through a weight-sharing network to form a characteristic stress distribution map for special-shaped piles, highlighting the stress distribution heterogeneity caused by geometric asymmetry.

[0021] The feature map output stage ensures that the dimensions of the two feature maps are aligned and their physical meanings are preserved. By adjusting the convolution kernel stride, padding parameters, and pooling operations at each network layer, the stress distribution feature maps for standard and shaped piles have the same spatial size (H'×W') and number of channels (C'), facilitating subsequent pixel-level differential operations. Each channel of the feature map is assigned specific physical semantics, for example, channel 1 encodes the friction gradient distribution, channel 2 encodes the end resistance concentration, and channel 3 encodes the stress direction characteristics. The feature vector (length C') of each pixel contains multi-dimensional stress information at that location, such as stress magnitude, direction, and its correlation with the surrounding area. This ensures that the feature map not only contains spatial location information but also reflects complex mechanical response characteristics.

[0022] In particular, implementation requires attention to the data source and optimization strategy for network training. Training data can be generated through finite element simulation. Using software such as ABAQUS and ANSYS, simulation models are established for various loading conditions, geological parameters (such as soil cohesion and internal friction angle), and pile geometry (such as rib size and expanded base diameter). Stress distribution maps for standard and special-shaped piles are calculated, forming a large-scale training dataset. To enhance the model's adaptability to real-world scenarios, some data needs to be calibrated against field measurements. For example, stress sensors can be embedded in pile foundations to collect stress data at different depths. This data can then be corrected to ensure that the feature detector can recognize the stress distribution characteristics of actual projects. A pre-training strategy is used for network parameter initialization. Models that perform well in common image feature extraction tasks (such as VGG and ResNet) are selected as a foundation, their pre-trained weights are loaded, and the network is then fine-tuned through end-to-end training. During the fine-tuning process, paired stress distribution maps are used as input, and the correlation between the difference in characteristic maps and the actual bearing capacity correction coefficient is used as the optimization goal. The network weights are adjusted through the back-propagation algorithm to enhance the recognition ability of the unique stress characteristics of special-shaped piles (such as stress concentration at the cross-section mutation point), so that the feature extraction process is deeply coupled with the subsequent bearing capacity correction requirements.

[0023] To address potential computational resource limitations, lightweight techniques are needed to improve model efficiency. For example, depthwise separable convolutions are used in convolutional layers instead of traditional convolution kernels, decomposing the standard 3×3 convolution into depthwise and pointwise convolutions. This reduces computational effort while maintaining feature extraction capabilities. Model pruning techniques are applied to remove connections or neurons in the network that contribute less to feature extraction, reducing model complexity. Quantization techniques are used to convert floating-point parameters to fixed-point, reducing memory usage and computational time. These optimization strategies ensure that the feature detector can run efficiently on portable devices or in distributed computing environments, meeting the needs of real-time processing or batch analysis of large numbers of stress distribution maps.

[0024] In step S3, the stress distribution characteristic diagram of the standard pile and the stress distribution characteristic diagram of the special-shaped pile are subjected to spatial differential feature enhancement to obtain a stress distribution spatial significant differential characteristic diagram. Specifically, Figure 2 FIG. 1 is a flow chart of step S3 in the method for determining bearing capacity based on analysis of bearing characteristics of special-shaped piles according to an embodiment of the present application. Figure 2 As shown, the step S3 includes: S31, calculating the stress distribution space difference feature map between the standard pile stress distribution feature map and the special-shaped pile stress distribution feature map; S32, performing regional feature association receptive field significance on the stress distribution space difference feature map to obtain the stress distribution space significant difference feature map.

[0025] In step S31, the stress distribution space difference characteristic diagram between the stress distribution characteristic diagram of the standard pile and the stress distribution characteristic diagram of the special-shaped pile is calculated. Accordingly, considering that the stress distribution of the standard pile and the special-shaped pile is different due to their different geometric shapes. The core defect of the traditional correction model is that it simplifies the mechanical difference between the special-shaped pile and the standard pile into a single scalar coefficient, while ignoring the stress space distribution reconstruction caused by the change in the geometric shape of the special-shaped pile. For example, the rib structure of the X-type pile will cause the pile side friction resistance to form a strip-shaped high stress area along the edge of the rib, while the friction resistance of the standard pile is evenly distributed in an axisymmetric manner; the expanded diameter area at the end of the expanded bottom pile may produce a ring-shaped end resistance peak cluster, which is significantly different from the uniform compressive stress pattern at the end of the standard pile. Based on this, the present application calculates the stress distribution space difference characteristic diagram between the stress distribution characteristic diagram of the standard pile and the stress distribution characteristic diagram of the special-shaped pile. Specifically, this step compares the high-dimensional feature maps of the two types of piles on a pixel-by-pixel or region-by-region basis. Using differential operations, it strips away the common underlying stress transfer patterns (such as the uniform frictional resistance distribution in the midsection of the pile body) and highlights the differentiated mechanical responses of the irregularly shaped piles due to sudden cross-sectional changes or geometric optimization. For example, for expanded-base piles, the differential feature map emphasizes the stress distribution differences between the expanded end region and the end of a standard pile, while weakening the common characteristics of the straight section of the pile body. For X-shaped piles, the differential feature map focuses on the friction gradient differences between the rib edge and the smooth side of a standard pile. This approach transforms the "global difference" implicit in traditional methods into an explicit "spatial difference distribution," providing a spatial basis for the dynamic generation of subsequent correction coefficients.

[0026] In step S32, the stress distribution spatial differential feature map is subjected to regional feature association receptive field saliency to obtain the stress distribution spatial significant differential feature map. Specifically, Figure 3 FIG. 1 is a flow chart of step S32 in the method for determining bearing capacity based on analysis of bearing characteristics of special-shaped piles according to an embodiment of the present application. Figure 3As shown, step S32 includes: S32-1, extracting a set of stress distribution spatial differential feature pixel-level initial vectors and a stress distribution spatial differential feature vector to be enhanced from the stress distribution spatial differential feature map; S32-2, determining the size of the stress distribution spatial differential feature receptive field of the stress distribution spatial differential feature vector to be enhanced; S32-3, based on the size of the stress distribution spatial differential feature receptive field, filtering a set of pixel-level initial vectors within the local receptive field of the stress distribution spatial differential feature from the set of stress distribution spatial differential feature pixel-level initial vectors; and S32-4, based on the set of pixel-level initial vectors within the local receptive field of the stress distribution spatial differential feature, performing significance enhancement on the stress distribution spatial differential feature vector to be enhanced to obtain an enhanced stress distribution spatial differential feature pixel-level vector. More specifically, in an embodiment of the present application, the enhanced stress distribution spatial differential feature pixel-level vector is the channel feature vector at the (i, j)th pixel position of the stress distribution spatial significant differential feature map.

[0027] It should be understood that the stress reconstruction effect caused by the geometric irregularity of special-shaped piles (such as the jump in friction resistance at the edge of the X-shaped pile rib and the annular stress diffusion at the end of the expanded-base pile) is often limited to specific areas, and traditional global correction models or local analysis methods with fixed windows have difficulty accurately capturing such spatially localized, nonlinear mechanical responses. For example, in an expanded-base pile, the high stress concentration in the expanded diameter area at the end and the difference in stress distribution in the straight pipe section of the pile body may only occupy a local area of ​​the entire pile body stress distribution map. However, traditional methods use uniform weighting or fixed-size receptive field processing, which will mix the difference signals in such key areas with noise or non-significant difference areas, resulting in the generation of correction coefficients being interfered with by irrelevant areas and unable to accurately reflect the actual bearing gain of the special-shaped pile. Based on this, in the technical solution of the present application, the stress distribution spatial differential feature map is subjected to regional feature association receptive field significance to obtain the stress distribution spatial significant differential feature map.

[0028] Specifically, the core feature vector at each pixel in the differential feature map is first extracted through feature decoupling and information compression. The optimal receptive field size at each location is then adaptively determined based on the spatial structure of the feature distribution. For example, in the high-difference region at the end of the expanded pile, the network automatically expands the receptive field to capture the circularly diffuse stress difference pattern by analyzing the spatial correlation of the compressed feature vectors. Meanwhile, in the linear difference band at the edge of the X-shaped pile rib, the receptive field is contracted to focus on the narrow region of high gradient difference. This dynamic anchoring mechanism enables the network to flexibly adjust its focus based on the spatial semantics of the difference features (such as shape, gradient direction, and local contrast). It then selects a set of local features directly related to the bearing capacity of the irregular pile (such as the multi-peak difference region at the end and the transition zone at the rib edge) from the original differential feature map and enhances their saliency through contextual information aggregation. This regional saliency optimization of the differential feature map allows for more precise capture and quantification of these nonlinear mechanical responses. Furthermore, it enables more accurate mapping of the spatial feature information into lateral friction and end resistance correction coefficients. The significant feature maps highlight the key information related to the correction coefficients, making the coefficient generation process based on these features more accurate and reliable, thereby improving the accuracy of the ultimate bearing capacity prediction of special-shaped piles and better meeting the actual needs of the project.

[0029] Specifically, in the embodiment of the present application, step S32-1 includes: performing feature decoupling on the stress distribution spatial differential feature map along the channel dimension to obtain a set of pixel-level initial vectors of the stress distribution spatial differential feature. This process can be expressed as follows:

[0030]

[0031] in, is the stress distribution spatial differential characteristic diagram, is the set of real numbers, and They are The height and width of each feature matrix along the channel dimension, yes The number of channels, It is feature decoupling, is each stress distribution space differential feature pixel-level initial vector in the set of stress distribution space differential feature pixel-level initial vectors; Extract the stress distribution spatial differential feature pixel-level initial vector at the (i, j)th pixel position from the set of stress distribution spatial differential feature pixel-level initial vectors as the stress distribution spatial differential feature vector to be enhanced. This process can be expressed as follows:

[0032] in, yes The stress distribution spatial differential feature pixel-level initial vector at the (i, j)th pixel position in , It will As the spatial differential eigenvector of the stress distribution to be enhanced.

[0033] Understandably, traditional methods, due to inter-channel coupling, result in global averaging of local mechanical response characteristics, making it impossible to precisely capture the spatial heterogeneity of stress distribution in irregular piles. For example, the annular stress diffusion at the end of the expanded-base pile and the low-gradient region of the straight pipe section of the pile body exhibit nonlinear mixing in channel characteristics. Direct global weighting would obscure key differential signals. By decoupling the spatial differential feature map of the stress distribution along the channel dimension and independently extracting the channel feature vector for each pixel, redundant inter-channel dependencies can be broken, allowing subsequent processing to independently focus on the fine-grained feature expression of each pixel (such as the high-contrast transition at the end or the gradient directionality at the rib edge). This provides a pristine feature substrate uncontaminated by channel coupling for subsequent dynamic receptive field adjustment, ensuring the independent representation of local nonlinear mechanical patterns (such as multi-peak differential regions).

[0034] Accordingly, stress differences in irregular piles exhibit significant spatial localization (e.g., the narrow, high-gradient bands at the edges of X-shaped pile ribs), necessitating pixel-by-pixel anchoring of key locations for targeted enhancement. Extracting the spatial differential eigenvector of the stress distribution to be enhanced, centered around the (i, j)th pixel, essentially spatially anchors regions of sudden changes in stress distribution (e.g., stress concentration points at the ends of expanded piles), avoiding the introduction of noise from non-correlated regions due to inadequate coverage of traditional fixed windows. This ensures that subsequent operations can focus on the core anchor points of the local mechanical response with high spatial resolution, providing precise initial feature anchors for semantically driven adjustments of the dynamic receptive field.

[0035] Specifically, in the embodiment of the present application, step S32-2 includes: performing information compression on the stress distribution space differential feature vector to be enhanced to obtain the stress distribution space differential feature distillation vector to be enhanced. This process can be expressed as:

[0036] in, To calculate the Euclidean norm of a vector, To calculate the square of the Euclidean norm of a vector, is the spatial differential feature distillation vector of the stress distribution to be enhanced; Based on the characteristic distribution space structure characteristics of the stress distribution space differential feature distillation vector to be enhanced, the size of the stress distribution space differential feature receptive field of the stress distribution space differential feature vector to be enhanced is determined. This process can be expressed by the formula:

[0037] in, is the logarithmic function value with base 2, yes The size of the receptive field of the stress distribution spatial differential feature.

[0038] It should be understood that the originally extracted spatial differential eigenvectors of the stress distribution to be enhanced contain redundant information (such as background noise or low-significance stress fluctuations). Directly using them for receptive field prediction will interfere with the model's identification of key mechanical modes (such as annular diffusion or linear transitions). By compressing the spatial differential eigenvectors of the stress distribution to be enhanced, redundant components (such as weak differential signals in uniform stress zones) can be stripped away, while retaining core features that are strongly correlated with bearing capacity correction (such as the spatial correlation of end stress peaks). For example, after information compression, the high-difference region at the end of the expanded pile has its annular diffusion pattern's characteristic distribution compactness enhanced, providing high signal-to-noise ratio input conditions for the subsequent adaptive prediction of the receptive field size, preventing irrelevant regions from interfering with model decisions.

[0039] Accordingly, due to the strong spatial heterogeneity of the mechanical response of irregularly shaped piles, a traditional fixed receptive field cannot simultaneously accommodate both the large-scale annular diffusion of the expanded base pile and the narrow, high-gradient differences at the rib edges. However, based on the spatial structural properties of the distilled features after information compression (such as gradient direction consistency and local contrast distribution), the model dynamically predicts the optimal receptive field size through semantic association: expanding the receptive field in the end diffusion region to capture the annular context and shrinking the receptive field at the rib edges to focus on the linear transition band. In other words, by mapping the semantics of the spatial differential distillation features to be enhanced (such as shape priors and local contrast) into receptive field parameters for the spatial differential features of the stress distribution, "structurally aware" contextual range selection can be achieved, aligning the saliency expression of key differential regions (such as multi-peak differential regions) with physical and mechanical mechanisms (such as stress diffusion paths).

[0040] Specifically, in the embodiment of the present application, step S32-3 includes: based on the size of the stress distribution spatial differential feature receptive field, screening out a set of pixel-level initial vectors within the stress distribution spatial differential feature local receptive field from the set of pixel-level initial vectors of the stress distribution spatial differential feature. This process can be expressed by the formula:

[0041] in, is the set of pixel-level initial vectors within the local receptive field of the stress distribution spatial differential feature. They are Middle ( ) , the ( ) , the ( ) , the ( ) and ( ) pixel position stress distribution spatial differential feature local receptive field pixel-level initial vector.

[0042] It should be understood that the dynamic receptive field needs to accurately cover contextual regions that are strongly correlated with the mechanical behavior of the anchor pixel. For example, the end anchor of the expanded pile needs to include all high-stress difference pixels within the annular neighborhood to model the diffusion effect, while the rib edge anchor only needs to extract narrow band features along the gradient direction. By filtering the set of pixel-level initial vectors within the local receptive field of the stress distribution spatial differential feature based on the predicted size, centered on the anchor point, a contextual information library that matches the current mechanical model (such as the end annular neighborhood or the rib edge linear neighborhood) can be constructed. In other words, this step ensures that subsequent saliency enhancement only incorporates difference signals from semantically relevant regions (such as the continuous gradient changes in the stress transition zone), avoiding interference from cross-regional noise (such as the low-difference background of the straight pipe section of the pile body) in the generation of correction coefficients.

[0043] Specifically, in an embodiment of the present application, step S32-4 includes: performing feature processing based on volume-boundary conformal regularization constraints on the set of pixel-level initial vectors within the local receptive field of the stress distribution spatial differential feature to obtain a first modulation weighting coefficient and a second modulation weighting coefficient; based on the first modulation weighting coefficient and the second modulation weighting coefficient, performing weighted fusion based on attention weights on the stress distribution spatial differential feature vector to be enhanced and the set of pixel-level initial vectors within the local receptive field of the stress distribution spatial differential feature to obtain the enhanced stress distribution spatial differential feature pixel-level vector. The above process can be expressed as follows:

[0044]

[0045] in, yes Middle ( ) pixel position stress distribution spatial differential feature local receptive field pixel-level initial vector, is the stress distribution spatial differential feature scoring weight vector, is matrix multiplication, yes function, yes The corresponding stress distribution spatial differential feature attention weight, and are the first modulation weighting coefficient and the second modulation weighting coefficient, respectively, yes The enhanced stress distribution spatial differential feature pixel-level vector after enhancement is the channel feature vector of the (i, j)th pixel position of the stress distribution spatial significant differential feature map.

[0046] In particular, the first modulation weighting coefficient and the second modulation weighting coefficient are first defined. That is, for the stress distribution space differential eigenvector to be enhanced The corresponding set of pixel-level initial vectors in the local receptive field of the stress distribution spatial differential feature In order to enhance the gain of significant feature channels and suppress the attenuation of non-significant feature channels, it is required that the voxel domain representation and the boundary topology representation must satisfy the geometric correspondence constraint in the manifold embedding space, that is, the two must satisfy the high-dimensional manifold homeomorphism. Therefore, the stress distribution spatial differential feature voxel domain representation vector is first determined as:

[0047] The stress distribution spatial differential characteristic boundary topology representation vector is:

[0048] Then, by modulating the weighting coefficient and , so that the interface-voxel tensor field has a regular constraint that satisfies the exchange relation, that is, the stress distribution space differential feature voxel domain representation vector and stress distribution spatial differential characteristic boundary topology representation vector The spatial two-norm representation of the difference vector tends to the coefficient and The product of:

[0049] in is the scaling factor.

[0050] On the premise of satisfying the interface morphology constraint criteria and utilizing the conformal exchange invariance of the voxel domain, the above method can ensure the conformality of the feature manifold geometry, thereby achieving high-precision conformal fusion of the set of pixel-level initial vectors within the local receptive field of the stress distribution space differential feature, significantly improving the context-aware saliency expression efficiency of the stress distribution space differential feature vector to be enhanced.

[0051] Because when performing weighted fusion later, and The sum of is 1, and thus, the first modulation weighting coefficient and the second modulation weighting coefficient can be obtained through the above processing.

[0052] It should be understood that the bearing capacity correction of irregular piles relies on the coordinated expression of local mechanical responses and global physical constraints. By aggregating pixel-level initial vectors within the local receptive field of spatial differential features of stress distribution (such as the multimodal differential features of the end ring neighborhood or the gradient coherence of the rib edge), and performing weighted fusion on the spatial differential feature vectors of the enhanced stress distribution, patterns strongly correlated with the correction coefficient (such as the spatial persistence of stress concentration zones) can be amplified, while random fluctuations or insignificant differences can be suppressed. For example, after the annular context enhancement of the end anchor point of the expanded pile, its multimodal differential features are highlighted in the channel dimension, directly mapping them into the quantitative basis for the end resistance correction coefficient. In other words, through semantically driven feature reconstruction, nonlinear mechanical responses can be transformed into interpretable and significant expressions, thereby improving the physical consistency and engineering applicability of the generated correction coefficients for lateral friction resistance and end resistance.

[0053] In step S4, the stress distribution spatial significant differential feature map is feature decoded to obtain a side friction correction coefficient and an end resistance correction coefficient. Specifically, in an embodiment of the present application, step S4 includes: inputting the stress distribution spatial significant differential feature map into a decoder-based side friction correction analyzer and a decoder-based end resistance correction analyzer, respectively, to obtain the side friction correction coefficient and the end resistance correction coefficient. It should be understood that the bearing capacity of a pile is composed of both side friction and end resistance, but their mechanisms of action and influencing factors are different. By inputting the stress distribution spatial significant differential feature map into different analyzers, the characteristic information related to side friction and end resistance can be specifically separated and analyzed, thereby more accurately grasping their respective correction coefficients. For example, factors such as the roughness of the pile side surface and the contact between the soil and the pile mainly affect side friction, while factors such as the shape of the pile end and the properties of the bearing layer have a greater impact on end resistance. Different analyzers can process the feature map according to their corresponding characteristics to extract accurate correction coefficients. It is worth mentioning that the decoder has the ability to map abstract features back to a specific parameter space in the neural network. The decoder-based analyzer can transform the complex feature information in the stress distribution spatially significant differential feature map into physically meaningful side friction and end resistance correction coefficients through learning and reasoning. In other words, the decoder can process high-dimensional feature data, compress it, and convert it into one-dimensional correction coefficients, effectively converting image features into actual engineering parameters.

[0054] In step S5, the total pile side friction and the total pile end resistance of the standard pile under the predetermined load are obtained. It should be understood that the total pile side friction of the standard pile under the predetermined load includes the sum of the shear forces generated by the relative displacement trend between the pile surface and the surrounding soil. Its value reflects the ability of the pile to transfer load through friction between the side and the soil. It specifically includes the distribution characteristics of the pile side friction at the interface of different soil layers (such as the degree of friction along the pile length, peak location, and attenuation pattern), as well as the contribution ratio of the friction resistance of each soil layer to the total friction. This parameter is directly related to soil properties (such as cohesion and internal friction angle), pile-soil interface characteristics (such as roughness), and pile surface area. The total resistance at the pile end includes the sum of the resistance forces generated by the soil at the pile end plane due to compression, reflecting the ability of the pile to transfer loads through the pile end and cause the soil below the pile end to undergo compression or shear failure. Its value is closely related to the bearing capacity characteristic value of the soil at the pile end, the geometric dimensions of the pile end (such as cross-sectional area) and the soil failure mode (such as overall shear, local shear or punching shear failure), reflecting the stress concentration effect and bearing performance of the soil at the pile end under load. In short, by obtaining the total pile side friction resistance and the total pile end resistance of the standard pile under the predetermined load, it is possible to provide benchmark reference data for the calculation of the ultimate bearing capacity of special-shaped piles and establish a connection point with the traditional pile foundation bearing capacity calculation method. In particular, in a specific embodiment of the present application, the calculation method of the total pile side friction resistance and the total pile end resistance is as follows:

[0055]

[0056] in, is the total friction resistance on the pile side, represents the effective perimeter of the pile side, represents the standard value of the ultimate friction resistance of the pile side of the i-th layer of soil, represents the thickness of the pile body through the i-th soil layer, is the total pile end resistance, Indicates the standard value of the ultimate end resistance of the pile end bearing layer. Indicates the effective projected area of ​​the pile tip.

[0057] In step S6, based on the side friction correction coefficient and the end resistance correction coefficient, the total pile side friction and the total pile end resistance are corrected to obtain the ultimate bearing capacity of the special-shaped pile under a predetermined load. It is understandable that due to their special geometric shapes (such as X-shaped, H-shaped, and expanded-base piles), the mechanical behavior of special-shaped piles under load is significantly different from that of standard piles. Their stress distribution is uneven, and the side friction and end resistance are also different from those of standard piles. Traditional bearing capacity calculation methods based on standard piles cannot accurately reflect the true bearing capacity of special-shaped piles. Therefore, it is necessary to correct the total pile side friction and the total pile end resistance of the standard pile based on the correction coefficient determined based on the difference in stress distribution between the special-shaped pile and the standard pile to adapt to the unique mechanical characteristics of the special-shaped pile. The side friction correction coefficient and the end resistance correction coefficient obtained in the previous steps quantify the impact of the stress distribution differences caused by the different geometric shapes of the special-shaped piles on the friction and end resistance. By adjusting the friction and end resistance of standard piles based on these correction coefficients, the influence of the geometric shape of special-shaped piles on the bearing capacity can be accurately reflected in the calculation results, making the bearing capacity calculation more scientific and reasonable. That is, the ultimate bearing capacity of special-shaped piles under a predetermined load ( ) = Side friction correction coefficient ( )×total friction resistance on pile side( ) + end resistance correction factor ( )×total pile end resistance ( For example, assume that the parameters of a concave-angle special-shaped pile are: pile surrounding layer: 2m thick silt ( ), 3m thick adhesive ( ); Pile end bearing layer: sand ( ), calculation parameters: , α=1.15 (concave angle increases lateral resistance), ,but 1.2x[40x2+60x3]+1.05x2000x0.25=477.6+525=1002.6 kN.

[0058] In summary, a bearing capacity determination method based on the bearing characteristics analysis of special-shaped piles based on the embodiment of the present application is explained, which first obtains the stress distribution diagram of the two types of piles under the same load, extracts their spatial features and calculates the differential features, and then uses the feature significant enhancement to enhance the identification of the stress concentration area unique to the special-shaped piles in the differential diagram to quantify the nonlinear mechanical response, and then maps the spatial significant differences into dynamic correction coefficients of friction resistance and end resistance through feature decoding, and finally completes the correction of the ultimate bearing capacity of the special-shaped piles in combination with the bearing capacity data of the standard piles. In this way, the dynamic changes in the friction resistance distribution caused by the geometric asymmetry of the pile body and the multi-peak concentration phenomenon of the end resistance can be accurately captured, and at the same time, the optimization efficiency of the correction parameters under complex working conditions can be significantly improved, which is conducive to meeting the core demands of new energy pile foundation projects for material saving and high reliability.

Claims

1. A method for determining bearing capacity based on bearing characteristics analysis of special-shaped piles, characterized in that: include: Obtaining stress distribution diagrams of standard piles and special-shaped piles under predetermined loads; Extracting stress distribution spatial characteristics from the stress distribution diagram of the standard pile and the stress distribution diagram of the special-shaped pile to obtain a stress distribution characteristic diagram of the standard pile and a stress distribution characteristic diagram of the special-shaped pile; Performing spatial differential feature enhancement on the stress distribution characteristic map of the standard pile and the stress distribution characteristic map of the special-shaped pile to obtain a stress distribution spatially significant differential feature map, including: calculating a stress distribution spatially significant differential feature map between the stress distribution characteristic map of the standard pile and the stress distribution characteristic map of the special-shaped pile; performing regional feature association receptive field saliency on the stress distribution spatially significant differential feature map to obtain the stress distribution spatially significant differential feature map; Performing feature decoding on the stress distribution spatial significant difference feature map to obtain a side friction correction coefficient and an end resistance correction coefficient; Obtaining the total pile side friction resistance and the total pile end resistance of the standard pile under the predetermined load; Based on the side friction correction coefficient and the end resistance correction coefficient, the total pile side friction and the total pile end resistance are corrected to obtain the ultimate bearing capacity of the special-shaped pile under a predetermined load.

2. The method for determining bearing capacity based on bearing property analysis of special-shaped piles according to claim 1, characterized in that: Extracting stress distribution spatial features from the stress distribution diagram of the standard pile and the stress distribution diagram of the special-shaped pile to obtain a stress distribution characteristic diagram of the standard pile and a stress distribution characteristic diagram of the special-shaped pile, including: using a stress distribution spatial feature detector based on a twin convolutional neural network to process the stress distribution diagram of the standard pile and the stress distribution diagram of the special-shaped pile to obtain the stress distribution characteristic diagram of the standard pile and the stress distribution characteristic diagram of the special-shaped pile.

3. The method for determining bearing capacity based on bearing property analysis of special-shaped piles according to claim 1, characterized in that: Performing regional feature association receptive field saliency on the stress distribution spatial differential feature map to obtain a stress distribution spatial significant differential feature map, including: Extracting a set of stress distribution spatial differential feature pixel-level initial vectors and a stress distribution spatial differential feature vector to be enhanced from the stress distribution spatial differential feature map; Determining the size of the stress distribution spatial differential feature receptive field of the stress distribution spatial differential feature vector to be enhanced; Based on the size of the stress distribution spatial differential feature receptive field, screening out a set of pixel-level initial vectors within the stress distribution spatial differential feature local receptive field from the set of pixel-level initial vectors of the stress distribution spatial differential feature; Based on a set of pixel-level initial vectors within the local receptive field of the stress distribution spatial differential feature, the stress distribution spatial differential feature vector to be enhanced is significantly enhanced to obtain an enhanced stress distribution spatial differential feature pixel-level vector.

4. The method for determining bearing capacity based on bearing property analysis of special-shaped piles according to claim 3, characterized in that: The enhanced stress distribution spatial differential feature pixel-level vector is a channel feature vector at the (i, j)th pixel position of the stress distribution spatial significant differential feature map.

5. The method for determining bearing capacity based on bearing property analysis of special-shaped piles according to claim 4, characterized in that: Extracting a set of stress distribution space differential feature pixel-level initial vectors and a stress distribution space differential feature vector to be enhanced from the stress distribution space differential feature map, including: Performing feature decoupling on the stress distribution spatial differential feature map along the channel dimension to obtain a set of pixel-level initial vectors of the stress distribution spatial differential feature; The stress distribution spatial differential feature pixel-level initial vector at the (i, j)th pixel position is extracted from the set of stress distribution spatial differential feature pixel-level initial vectors as the stress distribution spatial differential feature vector to be enhanced.

6. The method for determining bearing capacity based on bearing property analysis of special-shaped piles according to claim 5, characterized in that: Determining the size of the stress distribution spatial differential feature receptive field of the stress distribution spatial differential feature vector to be enhanced includes: Performing information compression on the stress distribution space differential feature vector to be enhanced to obtain a stress distribution space differential feature distillation vector to be enhanced; Based on the feature distribution space structure characteristics of the stress distribution space differential feature distillation vector to be enhanced, the size of the stress distribution space differential feature receptive field of the stress distribution space differential feature vector to be enhanced is determined.

7. The method for determining bearing capacity based on bearing property analysis of special-shaped piles according to claim 6, characterized in that: Based on a set of pixel-level initial vectors within a local receptive field of the stress distribution spatial differential feature, the stress distribution spatial differential feature vector to be enhanced is significantly enhanced to obtain an enhanced stress distribution spatial differential feature pixel-level vector, including: Performing feature processing based on volume-boundary conformal regularization constraints on a set of pixel-level initial vectors within a local receptive field of the stress distribution spatial differential feature to obtain a first modulation weighting coefficient and a second modulation weighting coefficient; Based on the first modulation weighting coefficient and the second modulation weighting coefficient, the stress distribution spatial differential feature vector to be enhanced and the set of pixel-level initial vectors within the local receptive field of the stress distribution spatial differential feature are weighted fused based on the attention weight to obtain the enhanced stress distribution spatial differential feature pixel-level vector.

8. The method for determining bearing capacity based on bearing property analysis of special-shaped piles according to claim 7, characterized in that: The stress distribution space significant differential characteristic map is feature decoded to obtain a side friction resistance correction coefficient and an end resistance correction coefficient, including: inputting the stress distribution space significant differential characteristic map into a decoder-based side friction resistance correction analyzer and a decoder-based end resistance correction analyzer respectively to obtain the side friction resistance correction coefficient and the end resistance correction coefficient.

Images