Earth site disease identification and fusion do old 3D printing fidelity repair method and system

Abstract

The application belongs to the field of cultural relic protection, and specifically discloses a 3D printing fidelity repair method and system for earthen site disease identification and fusion aging, which comprises the following steps: collecting multi-source information of a site body, performing fusion processing, outputting a standardized fusion data set, identifying diseases and quantifying parameters, constructing a virtual model of a masonry cultural relic body, dividing a masonry repair area into standard component models for numbering, saving texture parameters and repair parameters of each component model, preparing rammed earth repair pieces and masonry repair pieces, performing texture replication and gradient aging processing, assembling on a site to form a stable rammed earth-masonry composite structure. According to the technical scheme, accurate disease identification and quantification, accurate texture replication, laser modular aging, factory pre-matching and on-site accurate assembly are integrated into a closed-loop repair, a stable rammed earth-masonry composite structure is constructed, the repair fidelity and construction efficiency are improved, and secondary damage to cultural relics is avoided.

Description

Technical Field

[0001] This invention belongs to the field of cultural relic protection and relates to a 3D printing-based method and system for the accurate restoration of earthen sites by identifying and integrating aging techniques. Background Technology

[0002] The site is an immovable cultural relic with rammed earth and brick as the core construction material. It includes the Great Wall, ancient city walls, beacon towers, rammed earth building foundations and other types. It is an important part of my country's cultural heritage. Its materials are mainly soil, sand, stone and a small amount of cementing material. It is characterized by loose texture, poor durability and susceptibility to damage caused by natural environment (wind erosion, water erosion, temperature and humidity changes) and human factors.

[0003] Currently, archaeological site restoration mainly employs traditional manual restoration methods. These methods rely heavily on human experience and are highly subjective. Manual restoration has low precision, making it difficult to accurately replicate complex damage patterns, resulting in poor surface smoothness and consistency of form after restoration. The aging process is also crude, often using simple dyeing and polishing methods, and the aging effect does not blend with the original weathering degree and texture characteristics of the site.

[0004] Therefore, developing a site restoration technology that can accurately identify defects, faithfully replicate the restoration, and achieve a natural aging effect, while effectively improving restoration durability and avoiding secondary damage, has become an urgent technical challenge in the field of site protection.

[0005] Current 3D printing restoration methods only achieve a simple mix of the original soil, without precise control over the soil's porosity, permeability, and strength. This leads to frequent cracking, detachment, and recurrence of damage such as salting and efflorescence. Traditional restoration methods use imported bricks, stones, and mortar, which have poor compatibility with the original artifact's material, resulting in cracking, detachment, and a high recurrence rate. Hand-replicated textures have low precision, failing to reproduce the original artifact's textural details and weathering characteristics; the aging process is also crude, resulting in poor integration with the artifact.

[0006] Current modular restoration methods fail to achieve accurate texture replication and targeted aging, resulting in low splicing precision and difficulty in forming a stable rammed earth-brick composite structure. 3D printed restoration materials have poor compatibility with the original brick and stone materials, and the aging process relies on manual labor, leading to low efficiency and poor results. Laser engraving technology has not been specifically applied to the modular aging of brick and stone artifacts, and it has not been combined with deep learning to achieve accurate texture replication, thus failing to meet the requirements of "authentic restoration and natural integration" for brick and stone artifacts.

[0007] The existing restoration technology lacks an integrated system for disease identification, texture replication, modular segmentation, aging, and assembly, resulting in low construction efficiency. Furthermore, it is difficult to achieve non-destructive disassembly and post-repair maintenance, which can easily cause secondary damage to cultural relics. Summary of the Invention

[0008] The purpose of this invention is to address the aforementioned problems in existing technologies by proposing a 3D printing-based method and system for the accurate restoration of earthen archaeological sites by identifying and integrating aging techniques.

[0009] To achieve the above objectives, the basic solution of this invention is: a 3D printing-based method for the accurate restoration of earthen archaeological sites by identifying and integrating aging techniques, comprising the following steps:

[0010] Multi-source information was collected on the site itself, and the collected heterogeneous data from multiple sources were fused and processed to output a standardized fused dataset.

[0011] The standardized fusion dataset is input into the deep learning disease identification model for disease identification and parameter quantification. Based on the standardized fusion dataset and disease identification results, a virtual model of the brick and stone cultural relic is constructed to determine the thickness of the rammed earth restoration, as well as the boundaries of the rammed earth restoration area and the brick and stone restoration areas on both sides, and to clarify the texture replication range and aging requirements of the brick and stone restoration area.

[0012] Deep learning is used to replicate the texture of the defective area. The texture-replicated brick and stone repair area is divided into several independent standard component models and numbered. The texture parameters and repair parameters of each component model are associated and saved.

[0013] Based on the rammed earth restoration area, a recyclable double-layer cavity restoration mold was obtained by 3D printing. Using the original soil around the site as raw material, the rammed earth restoration parts were prepared by layered ramming and combined with low temperature curing treatment.

[0014] Based on the component model, texture parameters, and repair parameters of the brick and stone repair area segmentation, brick and stone repair parts are prepared, and CNC laser engraving technology is used to perform texture replication and gradient aging treatment on the exposed surface of the component.

[0015] All brick and stone components are pre-assembled to form a complete brick and stone restoration body;

[0016] The rammed earth restoration components and brick and stone restoration components were assembled on-site to form a stable rammed earth-brick and stone composite structure.

[0017] The working principle and beneficial effects of this basic solution are as follows: This technical solution achieves automatic identification and quantification of disease type, location, depth, area, severity level, etc. through multi-source information collection and deep learning model, overcoming the problems of strong subjectivity and low quantification accuracy of traditional methods.

[0018] Deep learning is used to replicate the texture of defective areas, ensuring that the texture direction, depth of depressions and convexities, and weathering gradient of the repaired area are highly consistent with the original site, achieving authentic restoration. CNC laser engraving technology is employed to perform gradient aging treatment on the exposed outer surfaces of the brick and stone components, simulating natural weathering, wear, and erosion effects, making the repaired area indistinguishable from the original site to the naked eye. A stable rammed earth-brick and stone composite structure is constructed to improve the fidelity of the restoration and construction efficiency, avoiding secondary damage to the cultural relics.

[0019] Furthermore, multi-source information was collected from the site itself, and the collected heterogeneous data from multiple sources was fused to output a standardized fused dataset, specifically:

[0020] Anchor holes were pre-drilled in the main body of the site;

[0021] Using 3D laser scanners or close-range photogrammetry, 3D morphological data of the damaged parts of the site and the surrounding undamaged areas are collected to identify the location, area, size of surface peeling and defects, as well as the surface texture of the intact areas.

[0022] Based on the three-dimensional morphological data of the site itself, and combined with the stress characteristics of the rammed earth-brick composite structure and the anchoring requirements of the restoration, the anchor installation parameters are set through structural stress simulation and spatial point adaptation. These parameters include the anchor installation location, thickness, and insertion depth, ensuring that the anchor matches the site structure. Specifically:

[0023] Based on the three-dimensional morphological data of the site itself, the anchor bolt safe installation points were locked, avoiding defective areas;

[0024] Based on the dimensions and weight of the masonry restoration, the stress threshold is calculated, and the anchor bolt diameter is matched (e.g., 8-12mm diameter for carbon fiber anchor bolts). Specifically:

[0025] Calculate the self-weight of the masonry restoration:

[0026] ,

[0027] Where G represents the total self-weight of the brick and stone restoration; ρ is the density of the antique brick and stone material; V is the three-dimensional volume of the brick and stone restoration; and g is the acceleration due to gravity.

[0028] Calculate the combined stress load:

[0029] ,

[0030] Among them, F 总 : Total stress threshold for anchor bolt design; k: Safety factor for cultural relic restoration, ranging from 1.2 to 1.5, which is the minimum stress bearing threshold that the anchor bolt must meet;

[0031] Match anchor bolt specifications according to stress threshold:

[0032] ,

[0033] Where A: effective cross-sectional area of ​​the anchor bolt; [σ]: allowable stress of the anchor bolt material;

[0034] The diameter of the anchor rod is calculated from the cross-sectional area, and the final anchor rod thickness specification is determined.

[0035] Based on the thickness of the rammed earth repair and the depth of the body stabilization layer, the anchor bolt implantation depth is set by three-dimensional spatial coordinate calibration to ensure that the implanted end penetrates into the body stabilization layer and the exposed end is compatible with the anchor hole of the brick and stone repair body.

[0036] Finally, all anchor bolt parameters were spatially registered and verified with the three-dimensional morphological data of the body to complete the parameter setting; an infrared thermal imager was used to detect the location and depth of hollow areas, hidden cracks and differences in water content inside the archaeological artifacts, and to identify loose and eroded areas inside, providing a basis for controlling the thickness and density of rammed earth restoration.

[0037] The composition of the soil (rammed earth area) on the surface of brick and stone cultural relics, the composition of brick and stone materials, the degree of weathering, and the distribution range of salt precipitation areas were analyzed by using a hyperspectral imager to determine the color gradient and weathering parameters of texture replication, providing data support for aging treatment;

[0038] Ultrasonic compaction testing was used to determine the compaction of the brick and stone body and the rammed earth area, identify the loose areas inside, and provide a basis for the layered ramming force and low-temperature curing parameters of rammed earth repair.

[0039] Collect dimensional tolerances, splicing gaps, and bonding parameters of the subgrade components, and combine them with anchor bolt installation requirements to provide data support for virtual segmentation numbering and factory assembly;

[0040] The collected multi-source heterogeneous data were normalized, noise removed, and missing values ​​filled in. Core correlation features for the restoration of brick and stone cultural relics were extracted from the preprocessed multi-source data, including geometric features: surface texture, defect outline, splicing tolerance, and anchoring positioning coordinates; material physicochemical features: brick and stone matrix composition, weathering grade, moisture content, porosity, and density; structural mechanical features: location of internal defects, hollow area, crack depth, and anchoring stress threshold; process adaptation features: restoration material matching parameters, anchoring depth, and splicing gap; and appearance fidelity features: brick and stone surface color, weathering texture gradient, and aging matching parameters.

[0041] Using the spatial coordinates of the three-dimensional morphological data of cultural relics as a benchmark, the remaining data are spatially / dimensionally registered with the morphological features to ensure one-to-one correspondence between the data and output a standardized fusion dataset.

[0042] By comprehensively acquiring information on the morphology, internal defects, composition, and density of the site, and through data normalization, registration, and fusion, a standardized dataset is output, providing a precise and quantifiable basis for subsequent disease identification and restoration design.

[0043] Furthermore, the deep learning disease identification model adopts the XGBoost model and combines SHAP interpretability analysis. The deep learning disease identification model includes an input layer (multi-source detection features), a feature preprocessing layer (normalization, missing value imputation), an XGBoost classification and regression layer (based on the number of learners X, maximum depth Y, and learning rate Z), a SHAP feature contribution analysis layer, and an output layer (disease type, quantification parameters, and repair parameters) connected in sequence.

[0044] Input layer: Receives multi-source detection data, covering core data such as site three-dimensional morphology data, hyperspectral material data, infrared detection data, and ultrasonic density, and captures key information such as surface texture, material composition, and internal defects, all in a standardized digital format to ensure compatibility with subsequent preprocessing layers;

[0045] Feature preprocessing layer:

[0046] Normalization: The min-max normalization algorithm is used to map the input multi-source detection features (such as texture parameters and density data) to the [0,1] interval, eliminating the dimensional differences between data of different dimensions;

[0047] Missing value imputation: For missing items that may appear in multi-source data, the nearest neighbor interpolation method is used to imput them, giving priority to preserving the authenticity of the original data and avoiding the impact of missing data on the accuracy of disease identification.

[0048] XGBoost Classification and Regression Layers:

[0049] Basic architecture: The gradient boosting tree is the core, containing X basic learners (for example, X is 100-150, which can be adapted to the accuracy requirements of site disease identification). All learners run in series to gradually optimize the identification accuracy.

[0050] Parameter configuration: The maximum depth Y is set to 3-5 (to avoid overfitting and adapt to the complexity of the site's damage), and the learning rate Z is set to 0.1-0.3 (to ensure model convergence and avoid gradient vanishing).

[0051] Functional modules:

[0052] (Classification): Classify the input preprocessed features to identify core defects (surface peeling, internal looseness, cracks, salt precipitation, etc.).

[0053] (Regression): Quantitatively calculate the size and severity of the defects and output specific values ​​(such as defect area and depth) to provide data support for setting subsequent repair parameters.

[0054] Iterative optimization: Receive the SHAP feature contribution analysis results, dynamically adjust the learner parameters, and improve the accuracy and stability of disease identification.

[0055] To quantify the contribution of different detection features to the disease identification results, the SHAP analysis method is introduced to conduct interpretability analysis on the trained disease identification model.

[0056] SHAP Feature Contribution Analysis Layer:

[0057] Extract all features from the XGBoost classification and regression layer outputs, and clarify the influence weight of each feature (such as texture parameters and density) on disease identification;

[0058] The contribution of each detection feature is quantified using SHAP values ​​to select the core features most critical for disease identification (such as the density data of hollow areas) and eliminate redundant features.

[0059] The feature contribution results are fed back to the XGBoost classification and regression layers to optimize model parameters, while the feature contribution ranking is output to ensure consistency with the interpretability of the overall repair method.

[0060] Output layer:

[0061] Clearly identify the core diseases (surface peeling, internal loosening, cracks, salt precipitation, etc.), and label the specific disease name and its category;

[0062] Output specific quantitative data on the defects, including defect area, depth, and distribution location, which correspond to the parameter requirements for subsequent repair thickness and texture replication;

[0063] The standardized fusion dataset is input into the deep learning disease identification model for disease identification and parameter quantification, specifically as follows:

[0064] The detection features, such as 3D morphology, infrared imaging, spectral detection, and substrate material, are normalized and then input into the XGBoost model.

[0065] The XGBoost model automatically classifies disease types, including efflorescence, spalling, cracking, and erosion, and outputs quantitative parameters, including disease location, area, and depth.

[0066] The contribution of each disease type and quantitative parameter is calculated by SHAP analysis, redundant features are eliminated, the identification accuracy is optimized, and the reliability of disease identification and quantification results is ensured.

[0067] The system outputs the disease type and ranks the contribution of quantitative parameters to provide data support for setting subsequent repair parameters.

[0068] Deep learning disease identification models have a simple structure and are easy to use.

[0069] Furthermore, based on the standardized fusion dataset and the results of disease identification, a virtual model of the brick and stone cultural relic was constructed to determine the thickness of the rammed earth restoration, as well as the boundaries of the rammed earth restoration area and the brick and stone restoration areas on both sides. The method for clarifying the texture replication range and aging requirements of the brick and stone restoration area is as follows:

[0070] Based on the maximum depth of the disease predicted by the XG Boost model, combined with the SHAP feature contribution ranking, and superimposed with the anchoring depth and safety margin, the thickness of the rammed earth repair is determined to ensure the strength and stability of the repair structure.

[0071] ,

[0072] Among them, SHAP i : SHAP value of the i-th detection feature (such as 3D morphology parameters, infrared detection data, hyperspectral material parameters, etc.), which is the quantified value of the contribution of the feature to the disease identification result. A positive SHAP value indicates that the feature promotes the identification of the corresponding disease, and a negative value indicates that it inhibits the identification of the corresponding disease. The larger the absolute value, the higher the feature contribution. F represents the set of multi-source detection features, covering core features such as 3D morphology, hyperspectral, infrared, and ultrasonic, corresponding to the multi-source detection data input to the input layer. i: Index of a single detection feature, corresponding to a specific detection index (such as defect depth, density, material composition content, etc.). S: Any subset of the feature set F that does not contain the i-th feature, used to simulate the identification result of the model after removing the feature. M: Total number of features in the feature set F, that is, the total dimension of the multi-source detection features, which is consistent with the feature dimension of the input layer and the feature preprocessing layer. After adding the i-th feature to the feature subset S, the disease identification prediction value (such as the disease type probability and defect quantification parameter prediction value) output by the XGBoost model. The disease identification prediction value output by the feature subset S after it is input into the XGBoost model;

[0073] ,

[0074] Where H represents the final thickness of the rammed earth repair; D dis The maximum depth of the disease after SHAP correction; D a The minimum depth for anchor bolts to be implanted into the rammed earth stabilized base is set by the anchor bolt installation parameters; ΔH safe For safety margin;

[0075] Based on the disease outline output by the XG Boost model, and combined with the SHAP contribution characteristics, the boundary between the rammed earth repair area and the brick and stone repair area is determined. The boundary extends 5-10mm outward to the complete substrate to ensure the stability of the repair overlap.

[0076] Based on the spatial coordinates of the cultural relic itself, spatial registration and point cloud fusion are performed on the standardized fusion data to construct a geometric grid model of the cultural relic itself.

[0077] The actual texture, color, and weathering characteristics of the cultural relic's surface are mapped onto a geometric mesh model to complete the construction of the virtual model of the artifact. Specifically:

[0078] Extract the edge features of the real texture on the surface of cultural relics, including the three-dimensional coordinates, direction angle, and transition gradient of the texture edge;

[0079] Using the spatial coordinates of the artifact itself as the sole reference, the texture edge feature dataset is precisely registered with the edge coordinates of the geometric mesh model. The least squares method is used to calculate the coordinate deviation between the texture edge and the mesh edge, ensuring that the deviation is ≤0.05mm.

[0080] The transition area of ​​the texture edge is smoothed, and the interpolation algorithm is used to fill the tiny gap between the texture edge and the mesh edge, so that the texture edge and the edge of the geometric mesh model are seamlessly connected.

[0081] By combining the complete regional texture patterns in the multi-source fusion data, the consistency of the aligned texture edges is checked to ensure that the direction and spacing of the texture edges are consistent with the real texture edges of the cultural relic.

[0082] Based on standardized and fused data, the defects, cracks, and weathering areas of cultural relics are identified and parametrically labeled. A virtual model of the brick and stone cultural relics is constructed to comprehensively analyze the type, distribution, size, and severity of the defects. Specifically:

[0083] Defect type determination: Based on the standardized fusion dataset, the deep learning defect identification model is used to classify and identify defects by combining three-dimensional morphology, infrared thermal imaging, hyperspectral and ultrasonic detection data, and determine the specific type of defect as one or more of the following: peeling, incompleteness, hollowness, microcracks, internal looseness, salting out or surface weathering.

[0084] Defect spatial distribution: Based on the three-dimensional coordinate system of the virtual model of the brick and stone cultural relic, mark the planar coordinates, depth direction position and continuous distribution range of defects on the surface of the cultural relic;

[0085] Defect size quantification: The geometric parameters of defects are automatically extracted from the point cloud data of the virtual model of the brick and stone cultural relic, and the length, width, maximum depth, projected area and volume of the defect are quantitatively calculated.

[0086] Defect severity classification: Based on defect size, porosity, and impact on structural stability, combined with SHAP feature contribution analysis, defects are classified into three levels: mild, moderate, and severe.

[0087] Adjustments were made based on the internal hollowness and looseness, clarifying the scope of texture replication and aging requirements for the brick and stone repair area, ensuring a stable rammed earth-brick and stone composite structure is formed after repair. Specifically:

[0088] Based on the original rammed earth and brick-stone interface of the cultural relic, and combined with the disease outline and structurally stable area, a digital closed boundary is delineated in the virtual model of the brick-stone cultural relic.

[0089] The maximum depth of the defect and the anchorage thickness of the anchor bolt are taken, and the minimum safe thickness to meet the structural stability is determined by dynamically adjusting the hollow / looseness based on infrared and ultrasonic detection.

[0090] Using the three-dimensional outline of the defective brick and stone area as the boundary, extract the texture features of the surrounding complete brick and stone, and lock the texture replication area in the virtual model of the brick and stone cultural relic.

[0091] Based on the weathering and color parameters obtained from hyperspectral imaging, the color, roughness, and natural aging characteristics of the original bricks and stones are matched to determine the gradient aging parameters.

[0092] Clearly define the scope of texture replication and aging requirements for the brick and stone restoration area to facilitate subsequent use. Determine the boundary between the rammed earth and brick and stone restoration areas to ensure a stable composite structure is formed after restoration.

[0093] Furthermore, texture replication of the defective areas is performed using deep learning, specifically as follows:

[0094] Texture samples of complete areas of the cultural relics were extracted from the standardized fusion dataset, classified according to brick and stone material and weathering degree, and detailed parameters were labeled, including: texture depth, spacing, direction, and weathering traces. These samples were then input into a deep learning disease identification model for training and optimization to ensure that the model can learn texture features of different materials and weathering degrees.

[0095] ,

[0096] Among them, h i,j Z represents the bump depth of the texture at pixel (i,j) (in mm). i,j Z0 is the 3D elevation value of the pixel (unit: mm); Z0 is the elevation value of the texture reference plane (unit: mm), with a value range of 0.01~1.5 mm.

[0097] ,

[0098] Where, d kThe distance between the k-th texture unit and the (k+1)-th texture unit (in mm); (x k ,y k ), (x k+1 ,y k+1 These are the planar coordinates of the centers of the two texture units, respectively.

[0099] ,

[0100] Where W is the weathering coefficient (dimensionless), with a value ranging from 0.3 to 0.9; the smaller the value, the more severe the weathering. w The grayscale value of the weathered area texture; I i The grayscale values ​​represent the texture of the complete, unweathered area.

[0101] Based on a trained deep learning-based defect identification model, and combining the boundary contours and size parameters of the defective region with the texture patterns of the surrounding intact region, the texture of the defective region is replicated:

[0102] ,

[0103] Among them, L min T is the loss value used for model training. pred,n T represents the texture parameter values ​​predicted by the model. true,n Here are the actual texture parameter values, λ is the regularization coefficient, and Ω(f) is the value of the texture parameter in the actual annotation. t ) is the complexity penalty term for the t-th decision tree; through iterative training, the model can accurately learn the texture distribution patterns and feature mapping relationships under different brick and stone materials and different degrees of weathering;

[0104] For areas with regular defects, texture features of the surrounding complete areas are extracted and copied to the areas with regular defects to ensure that the texture direction, spacing, and bump depth are consistent with the original texture.

[0105] ,

[0106] Among them, T defect (x,y) represents the replicated texture parameters at the defect region (x,y), T intact (x+a,y+b) represents the texture parameters of the template corresponding to the surrounding complete area, where a and b are translation offsets to ensure the continuity and consistency between the replicated texture and the original texture.

[0107] For irregularly shaped defect areas, interpolation and texture transfer techniques are used to transfer the texture features of the complete body area to the defect area, restoring the texture details of the defect area, while simulating natural weathering gradients, specifically:

[0108] The basic texture mesh of the defect area is restored using bilinear interpolation. :

[0109] ,

[0110] Where u=x−x0 and v=y−y0 are interpolation weights, and (x0,y0), (x1,y0), (x0,y1), and (x1,y1) are the coordinates of the four surrounding complete texture pixels;

[0111] Texture transfer technology is used to transfer texture features and weathering gradients from the complete area of ​​the body to the defect area:

[0112] ,

[0113] Among them, T source (x,y) represents the complete region texture source features, T target (x,y) represents the initial texture of the defect region, and W(x,y) represents the migration weight; the weathering gradient is simulated using the following formula:

[0114] ,

[0115] Among them, W center is the weathering coefficient at the center of the defect area, k is the weathering gradient coefficient, and D(x,y) is the distance from the pixel to the defect boundary, realizing a natural weathering transition from the defect center to the edge and restoring the weathering details of real bricks and stones. This represents a simulated natural weathering gradient;

[0116] Based on hyperspectral imaging data, weathering time-series features such as color difference gradation, surface hardness decay, and salt precipitation distribution of the brick and stone body are extracted globally. The influence weights of various environmental and material parameters on the weathering gradient are quantified using the SHAP feature contribution formula. Core feature variables that dominate the weathering evolution of the brick and stone are then selected, and a texture time-series weathering evolution model is constructed.

[0117] ,

[0118] Where T(W) represents the replicated texture parameter under the corresponding weathering coefficient, and T0 represents the original complete texture parameter of the brick / stone. η represents the texture parameters of heavily weathered material, W is the measured weathering coefficient of the body, and η is the weathering gradation adjustment coefficient.

[0119] The replicated texture is compared with the texture of the complete area of ​​the original. The texture's bump depth, spacing, and direction errors are detected by the comparison software CloudCompare. At the same time, the color gradient of the texture is calibrated by combining the collected hyperspectral imaging data to make the color of the replicated texture consistent with the weathering color of the original.

[0120] For regular and irregular defect areas, copy matching and interpolation + texture transfer techniques are used respectively to achieve accurate replication of complex texture shapes.

[0121] Furthermore, the repaired area after texture replication is divided into several independent standard components and numbered. The method for associating and saving the repair parameters of each component is as follows:

[0122] After the texture is replicated, the entire brick and stone restoration area is divided into several independent standard brick and stone components according to the structural characteristics of the cultural relic and the ease of construction, and then numbered.

[0123] The segmentation boundary must avoid the core texture area of ​​the cultural relic and the anchor installation position. In the virtual model of the brick and stone cultural relic body of each brick and stone component, the anchor hole position (corresponding one-to-one with the anchor hole of the body), splicing baseline, and aging area (only the outer exposed surface) are marked. At the same time, the number, texture parameters, aging parameters, and size parameters are synchronously associated with the standardized fusion dataset to ensure the traceability of each brick and stone component.

[0124] The restoration area is divided into independent standard components, which facilitates factory prefabrication and on-site assembly, while avoiding the core texture area and anchor bolt location, thus protecting the integrity of the cultural relics.

[0125] Furthermore, a 3D-printed, recyclable, double-layered cavity restoration mold was used, with the surrounding native soil as the raw material. Specifically:

[0126] Based on the outline of the repaired part and the reserved dimensions of the anchor bolt holes in the fused data, a double-layer cavity cooling mold model is designed. The inner forming cavity of the mold fits the outline of the repaired part, and holes matching the thickness of the anchor bolts are reserved at the corresponding anchor bolt positions.

[0127] A 1:1 restoration mold model was constructed using 3D modeling software. The inner cavity of the mold perfectly matched the outer dimensions of the restoration part. It also had pre-set rammed layer grooves and surface texture replicas that were consistent with the original site. During manual ramming, the stratification and texture features of the original site were directly restored.

[0128] The mold is designed as a modular structure with pre-installed plastic elastic demolding clips, and the joints are sealed.

[0129] The soil was prepared using native soil from the surrounding area of ​​the site: it was collected in layers (top, middle, and deep) to avoid compositional deviations caused by the introduction of foreign soil.

[0130] The collected soil is sieved and impurities are removed. Depending on the type of disease, trace amounts of functional additives can be added. For alkali-induced diseases, oxalic acid (desalination agent) + silane coupling agent (salt inhibitor) can be added. Deionized water is added to the prepared soil to adjust it to the optimal moisture content, achieving the best forming state for ramming and ensuring that the outer side of the rammed soil remains flat.

[0131] The 3D-printed mold features pre-designed grooves and textured ridges on its inner wall, directly replicating the original site's layers and textures during the ramming process. The mold is a modular structure, recyclable, and residue-free, preventing secondary pollution. Soil is taken from the surrounding area in layers, with functional additives used to specifically treat damage, ensuring material compatibility and restoration effectiveness.

[0132] Furthermore, layered compaction is performed. The thickness of each layer is determined based on the texture and depth of the inner cavity of the mold. Specifically:

[0133] S81, First layer filling: Fill the repair soil evenly to the thickness of the first layer, and use a fine brush to clean the repair soil in the gaps of the mold texture to ensure that the repair soil completely fills the texture grooves.

[0134] S82, Preliminary compaction: Gently tamp the surface of the repair material with a wooden tamping hammer, tamping evenly from the edge of the mold towards the center;

[0135] S83, Texture Replication: For areas with uneven surface texture, use a rubber pad to fit the texture of the inner wall of the mold, gently press the surface of the repair material to replicate the uneven details that match the texture of the mold, focus on pressing the grooves of the texture to ensure that the repair material and the mold texture are completely in contact.

[0136] S84, Subsequent Layering: Repeat steps S81-S83 above, filling, tamping, and replicating textures layer by layer until the top of the mold is filled. When tamping the last layer, calibrate the overall flatness and texture continuity of the repaired part surface to ensure consistency with the virtual model. At the same time, reserve the integrity of the anchor bolt holes to avoid clogging the holes during the tamping process.

[0137] The layer thickness is adjusted according to the depth of the texture, and a wooden hammer and rubber tamper are used to press and replicate the texture to ensure that the texture is clearly replicated without gaps and with high precision.

[0138] Furthermore, combined with low-temperature curing treatment, the steps for preparing rammed earth repair components are as follows:

[0139] After the entire repair part is compacted, the silicone sealing end cap is fastened and fixed with positioning buckles to ensure the mold is sealed. Granular dry ice is then filled through the mold filling port, with the filling amount being 80% of the cavity gap layer volume (to reserve expansion space). After filling, the silicone low-temperature resistant sealing cap is fastened to prevent the dry ice from sublimating rapidly.

[0140] After standing at room temperature for 2-4 hours, the free water in the soil of the repaired part is frozen by the low temperature of dry ice, and the micro-freezing heave effect of the ice fills the gaps between soil particles.

[0141] Throughout the entire process, the surface temperature of the repaired parts is controlled to be no lower than -10℃ to avoid excessive low temperature causing soil cracking.

[0142] After low-temperature curing is completed, open the sealing cap of the mold filling port to allow the CO2 gas in the cavity gap layer to be discharged naturally, and allow the mold and the repaired part to naturally warm up at room temperature (15~25℃).

[0143] After the temperature recovery is complete, remove the sealing end cap and connect the repair part to the main body according to the pre-set anchor bolt hole alignment point;

[0144] The anchor rods are made of lightweight and high-strength materials, and the surface of the anchor rods is micro-abraded to enhance the adhesion between the anchor rods and the original soil of the repair soil and the original soil of the body. An environmentally friendly anti-corrosion coating is applied to the surface of the anchor rods, and inorganic materials are used for micro-grouting at the joints to enhance the adhesion between the repair body and the original body.

[0145] Dry ice low-temperature solidification utilizes the micro-freeze-swell effect of ice to fill the gaps between soil particles, improving density and strength, and preventing soil cracking.

[0146] Furthermore, based on the parameters output from the virtual model of the brick and stone cultural relic, brick and stone components were prepared, and CNC laser engraving technology was used to replicate the texture and perform gradient aging on the exposed outer surface of the components. The specific steps are as follows:

[0147] Antique brick and stone materials that are consistent with the material of the cultural relic are selected, and individual brick and stone components are cut and prepared according to the size parameters of the virtual model of the brick and stone cultural relic.

[0148] After cutting, the splicing surfaces of the brick and stone components are polished to ensure that the interlocking structure of the splicing surfaces is consistent with the virtual model, and at the same time, surface dust and burrs are cleaned.

[0149] The virtual model and texture parameters of each brick and stone artifact are imported into a CNC laser engraving device, and texture engraving is performed on the exposed outer surface of the brick and stone artifact.

[0150] Based on the hardness of the brick and stone material, the laser engraving parameters are adjusted to ensure that the engraving depth matches the texture depth in the virtual model of the brick and stone artifact.

[0151] The brick and stone components are fixed on the carving workbench, and the brick and stone components are precisely aligned with the virtual model through the CCD vision positioning system.

[0152] The process employs a layered carving method, first carving the outline and main structure of the texture, and then carving the details of the texture. The carving effect is monitored in real time during the carving process, and the carving parameters are adjusted in a timely manner if texture deviation occurs.

[0153] For areas with a texture depth of 0.5mm ≤ 1.5mm, they are considered to have a large texture depth, and the number of engraving layers is increased to ensure that the three-dimensionality of the texture is consistent with the body.

[0154] After the carving is completed, use a high-pressure blower to blow away the carving dust on the surface of the brick and stone components, and then use a fine brush to clean the residual dust in the texture gaps to ensure that the texture is clear and free of dust residue. At the same time, check the texture replication effect and compare it with the virtual model.

[0155] Based on the aging gradient parameters output by a deep learning model, combined with the weathering degree of the artifact itself, CNC laser engraving equipment is used to perform targeted aging treatment on the exposed outer surfaces of the brick and stone components. This simulates the effects of natural weathering, wear, and erosion without altering the internal structure and splicing precision of the brick and stone components. Specifically:

[0156] Different laser aging parameters are set according to the weathering level of the cultural relic;

[0157] Based on the weathering gradient of the exposed outer surface of the brick and stone components, laser aging is carried out in different areas to create a natural transition from lightly weathered areas to heavily weathered areas.

[0158] The antique-style bricks and stones are made of the same material as the original cultural relics. Based on CCD visual positioning and layered carving operations, the texture carving is highly accurate and has a strong three-dimensional effect.

[0159] Furthermore, all the brick and stone components are pre-assembled to form a complete brick and stone restoration body. The specific method is as follows:

[0160] Arrange the aged brick and stone components in numerical order and check the splicing surfaces and anchor bolt holes of each component;

[0161] Prepare the materials for assembly; the original soil used should be the same as that used in the rammed earth repair.

[0162] Following the assembly sequence of the virtual model of the brick and stone cultural relic, the brick and stone components are assembled layer by layer from bottom to top. The splicing material is evenly applied to the splicing surfaces of the brick and stone components to ensure that the splicing surfaces are completely covered without any omissions.

[0163] Align adjacent brick and stone components according to the splicing baseline, so that the splicing surfaces fit together and the anchor bolt holes are aligned.

[0164] Each layer is assembled and fixed with clamps. Once the assembled material has initially solidified, the next layer is assembled.

[0165] After all the brick and stone components are assembled and the assembly materials have completely solidified, anchor rods are driven into the preset anchor rod holes. The anchor rods are made of lightweight and high-strength carbon fiber material, and their diameter matches the size of the holes.

[0166] Before the anchor bolts are installed, the surface is micro-sanded to enhance the adhesion with the splicing materials and rammed earth, and an environmentally friendly anti-corrosion coating is applied at the same time.

[0167] During implantation, the top of the anchor rod is tapped with a rubber mallet to ensure that the implantation depth of the anchor rod is consistent with the virtual model of the brick and stone cultural relic, and the length of the anchor rod extending out of the brick and stone component matches the depth of the anchor hole in the rammed earth restoration area. After implantation, inorganic materials are used to micro-grout the gap between the anchor rod and the hole to compact and fix it.

[0168] Assemble the components layer by layer according to their numbers. Fixtures are used to prevent misalignment, ensuring a secure connection and facilitating installation.

[0169] Furthermore, the rammed earth restoration components and the brick and stone restoration body are assembled on-site at the site to form a stable rammed earth-brick and stone composite structure. The specific method is as follows:

[0170] Clean the surface dust of the rammed earth repair area, check the flatness of the rammed earth surface and the anchor bolt hole positions to ensure that it matches the anchor bolts and splicing surfaces of the brick and stone repair body; apply a thin, even layer of inorganic mortar material to the splicing area of ​​the rammed earth surface to enhance the adhesion.

[0171] Using hoisting equipment, the assembled brick and stone restoration body is hoisted to the preset position and aligned with the rammed earth restoration area according to the splicing baseline. This ensures that the anchor rods of the brick and stone restoration body are inserted into the anchor holes in the rammed earth area, and that the splicing surfaces fit together without gaps. During the assembly process, a total station is used to monitor the assembly accuracy in real time and adjust the position of the brick and stone restoration body to ensure that the overall flatness is consistent with the original cultural relic.

[0172] After assembly, inorganic materials are used for micro-grouting at the joints between the brick and stone restoration and the rammed earth, so that the joints can be naturally integrated with the surrounding structure.

[0173] After assembly, non-woven fabric bandages are used to flexibly fix the connecting parts, and the whole body is cured at room temperature for 7 to 14 days. After the curing period, the flexible fixing device is removed and the exposed outer surface of the brick and stone restoration is cleaned.

[0174] Real-time monitoring of assembly accuracy ensures that the repaired body is as flat as the original body, and guarantees stable curing of the connection points.

[0175] The present invention also provides a 3D-printed recyclable double-layer cavity repair mold for use in the method described herein, comprising:

[0176] The inner mold has a forming cavity that fits the outline of the rammed earth repair component. The inner mold has pre-reserved anchor holes at the corresponding anchor installation positions, and the inner wall is provided with grooves for replicating the rammed earth layering and raised textures for replicating the surface texture.

[0177] The outer mold is fitted over the inner mold to provide structural rigidity for the mold.

[0178] A cavity gap layer, formed in the closed gap between the inner mold and the outer mold, is used to fill the refrigerant to achieve low-temperature curing.

[0179] The inner mold conforms to the contour of the repair part and has pre-drilled anchor holes. The inner wall has grooves / textures, which directly replicate the layering and texture during ramming. The outer mold provides structural rigidity, and the cavity gap layer is used to fill the refrigerant to achieve low-temperature curing.

[0180] The present invention also provides a repair structure for use in the method described herein, comprising:

[0181] The intermediate rammed earth layer is formed by layering and ramming the original soil around the site and then curing it at low temperature. Its outer surface is kept flat, and its inner surface is connected to the site body. Anchor rods are pre-embedded in the intermediate rammed earth layer, and one end of the anchor rods is anchored inside the site body.

[0182] The outer brick and stone layer covers the outer surface of the middle rammed earth layer. The outer brick and stone layer is composed of several independent brick and stone components. The exposed outer surface of each brick and stone component is decorated with an antique texture formed by CNC laser engraving, which is integrated with the surface texture and weathering degree of the site.

[0183] The intermediate rammed earth layer and the outer brick and stone layer are fixedly connected by anchor bolts and / or bonding materials to form a stable rammed earth-brick and stone composite structure.

[0184] The intermediate rammed earth layer is constructed using native soil and cured at low temperatures, with a smooth outer surface to provide a stable base for the brick and stone layers. The brick and stone layers are assembled from independent components, with the exposed outer surfaces laser-engraved to create an antique texture that blends naturally with the main body. Attached Figure Description

[0185] Figure 1 This is a flowchart illustrating the 3D printing-based method for the accurate restoration of earthen archaeological sites, which involves identifying and fusing aging techniques.

[0186] Figure 2 This is a schematic diagram illustrating the state of the three-dimensional morphological data of the damaged parts of the site and the surrounding area in the 3D printing-based authentic restoration method for identifying and fusing aging effects of earthen site defects according to the present invention. Detailed Implementation

[0187] Embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0188] In the description of this invention, it should be understood that the terms "longitudinal", "lateral", "up", "down", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0189] In the description of this invention, unless otherwise specified and limited, it should be noted that the terms "installation", "connection" and "linking" should be interpreted broadly. For example, they can refer to mechanical or electrical connections, or internal connections between two components. They can be direct connections or indirect connections through an intermediate medium. Those skilled in the art can understand the specific meaning of the above terms according to the specific circumstances.

[0190] This invention discloses a 3D printing-based method for the accurate identification and aging of rammed earth archaeological sites. Addressing the problems in existing artifact restoration methods, such as subjective identification of defects, low quantitative accuracy, inaccurate replication of brick and stone surface textures, harsh aging effects with poor integration with the original structure, and the lack of targeted modular splicing in restoration, which fails to create a stable rammed earth-brick composite structure, this invention provides an integrated closed-loop restoration method encompassing precise defect identification and quantification, accurate texture replication, laser modular aging, factory pre-assembly, and precise on-site assembly. This method constructs a stable rammed earth-brick composite structure, improving restoration fidelity and construction efficiency while avoiding secondary damage to the artifacts.

[0191] The technical solution of this invention centers on "disease identification - virtual repair - rammed earth repair - brick and stone texture processing - modular assembly - on-site assembly," first completing the rammed earth repair (keeping the outer side flat) to form a rammed earth-brick and stone composite structure. For example... Figure 1 As shown, the 3D printing-based method for authentic restoration of earthen archaeological sites, which involves identifying and integrating aging techniques, includes the following steps:

[0192] Based on the field survey, and combined with the material characteristics and damage features of brick and stone cultural relics, multi-source information was collected on the site itself. The collected multi-source heterogeneous data was then fused and processed to output a standardized fused dataset.

[0193] The standardized fusion dataset is input into the deep learning disease identification model for disease identification and parameter quantification. Based on the standardized fusion dataset and disease identification results, a virtual model of the brick and stone cultural relic is constructed to determine the thickness of the rammed earth restoration, as well as the boundaries of the rammed earth restoration area and the brick and stone restoration areas on both sides, and to clarify the texture replication range and aging requirements of the brick and stone restoration area.

[0194] Deep learning is used to replicate the texture of the defective area. The texture-replicated brick and stone repair area is divided into several independent standard component models and numbered. The texture parameters and repair parameters of each component model are associated and saved.

[0195] Based on the rammed earth restoration area, a recyclable double-layer cavity restoration mold is obtained by 3D printing (existing 3D printing methods can be used). Using the original soil around the site as raw material, rammed earth restoration parts are prepared by layered ramming and combined with low temperature curing treatment.

[0196] Based on the component model, texture parameters, and repair parameters of the brick and stone repair area segmentation, brick and stone repair parts are prepared, and CNC laser engraving technology is used to perform texture replication and gradient aging treatment on the exposed surface of the component.

[0197] All brick and stone components are pre-assembled to form a complete brick and stone restoration body;

[0198] The rammed earth restoration components and brick and stone restoration components were assembled on-site to form a stable rammed earth-brick and stone composite structure.

[0199] In a preferred embodiment of the present invention, multi-source information is collected on the site itself, and the collected heterogeneous multi-source data is fused to output a standardized fused dataset, specifically as follows:

[0200] Anchor bolt holes are pre-drilled in the main body of the site; the parameters of the anchor bolts to be driven into the main body are confirmed in advance, and the actual physical parameters of the anchor bolts are used as the sole reference in subsequent virtual modeling, anchor hole reservation of brick and stone components, etc., to avoid positioning deviations and anchor hole misalignment caused by scanning and modeling first and then drilling and anchoring, thus avoiding errors.

[0201] like Figure 2 As shown, a 3D laser scanner or close-range photogrammetry technique is used to collect 3D morphological data of the damaged parts of the site and the surrounding undamaged areas. This identifies the location, area, and size of surface peeling and defects, as well as the surface texture of the intact areas. For defects that are smaller on the outside and larger on the inside, single laser scanning is limited by line-of-sight obstruction and cannot scan the internal voids. It can only obtain the shape of the external opening and cannot fully characterize the true volume of the defect. It can be combined with close-range photogrammetry, infrared thermal imaging internal detection, and ultrasonic density detection. Through multi-source heterogeneous data fusion, the visual blind spots of laser scanning can be filled, and the internal depth, cavity range, and erosion volume of defects that are smaller on the outside and larger on the inside can be calculated in reverse. This fully restores the parameters of the defects in all dimensions and ensures the accuracy of defect identification and repair design.

[0202] For example, a 3D laser scanner, such as the Faro Focus S350, can be used to scan the damaged parts of the site and a certain range (e.g., 0.5cm) of the surrounding complete area to obtain high-precision point cloud data; at the same time, a camera can be used to take pictures of the damaged area from multiple different angles, and 3D reconstruction can be performed using Agisoft Metashape software to obtain a textured surface model, and identify the location, area, size, and surface texture of the surface peeling and missing parts.

[0203] For defects that are smaller on the outside and larger on the inside, taking infrared thermal imaging for internal detection as an example, an infrared thermal imager is used to continuously capture thermal images of the defect area to obtain temperature field distribution data. Using the average temperature of the intact area as a benchmark, areas with a temperature difference ≥1℃ are identified as suspected internal defect areas. Combined with a pre-set temperature difference-depth calibration curve (calibrated based on the thermal conductivity characteristics of rammed earth / brick and stone sites), the depth range and boundary contour of the defect are initially obtained. An ultrasonic detector is used to detect the wave velocity in the defect area. Based on the correlation between the density of the rammed earth / brick and stone site and the wave velocity (e.g., wave velocity <2000m / s is considered a defect area), the density distribution at each measuring point is calculated to determine the range and wall thickness of the internal cavity.

[0204] Using the coordinate system of the laser scanning point cloud as a reference, the iterative nearest point (ICP) algorithm is used to uniformly register the near-field photogrammetry point cloud, infrared thermal imaging temperature field data, and ultrasonic wave velocity data into the same three-dimensional spatial coordinate system, thereby achieving spatial alignment of multi-source data. The external opening shape, the depth range of infrared thermal imaging, and the density data of ultrasonic detection are input into the preset reverse calculation model. Through the three-dimensional surface fitting algorithm, the internal cavity shape of the defect is reconstructed, and the full-dimensional defect parameters such as internal depth, cavity range, and erosion volume are calculated.

[0205] Based on the three-dimensional morphological data of the site itself, and combined with the stress characteristics of the rammed earth-brick composite structure and the anchoring requirements of the restoration, the anchor installation parameters, including the anchor installation location, thickness, and insertion depth, are set through structural stress simulation and spatial point adaptation to ensure that the anchor matches the site structure. Specifically:

[0206] Based on the three-dimensional morphological data of the site itself, the safe installation points for anchor bolts were located, avoiding defective areas. Specifically:

[0207] The deep, dense, original matrix region below the surface weathered layer and loose layer.

[0208] Completely avoid areas of the solid body where cracks, hollows, erosion, and other defects are distributed.

[0209] A solid transitional base section of a rammed earth-brick composite structure.

[0210] There are no hidden cracks or potential localized damage in the surrounding area.

[0211] Based on the dimensions and weight of the masonry restoration, the stress threshold is calculated, and the anchor bolt diameter is matched (e.g., 8-12mm diameter for carbon fiber anchor bolts).

[0212] Calculate the self-weight of the masonry restoration:

[0213] ,

[0214] Where G represents the total self-weight of the brick and stone restoration; ρ is the density of the antique brick and stone material; V is the three-dimensional volume of the brick and stone restoration; and g is the acceleration due to gravity.

[0215] Calculate the combined stress load:

[0216] ,

[0217] Among them, F 总 : Total stress threshold for anchor bolt design; k: Safety factor for cultural relic restoration, ranging from 1.2 to 1.5, which is the minimum stress bearing threshold that the anchor bolt must meet;

[0218] Match anchor bolt specifications according to stress threshold:

[0219] ,

[0220] Where A: effective cross-sectional area of ​​the anchor bolt; [σ]: allowable stress of the anchor bolt material;

[0221] The diameter of the anchor rod is calculated from the cross-sectional area, and the final anchor rod thickness specification is determined.

[0222] Based on the thickness of the rammed earth repair and the depth of the body stabilization layer, the anchor bolt implantation depth is set by three-dimensional spatial coordinate calibration to ensure that the implanted end penetrates into the body stabilization layer and the exposed end is compatible with the anchor hole of the brick and stone repair body.

[0223] Finally, all anchor bolt parameters were spatially registered and verified with the three-dimensional morphological data of the body to complete the parameter setting; an infrared thermal imager was used to detect the location and depth of hollow areas, hidden cracks and differences in water content inside the archaeological artifacts, and to identify loose and eroded areas inside, providing a basis for controlling the thickness and density of rammed earth restoration.

[0224] The composition of the soil (rammed earth area) on the surface of brick and stone cultural relics, the composition of brick and stone materials, the degree of weathering, and the distribution range of salt precipitation areas were analyzed by using a hyperspectral imager to determine the color gradient and weathering parameters of texture replication, providing data support for aging treatment;

[0225] Ultrasonic compaction testing was used to determine the compaction of the brick and stone body and the rammed earth area, identify the loose areas inside, and provide a basis for the layered ramming force and low-temperature curing parameters of rammed earth repair.

[0226] Collect dimensional tolerances, splicing gaps, and bonding parameters of the subgrade components, and combine them with anchor bolt installation requirements to provide data support for virtual segmentation numbering and factory assembly;

[0227] The collected multi-source heterogeneous data were normalized, noise was removed, and missing values ​​were filled in. Core correlation features for the restoration of brick and stone cultural relics were extracted from the preprocessed multi-source data (3D topographic point clouds obtained through 3D laser scanning or close-range photogrammetry; internal density obtained through infrared imaging and ultrasonic testing, etc.). These features include: geometric morphological features: surface texture, defect outline, splicing tolerance, and anchoring coordinates; material physicochemical features: brick and stone matrix composition, weathering grade, moisture content, porosity, and density; structural mechanical features: internal defect location, hollow area, crack depth, and anchoring stress threshold; process adaptation features: restoration material matching parameters, anchoring depth, and splicing gap; and appearance fidelity features: brick and stone surface color, weathering texture gradient, and aging matching parameters.

[0228] Using the spatial coordinates of the three-dimensional morphological data of cultural relics as a benchmark, the remaining data are spatially / dimensionally registered with the morphological features to ensure one-to-one correspondence between the data and output a standardized fusion dataset.

[0229] In a preferred embodiment of the present invention, the deep learning disease identification model adopts the XGBoost model and combines it with SHAP interpretability analysis. The deep learning disease identification model includes an input layer (multi-source detection features), a feature preprocessing layer (normalization, missing value imputation), an XGBoost classification and regression layer (based on the number of learners X, maximum depth Y, and learning rate Z), a SHAP feature contribution analysis layer, and an output layer (disease type, quantification parameters, and repair parameters) connected in sequence.

[0230] Input layer: Receives multi-source detection data, covering core data such as site three-dimensional morphology data, hyperspectral material data, infrared detection data, and ultrasonic density, and captures key information such as surface texture, material composition, and internal defects, all in a standardized digital format to ensure compatibility with subsequent preprocessing layers;

[0231] Feature preprocessing layer:

[0232] The min-max normalization algorithm is used to map the input multi-source detection features (such as texture parameters and density data) to the [0,1] interval, thereby eliminating the dimensional differences between data of different dimensions;

[0233] To address potential missing items in multi-source data, nearest neighbor interpolation is used to fill them in, prioritizing the preservation of the original data's authenticity and avoiding the impact of missing data on disease identification accuracy.

[0234] XGBoost Classification and Regression Layers:

[0235] The algorithm is based on a gradient boosting tree and contains X basic learners (for example, X can be 100-150 to meet the accuracy requirements of site disease identification). All learners are run in series to gradually optimize the identification accuracy.

[0236] The maximum depth Y is set to 3-5 (to avoid overfitting and adapt to the complexity of the site's damage), and the learning rate Z is set to 0.1-0.3 (to ensure model convergence and avoid gradient vanishing).

[0237] Functional modules:

[0238] The input preprocessed features are classified to identify core defects (surface peeling, internal looseness, cracks, salt precipitation, etc.).

[0239] The size and severity of the defects are quantitatively calculated, and specific values ​​(such as defect area and depth) are output to provide data support for setting subsequent repair parameters.

[0240] Receive the SHAP feature contribution analysis results, dynamically adjust the learner parameters, and improve the accuracy and stability of disease identification.

[0241] To quantify the contribution of different detection features to the disease identification results, the SHAP analysis method is introduced to conduct interpretability analysis on the trained disease identification model.

[0242] SHAP Feature Contribution Analysis Layer:

[0243] Extract all features from the XGBoost classification and regression layer outputs, and clarify the influence weight of each feature (such as texture parameters and density) on disease identification;

[0244] The contribution of each detection feature is quantified using SHAP values ​​to select the core features most critical for disease identification (such as the density data of hollow areas) and eliminate redundant features.

[0245] The feature contribution results are fed back to the XGBoost classification and regression layers to optimize model parameters, while the feature contribution ranking is output to ensure consistency with the interpretability of the overall repair method.

[0246] Output layer:

[0247] Clearly identify the core diseases (surface peeling, internal loosening, cracks, salt precipitation, etc.), and label the specific disease name and its category;

[0248] Output specific quantitative data on the defects, including defect area, depth, and distribution location, which correspond to the parameter requirements for subsequent repair thickness and texture replication;

[0249] The standardized fusion dataset is input into the deep learning disease identification model for disease identification and parameter quantification, specifically as follows:

[0250] The detection features, such as 3D morphology, infrared imaging, spectral detection, and substrate material, are normalized and then input into the XGBoost model.

[0251] The XGBoost model automatically classifies disease types, including efflorescence, spalling, cracking, and erosion, and outputs quantitative parameters, including disease location, area, and depth.

[0252] The contribution of each disease type and quantitative parameter is calculated by SHAP analysis, redundant features are eliminated, the identification accuracy is optimized, and the reliability of disease identification and quantification results is ensured.

[0253] The system outputs the disease type and ranks the contribution of quantitative parameters to provide data support for setting subsequent repair parameters.

[0254] In a preferred embodiment of the present invention, a virtual model of the brick and stone cultural relic is constructed based on a standardized fusion dataset and the results of disease identification. The thickness of the rammed earth restoration, as well as the boundaries of the rammed earth restoration area and the brick and stone restoration areas on both sides, are determined. The method for clarifying the texture replication range and aging requirements of the brick and stone restoration area is as follows:

[0255] The maximum depth of the defect is predicted based on the XG Boost model (the rammed earth repair thickness is determined by combining the maximum defect depth with the anchor bolt thickness, dynamically adjusted based on the degree of hollowness / looseness detected by infrared and ultrasonic testing, to determine the minimum safe thickness to ensure structural stability). This is combined with the SHAP feature contribution ranking (the greater the feature contribution, the greater the maximum defect depth), and the anchor bolt insertion depth and safety margin are superimposed to determine the rammed earth repair thickness, ensuring the strength and stability of the repaired structure.

[0256] ,

[0257] Among them, SHAP i: SHAP value of the i-th detection feature (such as 3D morphology parameters, infrared detection data, hyperspectral material parameters, etc.), which is the quantified value of the contribution of the feature to the disease identification result. A positive SHAP value indicates that the feature promotes the identification of the corresponding disease, and a negative value indicates that it inhibits the identification of the corresponding disease. The larger the absolute value, the higher the feature contribution. F represents the set of multi-source detection features, covering core features such as 3D morphology, hyperspectral, infrared, and ultrasonic, corresponding to the multi-source detection data input to the input layer. i: Index of a single detection feature, corresponding to a specific detection index (such as defect depth, density, material composition content, etc.). S: Any subset of the feature set F that does not contain the i-th feature, used to simulate the identification result of the model after removing the feature. M: Total number of features in the feature set F, that is, the total dimension of the multi-source detection features, which is consistent with the feature dimension of the input layer and the feature preprocessing layer. After adding the i-th feature to the feature subset S, the disease identification prediction value (such as the disease type probability and defect quantification parameter prediction value) output by the XGBoost model. The disease identification prediction value output by the feature subset S after it is input into the XGBoost model;

[0258] ,

[0259] Where H represents the final thickness of the rammed earth repair; D dis The maximum depth of the disease after SHAP correction (unit: mm); D a The minimum depth (in mm) for anchor bolts to be implanted into the rammed earth stabilized matrix is ​​set by the anchor bolt installation parameters. In cultural relic restoration scenarios, it is generally ≥50mm to ensure the anchor bolt's grip strength and pull-out bearing capacity; ΔH safe As a safety margin (unit: mm), it is generally taken as 10~20mm to offset the effects of construction errors, minor environmental disturbances and long-term weathering.

[0260] Based on the defect contours output by the XG Boost model, and combined with the SHAP contribution characteristics, the boundary between the rammed earth restoration area and the brick and stone restoration area was determined (based on a standardized fusion dataset, quantitative analysis was performed using the XG Boost-SHAP defect identification model, and defect information was annotated in the virtual model with three-dimensional coordinates, and classified into light, medium, and severe levels according to defect size and structural impact). Based on the original rammed earth-brick and stone interface of the cultural relic, and combined with the defect contours and structurally stable areas, a digital closed boundary was delineated in the virtual model), with the boundary extending 5-10mm outward to the intact substrate to ensure the stability of the repair joint.

[0261] The XGBoost model relies on multi-source fusion features:

[0262] On the one hand, it identifies and distinguishes the material boundary contours of rammed earth and brick masonry substrates; on the other hand, it outputs the closed contour lines of diseases such as spalling, loosening, and cracking, and preliminarily delineates the disease coverage area and the original boundary line between the two substrates.

[0263] Through SHAP contribution analysis, high contribution distinguishing features were extracted: core indicators such as rammed earth porosity, density, surface weathering parameters, brick and stone texture features, and material hardness.

[0264] Using the outer contour of the lesion output by XGBoost as a constraint, and combining the actual material boundary between rammed earth and brick / stone determined by SHAP, the initial boundary division of the rammed earth repair zone and the brick / stone repair zone is generated, specifically as follows:

[0265] Input the standardized fusion dataset into the trained XGBoost disease identification model;

[0266] By setting a preset probability threshold (e.g., a probability ≥ 0.8 is used to determine a disease point), disease points are clustered and edge detected to generate a closed outer contour line of the disease, which serves as the minimum range constraint that the repair area must cover, ensuring that there are no disease residues.

[0267] Using the SHAP feature contribution formula, the contribution of each input feature to "material classification (rammed earth / brick and stone)" is calculated, core distinguishing features are selected, and low-contribution interfering features are removed.

[0268] Correct the material boundaries in the initial output of XGBoost (such as removing false boundaries caused by surface salting and weathering layers) to obtain the true rammed earth-brick boundary lines determined only by the properties of the matrix material.

[0269] Based on the spatial coordinates of the cultural relic itself, spatial registration and point cloud fusion are performed on the standardized fusion data to construct a geometric grid model of the cultural relic itself.

[0270] The actual texture, color, and weathering characteristics of the cultural relic's surface are mapped onto a geometric mesh model to complete the construction of the virtual model of the artifact. Specifically:

[0271] Extract the edge features of the real texture on the surface of cultural relics, including the three-dimensional coordinates, direction angle, and transition gradient of the texture edge;

[0272] Using the spatial coordinates of the artifact itself as the sole reference, the texture edge feature dataset is precisely registered with the edge coordinates of the geometric mesh model. The least squares method is used to calculate the coordinate deviation between the texture edge and the mesh edge, ensuring that the deviation is ≤0.05mm.

[0273] The transition area of ​​the texture edge is smoothed, and the interpolation algorithm is used to fill the tiny gap between the texture edge and the mesh edge, so that the texture edge and the edge of the geometric mesh model are seamlessly connected.

[0274] By combining the complete regional texture patterns in the multi-source fusion data, the consistency of the aligned texture edges is checked to ensure that the direction and spacing of the texture edges are consistent with the real texture edges of the cultural relic.

[0275] Based on standardized and fused data, the defects, cracks, and weathering areas of cultural relics are identified and parametrically labeled. A virtual model of the brick and stone cultural relics is constructed to comprehensively analyze the type, distribution, size, and severity of the defects. Specifically:

[0276] Based on a standardized fusion dataset, a deep learning disease identification model is used to classify and identify defects by combining three-dimensional morphology, infrared thermal imaging, hyperspectral and ultrasonic detection data, and determine the specific type of defect as one or more of the following: peeling, incompleteness, hollowness, microcracks, internal looseness, salting out or surface weathering.

[0277] Using the three-dimensional coordinate system of the virtual model of the brick and stone cultural relic as a reference, the planar coordinates, depth position and continuous distribution range of defects on the surface of the cultural relic are marked.

[0278] The geometric parameters of defects are automatically extracted from the point cloud data of the virtual model of the brick and stone cultural relics, and the length, width, maximum depth, projected area and volume of the defects are quantitatively calculated.

[0279] Based on the defect size, porosity, and impact on structural stability, and combined with SHAP feature contribution analysis, defects are classified into three levels: mild, moderate, and severe.

[0280] Adjustments were made based on the internal hollowness and looseness, clarifying the scope of texture replication and aging requirements for the brick and stone repair area, ensuring a stable rammed earth-brick and stone composite structure is formed after repair. Specifically:

[0281] Based on the original rammed earth and brick-stone interface of the cultural relic, and combined with the disease outline and structurally stable area, a digital closed boundary is delineated in the virtual model of the brick-stone cultural relic.

[0282] The maximum depth of the defect and the anchorage thickness of the anchor bolt are taken, and the minimum safe thickness to meet the structural stability is determined by dynamically adjusting the hollow / looseness based on infrared and ultrasonic detection.

[0283] Using the three-dimensional outline of the defective brick and stone area as the boundary, extract the texture features of the surrounding complete brick and stone, and lock the texture replication area in the virtual model of the brick and stone cultural relic.

[0284] Based on the weathering and color parameters obtained from hyperspectral imaging, the color, roughness, and natural aging characteristics of the original bricks and stones are matched to determine the gradient aging parameters.

[0285] In a preferred embodiment of the present invention, texture replication of the defective region is performed using deep learning, specifically as follows:

[0286] Texture samples of complete areas of the cultural relics were extracted from the standardized fusion dataset, classified according to brick and stone material and weathering degree, and detailed parameters were labeled, including: texture depth, spacing, direction, and weathering traces. These samples were then input into a deep learning disease identification model for training and optimization to ensure that the model can learn texture features of different materials and weathering degrees.

[0287] ,

[0288] Among them, h i,j Z represents the bump depth of the texture at pixel (i,j) (in mm). i,j Z0 is the 3D elevation value of the pixel (unit: mm); Z0 is the elevation value of the texture reference plane (unit: mm), with a value range of 0.01~1.5 mm.

[0289] ,

[0290] Where, d k The distance between the k-th texture unit and the (k+1)-th texture unit (in mm); (x k ,y k ), (x k+1 ,y k+1 These are the planar coordinates of the centers of the two texture units, respectively.

[0291] ,

[0292] Where W is the weathering coefficient (dimensionless), with a value ranging from 0.3 to 0.9; the smaller the value, the more severe the weathering. w The grayscale value of the weathered area texture; I i The grayscale values ​​represent the texture of the complete, unweathered area.

[0293] Based on a trained deep learning-based disease identification model, and combining the boundary contours and size parameters of the defective region with the texture patterns of the surrounding intact region, texture replication is performed on the defective region. The annotated texture feature parameter library is then input into the deep learning-based disease identification model (XGBoost model) for supervised training and optimization. The model training objective is to minimize the texture feature prediction error.

[0294] ,

[0295] Among them, L min T is the loss value used for model training. pred,n T represents the texture parameter values ​​predicted by the model. true,nHere are the actual texture parameter values, λ is the regularization coefficient, and Ω(f) is the value of the texture parameter in the actual annotation. t ) is the complexity penalty term for the t-th decision tree; through iterative training, the model can accurately learn the texture distribution patterns and feature mapping relationships under different brick and stone materials and different degrees of weathering;

[0296] For regular defect areas (referring to geometrically regular, straight-boundary, and symmetrically shaped defects on the surface of bricks and stones, including rectangular, square, and strip-shaped defects with standard geometric features, such as regular block-shaped missing parts), a feature matching and copying method is used. This method extracts the texture features of the surrounding complete area and copies and matches them to the regular defect area, ensuring that the texture direction, spacing, and depth of concavity and convexity are consistent with the original body.

[0297] ,

[0298] Among them, T defect (x,y) represents the replicated texture parameters at the defect region (x,y), including bump depth, texture spacing, orientation angle, weathering coefficient, and color characteristics, used to fully characterize the texture morphology and weathering state at this location; T intact (x+a,y+b) represents the texture parameters of the template corresponding to the surrounding complete region, and a and b are the translation offsets to ensure the continuity and consistency between the replicated texture and the original texture. The translation offsets a and b are solved by boundary feature matching and the least squares method. The goal is to minimize the texture parameter error of the common boundary between the defective region and the complete region to determine the optimal translation amount, so as to ensure the continuity and overall consistency between the replicated texture and the original texture at the boundary.

[0299] For irregularly shaped defect areas (referring to defect areas on the brick and stone surface with irregular outlines, curved / broken / non-fixed geometric shapes, asymmetrical sizes, and complex shapes, including missing corners, local erosion, irregular peeling, and other defective parts without standard geometric features), interpolation and texture transfer techniques are used to transfer the texture features of the complete area of ​​the brick / stone to the defect area, restoring the texture details of the defect area, while simulating the natural weathering gradient, specifically:

[0300] The basic texture mesh of the defect area is restored using bilinear interpolation. :

[0301] ,

[0302] Where u=x−x0 and v=y−y0 are interpolation weights, and (x0,y0), (x1,y0), (x0,y1), and (x1,y1) are the coordinates of the four surrounding complete texture pixels;

[0303] Then, using texture transfer technology, the texture features and weathering gradients of the complete area of ​​the body are transferred to the defective area:

[0304] ,

[0305] Among them, T source (x,y) represents the complete region texture source features, T target (x,y) represents the initial texture of the defect region, and W(x,y) represents the migration weight; the weathering gradient is simulated using the following formula:

[0306] ,

[0307] Among them, W center is the weathering coefficient at the center of the defect area, k is the weathering gradient coefficient, and D(x,y) is the distance from the pixel to the defect boundary, realizing a natural weathering transition from the defect center to the edge and restoring the weathering details of real bricks and stones. This represents a simulated natural weathering gradient;

[0308] Based on hyperspectral imaging data, weathering time-series features such as color difference gradation, surface hardness decay, and salt precipitation distribution of the brick and stone body are extracted globally. The influence weights of various environmental and material parameters on the weathering gradient are quantified using the SHAP feature contribution formula. Core feature variables that dominate the weathering evolution of the brick and stone are then selected, and a texture time-series weathering evolution model is constructed.

[0309] ,

[0310] Where T(W) represents the replicated texture parameter under the corresponding weathering coefficient, and T0 represents the original complete texture parameter of the brick / stone. Here, W represents the measured weathering coefficient of the original artifact, and η represents the weathering gradient adjustment coefficient. Based on this model, the defect area is replicated in a zoned and differentiated manner. High-weathered textures are replicated in the overlapping boundary area near the intact substrate, while moderately weathered textures are matched to the central area of ​​the defect, following the true deterioration pattern of cultural relics: "heavier weathering at the edges and lighter weathering at the interior." Simultaneously, combined with the weathering gradient simulation formula, a natural gradient transition effect completely consistent with the original artifact is replicated. Furthermore, hyperspectral color calibration eliminates color difference deviations, ensuring that the restored texture not only achieves geometric consistency but also achieves three-dimensional synchronous fidelity in weathering sequence, texture layering, and color gradient. This solves the industry problem of traditional replications lacking weathering transition and exhibiting obvious artificial traces.

[0311] The replicated texture is compared with the texture of the complete area of ​​the original. The software CloudCompare is used to compare the replicated texture model with the original texture model and complete the spatial coordinate registration; the software automatically extracts and compares the texture depth, spacing and output error values ​​to detect the texture depth, spacing and orientation error. At the same time, combined with the collected hyperspectral imaging data (using a hyperspectral imager to analyze the soil composition (rammed earth area) of the brick and stone cultural relics, the composition of brick and stone materials, the degree of weathering and the distribution range of salt precipitation areas on the surface of brick and stone cultural relics, to determine the color gradient and weathering parameters of the texture replication), the color gradient of the texture is calibrated so that the color of the replicated texture is consistent with the weathering color of the original, providing an accurate texture template for subsequent laser aging.

[0312] This invention employs an integrated and collaborative replication technology that combines texture morphology and adhesive interface. Breaking away from the traditional approach of separating appearance replication from structural adaptation, it innovatively constructs a coupled replication system of surface texture and internal adhesive interface, achieving a two-way synergy between appearance fidelity and structural stability.

[0313] By collecting micro-roughness, interface concavity and convexity structure, and boundary interlocking contour parameters of the defect cross-section using 3D point cloud data, a spatial mapping relationship between surface texture features and bonding interface morphology features is established. Based on the XG Boost model, the stress concentration distribution pattern at the interface is predicted. Under the premise of not altering the appearance and texture of the cultural relic or damaging its historical features, the microstructure of the bonding interface is adaptively optimized. Through optimization of the micro-concavity and convexity interlocking structure and gradient transition interface, the effective bonding area between the restoration and the original material is increased. Simultaneously, the interface mechanical deformation characteristics are matched, weakening the deformation differences of the rammed earth-brick composite structure.

[0314] In a preferred embodiment of the present invention, the method for dividing the repaired area after texture replication into several independent standard components and numbering them, and for associating and storing the repair parameters of each component, is as follows:

[0315] After the texture is replicated, the entire brick and stone restoration area is divided into several independent standard brick and stone components according to the structural characteristics of the cultural relic and the ease of construction, and then numbered.

[0316] The segmentation boundary must avoid the core texture area of ​​the cultural relic and the anchor installation position. In the virtual model of the brick and stone cultural relic body of each brick and stone component, the anchor hole position (corresponding one-to-one with the anchor hole of the body), splicing baseline, and aging area (only the outer exposed surface) are marked. At the same time, the number, texture parameters, aging parameters, and size parameters are synchronously associated with the standardized fusion dataset to ensure the traceability of each brick and stone component.

[0317] In a preferred embodiment of the present invention, a 3D-printed recyclable double-layer cavity repair mold is used, with the original soil surrounding the site as the raw material, specifically:

[0318] The recyclable double-layer cavity repair mold is 3D printed. The material strength meets the stress requirements of manual ramming and has no adhesion to the original soil. There is no material residue after demolding, avoiding secondary pollution to the repaired parts and the original site.

[0319] Based on the outline of the repaired part and the reserved dimensions of the anchor bolt holes in the fused data, a double-layer cavity cooling mold model is designed. The inner forming cavity of the mold fits the outline of the repaired part, and holes matching the thickness of the anchor bolts are reserved at the corresponding anchor bolt positions. Preferably, two handrails are set on the outer membrane to facilitate later installation, and a support plate with a certain thickness is provided at the bottom to facilitate later maintenance. The mold is formed by 3D printing to ensure that the mold structure is stable and the hole positions are accurate.

[0320] A 1:1 restoration mold model is constructed using 3D modeling software. The inner cavity of the mold perfectly matches the outer dimensions of the restoration part, and it is pre-set with rammed layer grooves and surface texture replicas that are consistent with the original site. During manual ramming, the layering and texture features of the original site are directly restored. The outer cavity shape matches the inner mold, and the overall size is 10-15mm larger than the inner mold, with a thickness of 5-8mm. This provides structural rigidity to the mold and prevents deformation during ramming / cooling. The cavity gap is a closed gap between the inner and outer molds, used to fill refrigerant. It is 5-8mm wide and has one refrigerant filling port with a diameter of 8-10mm and one vent port with a diameter of 4-6mm (to prevent gas expansion during cooling from causing mold deformation). The filling port is fitted with a silicone low-temperature resistant sealing cap.

[0321] The mold is designed as a modular structure with pre-installed plastic elastic demolding clips, and the joints are sealed to prevent the rammed earth from leaking out and the refrigerant from leaking out.

[0322] The soil was prepared using native soil from the surrounding area of ​​the site: it was collected in layers (top, middle, and deep) to avoid compositional deviations caused by the introduction of foreign soil.

[0323] The collected soil is sieved and impurities are removed. Depending on the type of disease, trace amounts of functional additives can be added. For alkali-induced diseases, oxalic acid (desalination agent) and silane coupling agent (salt inhibitor) can be added. Deionized water is added to the prepared soil to adjust it to the optimal moisture content, achieving the best forming state for ramming. This ensures that the density of the repaired parts after ramming is consistent with the original site soil.

[0324] Ensure that the outer side of the rammed earth remains flat to provide a flat benchmark for subsequent brick and stone component splicing.

[0325] In a preferred embodiment of the present invention, layered compaction is performed, and the thickness of each layer is determined based on the texture depth of the inner cavity of the mold. Specifically:

[0326] The thickness of each layer is determined based on the depth of the texture in the inner cavity of the mold. The overall thickness is controlled at 2-3 cm per layer. In areas with a greater depth of texture, the layer thickness can be adjusted to 1-1.5 cm per layer to avoid unclear texture replication and internal voids caused by thick layer tamping.

[0327] S81, First layer filling: Fill the repair soil evenly to the thickness of the first layer, and use a fine brush to clean the repair soil in the gaps of the mold texture to ensure that the repair soil completely fills the texture grooves.

[0328] S82, Preliminary compaction: Gently tamp the surface of the repair material with a wooden tamping hammer, tamping evenly from the edge of the mold towards the center;

[0329] S83, Texture Replication: For areas with uneven surface texture, use a rubber pad to fit the texture of the inner wall of the mold, gently press the surface of the repair material to replicate the uneven details that match the texture of the mold, focus on pressing the grooves of the texture to ensure that the repair material and the mold texture are completely in contact.

[0330] S84, Subsequent Layering: Repeat steps S81-S83 above, filling, tamping, and replicating textures layer by layer until the top of the mold is filled. When tamping the last layer, calibrate the overall flatness and texture continuity of the repaired part surface to ensure consistency with the virtual model. At the same time, reserve the integrity of the anchor bolt holes to avoid clogging the holes during the tamping process.

[0331] In a preferred embodiment of the present invention, the steps for preparing the rammed earth repair component, combined with low-temperature curing treatment, are as follows:

[0332] After the entire repair part is compacted, the silicone sealing end cap is fastened and fixed with positioning buckles to ensure the mold is sealed. Granular dry ice is then filled through the mold filling port, with the filling amount being 80% of the cavity gap layer volume (to reserve expansion space). After filling, the silicone low-temperature resistant sealing cap is fastened to prevent the dry ice from sublimating rapidly.

[0333] After standing for 2-4 hours at room temperature (15~25℃), the free water in the soil of the repaired part is frozen by the low temperature of dry ice. The micro-freezing heave effect of the ice fills the gaps between soil particles, making the particles more tightly interlocked, thereby improving the soil density and strength.

[0334] Throughout the entire process, the surface temperature of the repaired parts is controlled to be no lower than -10℃ to avoid excessive low temperature causing soil cracking.

[0335] After low-temperature curing, open the sealing cap of the mold filling port to allow the CO2 gas in the cavity gap layer to be discharged naturally. The mold and the repaired part are allowed to warm up naturally at room temperature (15~25℃) to avoid rapid warming up which would increase the porosity of the soil. The cured soil particles remain tightly interlocked.

[0336] After the temperature recovery is complete, remove the sealing end cap and connect the repair part to the main body according to the pre-set anchor bolt hole alignment point;

[0337] The anchor bolts are made of lightweight, high-strength materials (such as carbon fiber), and their surfaces are micro-abraded to enhance adhesion to the restoration soil and the original soil. An environmentally friendly anti-corrosion coating is applied to the anchor bolt surface to prevent rusting from prolonged exposure to a humid environment, which could affect connection stability. Inorganic materials are used for micro-grouting at the joints to further strengthen the bond between the restoration and the original soil.

[0338] In a preferred embodiment of the present invention, brick and stone components are prepared based on parameters output from a virtual model of the brick and stone artifact, and CNC laser engraving technology is used to perform texture replication and gradient aging treatment on the exposed outer surface of the components. The specific steps are as follows:

[0339] Antique brick and stone materials consistent with the material of the cultural relic are selected. Based on the size parameters of the virtual model of the brick and stone cultural relic, individual brick and stone components are cut and prepared with the cutting accuracy controlled within ±0.1mm.

[0340] After cutting, the splicing surfaces of the brick and stone components are polished to ensure that the interlocking structure of the splicing surfaces is consistent with the virtual model, and at the same time, surface dust and burrs are cleaned.

[0341] The virtual model and texture parameters of each brick and stone artifact are imported into a CNC laser engraving device, and texture engraving is performed on the exposed outer surface of the brick and stone artifact.

[0342] Based on the hardness of the brick and stone material (e.g., hardness of blue brick is 5~7H), adjust the laser engraving parameters: power 50~80W (blue brick), engraving speed 100~200mm / min, focal length 20~30mm. Adjust the laser engraving parameters so that the engraving depth is consistent with the concave and convex depth of the texture in the virtual model of the brick and stone cultural relic, and ensure that the details of the engraved texture are completely matched with the replicated texture.

[0343] The brick and stone components are fixed on the carving workbench. The CCD vision positioning system is used to precisely align the brick and stone components with the virtual model to ensure that the position of the carved texture is consistent with the virtual model and to avoid texture misalignment.

[0344] A layered carving method is adopted, first carving the outline and main structure of the texture, and then carving the details of the texture. The carving depth of each layer is controlled at 0.05~0.1mm. The carving effect is monitored in real time during the carving process. If texture deviation occurs, the carving parameters are adjusted in time (the texture carving outline and depth data are collected in real time through the CCD vision positioning system, and the real-time data is compared with the texture parameters of the virtual model).

[0345] For areas with a texture depth of 0.5mm ≤ 1.5mm, they are considered to have a large texture depth, and the number of engraving layers is increased to ensure that the three-dimensionality of the texture is consistent with the body.

[0346] After the carving is completed, use a high-pressure blower to blow away the carving dust on the surface of the brick and stone components, and then use a fine brush to clean the residual dust in the texture gaps to ensure that the texture is clear and free of dust residue. At the same time, check the texture replication effect and compare it with the virtual model (using CloudCompare software, after spatial coordinate registration, the software automatically compares the deviation values ​​of texture concavity and convexity depth, spacing and contour).

[0347] Based on the aging gradient parameters output by a deep learning model, combined with the weathering degree of the artifact itself, CNC laser engraving equipment is used to perform targeted aging treatment on the exposed outer surfaces of the brick and stone components. This simulates the effects of natural weathering, wear, and erosion without altering the internal structure and splicing precision of the brick and stone components. Specifically:

[0348] Based on the weathering level of the artifact, different laser aging parameters are set: For lightly weathered areas: power 30~50W, engraving speed 200~300mm / min, using point carving to simulate slight wear marks, point carving density 5~10 dots / cm², dot diameter 0.1~0.2mm; For moderately weathered areas: a combination of line carving and point carving is used to simulate erosion grooves and wear marks, groove width 0.2~0.5mm, depth 0.1~0.3mm; For heavily weathered areas: a combination of surface carving, line carving, and point carving is used to simulate severe wear and peeling marks, surface carving area roughness Ra 1.6~3.2μm, ensuring the aging effect is consistent with the weathering level of the artifact.

[0349] Based on the weathering gradient of the exposed outer surface of the brick and stone components, laser aging is performed in sections, with a natural transition from lightly weathered to heavily weathered areas to avoid abrupt changes in the aging effect. The focus is on simulating weathering characteristics under natural conditions, such as edge wear, surface erosion spots, and mineral precipitation marks, ensuring that the aged texture blends naturally with the original texture, making it difficult to distinguish the restored area from the original area with the naked eye.

[0350] Antique brick and stone materials consistent with the material of the cultural relic are selected. Based on the size parameters of the virtual model, individual brick and stone components are cut and prepared with a cutting accuracy controlled within ±0.1mm. After cutting, the splicing surfaces of the brick and stone components are polished to ensure that the interlocking structure of the splicing surfaces is consistent with the virtual model. At the same time, surface dust and burrs are cleaned to prepare for subsequent laser engraving and aging.

[0351] In a preferred embodiment of the present invention, all brick and stone components are pre-assembled to form a complete brick and stone repair body. The specific method is as follows:

[0352] Arrange the aged brick and stone components in numerical order, and check the splicing surfaces and anchor bolt holes of each component to ensure consistency with the virtual model;

[0353] Prepare the splicing materials. Select the original soil that is consistent with the rammed earth repair to ensure that the splicing materials are compatible with the brick and stone materials and the rammed earth materials, and that the bonding strength meets the standards.

[0354] Following the assembly sequence of the virtual model of the brick and stone cultural relic, the brick and stone components are assembled layer by layer from bottom to top. The splicing material is evenly applied to the splicing surfaces of the brick and stone components to ensure that the splicing surfaces are completely covered without any omissions.

[0355] Align adjacent brick and stone components according to the splicing baseline, so that the splicing surfaces fit together and the anchor bolt holes are aligned.

[0356] After each layer is assembled, it is fixed with clamps and left to stand for 10 to 15 minutes to allow the assembling material to initially solidify before assembling the next layer to avoid misalignment. During the assembly process, any excess assembling material that overflows from the joints should be cleaned up promptly to ensure that the aging effect on the exposed outer surface of the brick and stone components is not contaminated.

[0357] After all the brick and stone components are assembled and the assembly materials have completely solidified, anchor rods are driven into the preset anchor rod holes. The anchor rods are made of lightweight and high-strength carbon fiber material, and their diameter matches the size of the holes.

[0358] Before the anchor bolts are installed, the surface is micro-sanded to enhance the adhesion with the splicing materials and rammed earth, and an environmentally friendly anti-corrosion coating is applied to prevent rusting and failure.

[0359] During implantation, the top of the anchor rod is tapped with a rubber mallet to ensure that the implantation depth of the anchor rod is consistent with the virtual model of the brick and stone cultural relic, and the length of the anchor rod extending out of the brick and stone component matches the depth of the anchor hole in the rammed earth restoration area. After implantation, inorganic materials are used to micro-grout the gap between the anchor rod and the hole to compact and fix it to prevent loosening.

[0360] Based on the modular segmentation numbering and splicing parameters of virtual repair, all masonry components are pre-assembled in the factory to form a complete masonry repair whole, while anchor bolts are fixed to facilitate on-site assembly.

[0361] In a preferred embodiment of the present invention, the rammed earth restoration component and the brick and stone restoration body are assembled on-site at the archaeological site to form a stable rammed earth-brick and stone composite structure. The specific method is as follows:

[0362] Clean the surface dust of the rammed earth repair area, check the flatness of the rammed earth surface and the anchor bolt hole positions to ensure that it matches the anchor bolts and splicing surfaces of the brick and stone repair body; apply a thin, even layer of inorganic mortar material to the splicing area of ​​the rammed earth surface to enhance the adhesion.

[0363] Using hoisting equipment, the assembled brick and stone restoration was lifted to the preset position and aligned with the rammed earth restoration area according to the splicing baseline. This ensured that the anchor rods of the brick and stone restoration were inserted into the anchor holes in the rammed earth area, and that the spliced ​​surfaces fit together seamlessly. During assembly, a total station was used to monitor the assembly accuracy in real time, adjusting the position of the brick and stone restoration to ensure overall flatness consistent with the original artifact. The total station collected real-time three-dimensional coordinate data of the brick and stone restoration, the rammed earth restoration area, and the original artifact. The measured data was compared with the preset baseline coordinates of the virtual model, automatically calculating the plane, elevation, and angular deviations. Operators manually controlled the hoisting equipment for fine-tuning based on the deviation values ​​fed back by the total station in real time.

[0364] After assembly, inorganic materials are used for micro-grouting at the joints between the brick and stone restoration and the rammed earth, so that the joints can be naturally integrated with the surrounding structure.

[0365] Non-woven fabric bandages are used to flexibly fix the connection points, and the entire structure is kept at room temperature for 7 to 14 days. After the curing period, the flexible fixing device is removed, and the exposed outer surface of the brick and stone restoration is cleaned.

[0366] The present invention also provides a 3D-printed recyclable double-layer cavity repair mold for use in the method described herein, comprising:

[0367] The inner mold has a forming cavity that fits the outline of the rammed earth repair component. The inner mold has pre-reserved anchor holes at the corresponding anchor installation positions, and the inner wall is provided with grooves for replicating the rammed earth layering and raised textures for replicating the surface texture.

[0368] The outer mold is fitted over the inner mold to provide structural rigidity for the mold.

[0369] A cavity gap layer, formed in the closed gap between the inner mold and the outer mold, is used to fill the refrigerant to achieve low-temperature curing.

[0370] The present invention also provides a repair structure for use in the method described herein, comprising an intermediate rammed earth layer and an outer brick and stone layer;

[0371] The middle rammed earth layer is formed by layering and compacting the original soil around the site and then treating it with low temperature. Its outer surface is kept flat, while its inner surface is connected to the site itself. Anchor rods are pre-embedded in the middle rammed earth layer, with one end of the anchor rods anchored inside the site itself.

[0372] The outer brick and stone layer covers the outer surface of the middle rammed earth layer. The outer brick and stone layer is composed of several independent brick and stone components. The exposed outer surface of each brick and stone component is decorated with an antique texture formed by CNC laser engraving, which is integrated with the surface texture and weathering degree of the site.

[0373] The intermediate rammed earth layer and the outer brick and stone layer are fixedly connected by anchor bolts and / or bonding materials to form a stable rammed earth-brick and stone composite structure.

[0374] The specific embodiments described herein are merely illustrative examples of the present invention. Those skilled in the art can make various modifications or additions to the described embodiments or use similar methods to substitute them, without departing from the technology of the present invention or exceeding the scope defined by the appended claims.

[0375] In the embodiments of this application, terms such as "fixed," "fixed connection," and "fixed connection" refer to common fixing methods in the prior art, such as welding, riveting, and screws. "Rotary connection" refers to common rotary connection methods in the prior art, such as hinges and bearing rotation. If electrical components are provided, the functions, control, and power supply methods of all electrical components are common technical means in the prior art. This application has not improved them and they are not within the protection scope of this application. Therefore, this application will not elaborate on them.

[0376] Furthermore, the selection of materials and strength limitations for all components in this application can be made and arranged by those skilled in the art based on the site environment and the requirements of relevant national or industry standards, and are not within the scope of protection of this application. Therefore, this application will not elaborate on these points.

Claims

1. A method for authentic restoration of earthen archaeological sites using 3D printing, characterized in that: Includes the following steps: Multi-source information was collected on the site itself, and the collected heterogeneous data from multiple sources were fused and processed to output a standardized fused dataset. The standardized fusion dataset is input into the deep learning disease identification model for disease identification and parameter quantification. Based on the standardized fusion dataset and disease identification results, a virtual model of the brick and stone cultural relic is constructed to determine the thickness of the rammed earth restoration, as well as the boundaries of the rammed earth restoration area and the brick and stone restoration areas on both sides, and to clarify the texture replication range and aging requirements of the brick and stone restoration area. Deep learning is used to replicate the texture of the defective area. The texture-replicated brick and stone repair area is divided into several independent standard component models and numbered. The texture parameters and repair parameters of each component model are associated and saved. Based on the rammed earth restoration area, a recyclable double-layer cavity restoration mold was obtained by 3D printing. Using the original soil around the site as raw material, the rammed earth restoration parts were prepared by layered ramming and combined with low temperature curing treatment. Based on the component model, texture parameters, and repair parameters of the brick and stone repair area segmentation, brick and stone repair parts are prepared, and CNC laser engraving technology is used to perform texture replication and gradient aging treatment on the exposed surface of the component. All brick and stone components are pre-assembled to form a complete brick and stone restoration body; The rammed earth restoration components and brick and stone restoration components were assembled on-site to form a stable rammed earth-brick and stone composite structure.

2. The 3D printing-based method for authentic restoration of earthen archaeological sites by identifying and integrating aging techniques, as described in claim 1, is characterized in that... Multi-source information was collected from the site itself, and the collected heterogeneous data from multiple sources was fused to output a standardized fused dataset, specifically: Anchor holes were pre-drilled in the main body of the site; Using 3D laser scanners or close-range photogrammetry, 3D morphological data of the damaged parts of the site and the surrounding undamaged areas are collected to identify the location, area, size of surface peeling and defects, as well as the surface texture of the intact areas. Based on the three-dimensional morphological data of the site, and combined with the stress characteristics of the rammed earth-brick composite structure and the anchoring requirements of the restoration, the anchor installation parameters, including the anchor installation position, thickness, and insertion depth, are determined by structural stress simulation and spatial point adaptation to ensure that the anchor matches the structure. Based on the thickness of the rammed earth repair and the depth of the body stabilization layer, the anchor bolt implantation depth is set by three-dimensional spatial coordinate calibration to ensure that the implanted end penetrates into the body stabilization layer and the exposed end is compatible with the anchor hole of the brick and stone repair body. All anchor bolt parameters were spatially registered and verified with the three-dimensional morphological data of the body to complete the parameter setting. Infrared thermal imagers were used to detect the location and depth of hollow areas, hidden cracks, and differences in moisture content inside the artifacts at the site, and to identify loose and eroded areas, providing a basis for controlling the thickness and density of rammed earth restoration. Hyperspectral imaging was used to analyze the soil composition, brick and stone material composition, weathering degree, and salt precipitation area distribution of the surface of brick and stone cultural relics, and to determine the color gradient and weathering degree parameters of the texture reproduction. Ultrasonic density testing was used to determine the density of the brick and stone body and the rammed earth area, and to identify the loose areas inside. Collect dimensional tolerances, splicing gaps, and bonding parameters of the subgrade components, and combine them with anchor bolt installation requirements; The collected multi-source heterogeneous data were normalized, noise removed, and missing values ​​filled in. Core correlation features for the restoration of brick and stone cultural relics were extracted from the preprocessed multi-source data, including geometric features: surface texture, defect outline, splicing tolerance, and anchoring positioning coordinates; material physicochemical features: brick and stone matrix composition, weathering grade, moisture content, porosity, and density; structural mechanical features: location of internal defects, hollow area, crack depth, and anchoring stress threshold; process adaptation features: restoration material matching parameters, anchoring depth, and splicing gap; and appearance fidelity features: brick and stone surface color, weathering texture gradient, and aging matching parameters. Using the spatial coordinates of the three-dimensional morphological data of cultural relics as a benchmark, the remaining data are spatially / dimensionally registered with the morphological features to ensure one-to-one correspondence between the data and output a standardized fusion dataset.

3. The 3D printing-based method for authentic restoration of earthen archaeological sites by identifying and integrating aging techniques, as described in claim 1, is characterized in that... The deep learning disease identification model adopts the XGBoost model and combines SHAP interpretability analysis. The deep learning disease identification model includes an input layer, a feature preprocessing layer, an XGBoost classification and regression layer, a SHAP feature contribution analysis layer, and an output layer connected in sequence. The standardized fusion dataset is input into the deep learning disease identification model for disease identification and parameter quantification, specifically as follows: The features of 3D topography, infrared imaging, spectral detection, and substrate material detection are normalized and then input into the XGBoost model. The XGBoost model automatically classifies disease types, including efflorescence, spalling, cracking, and erosion, and outputs quantitative parameters, including disease location, area, and depth. The contribution of each disease type and quantitative parameter was calculated using the SHAP analysis method, redundant features were eliminated, and the identification accuracy was optimized. Output disease types and quantified parameters in order of contribution.

4. The 3D printing-based method for authentic restoration of earthen archaeological sites by identifying and integrating aging techniques, as described in claim 3, is characterized in that... A virtual model of the brick and stone cultural relic was constructed based on the standardized fusion dataset and the results of the defect identification. The thickness of the rammed earth restoration, as well as the boundaries of the rammed earth restoration area and the brick and stone restoration areas on both sides, were determined. The method for clarifying the texture replication range and aging requirements of the brick and stone restoration area is as follows: Based on the maximum depth of the disease predicted by the XG Boost model, combined with the SHAP feature contribution ranking, and superimposed with the anchoring depth and safety margin, the thickness of the rammed earth repair is determined to ensure the strength and stability of the repair structure. ， Among them, SHAP i : The SHAP value of the i-th detection feature, which is the quantified value of the feature's contribution to the disease identification result. A positive SHAP value indicates that the feature promotes the identification of the corresponding disease, while a negative value indicates that it inhibits the identification of the corresponding disease. The larger the absolute value, the higher the feature contribution. F represents the set of multi-source detection features, corresponding to the multi-source detection data input to the input layer. i: The index of a single detection feature, corresponding to a specific detection index. S: Any subset of the feature set F that does not contain the i-th feature, used to simulate the model's identification result after removing the feature. M: The total number of features in the feature set F, i.e., the total dimension of the multi-source detection features, which is consistent with the feature dimensions of the input layer and the feature preprocessing layer. The disease identification prediction value output by the XGBoost model after adding the i-th feature to the feature subset S; The disease identification prediction value output by the feature subset S after it is input into the XGBoost model; ， Where H represents the final thickness of the rammed earth repair; D dis The maximum depth of the disease after SHAP correction; D a The minimum depth for anchor bolts to be implanted into the rammed earth stabilized base is set by the anchor bolt installation parameters; ΔH safe For safety margin; Based on the disease outline output by the XG Boost model, and combined with the SHAP contribution characteristics, the boundary between the rammed earth repair area and the brick and stone repair area is determined. The boundary extends 5-10mm outward to the complete substrate to ensure the stability of the repair overlap. Based on the spatial coordinates of the cultural relic itself, spatial registration and point cloud fusion are performed on the standardized fusion data to construct a geometric grid model of the cultural relic itself. The actual texture, color, and weathering characteristics of the cultural relic's surface are mapped onto a geometric mesh model to complete the construction of the virtual model of the artifact. Specifically: Extract the edge features of the real texture on the surface of the cultural relic, including the three-dimensional coordinates, direction angle, and transition gradient of the texture edge; Using the spatial coordinates of the artifact itself as the sole reference, the texture edge feature dataset is registered with the edge coordinates of the geometric mesh model, and the least squares method is used to calculate the coordinate deviation between the texture edge and the mesh edge. The transition area of ​​the texture edge is smoothed, and the interpolation algorithm is used to fill the tiny gap between the texture edge and the mesh edge, so that the texture edge and the edge of the geometric mesh model are seamlessly connected. By combining the complete regional texture patterns in the multi-source fusion data, the consistency of the aligned texture edges is checked to ensure that the direction and spacing of the texture edges are consistent with the real texture edges of the cultural relic. Based on standardized fusion data, the defects, cracks and weathering areas of cultural relics are identified and parameterized, and a virtual model of the brick and stone cultural relics is constructed to comprehensively analyze the type, distribution, size and severity of defects in cultural relics. Adjustments were made based on the internal hollowness and looseness, clarifying the scope of texture replication and aging requirements for the brick and stone repair area, ensuring a stable rammed earth-brick and stone composite structure is formed after repair. Specifically: Based on the original rammed earth and brick-stone interface of the cultural relic, and combined with the disease outline and structurally stable area, a digital closed boundary is delineated in the virtual model of the brick-stone cultural relic. The maximum depth of the defect and the anchorage thickness of the anchor bolt are taken, and the minimum safe thickness to meet the structural stability is determined by dynamically adjusting the hollow / looseness based on infrared and ultrasonic detection. Using the three-dimensional outline of the defective brick and stone area as the boundary, extract the texture features of the surrounding complete brick and stone, and lock the texture replication area in the virtual model of the brick and stone cultural relic. Based on the weathering and color parameters obtained from hyperspectral imaging, the color, roughness, and natural aging characteristics of the original bricks and stones are matched to determine the gradient aging parameters.

5. The 3D printing-based method for authentic restoration of earthen archaeological sites by identifying and integrating aging techniques according to claim 4, characterized in that... Texture replication of defective regions is performed using deep learning, specifically as follows: Texture samples of complete areas of the cultural relics were extracted from the standardized fusion dataset, classified according to brick and stone material and weathering degree, and detailed parameters were labeled, including: texture depth, spacing, direction, and weathering traces. These samples were then input into a deep learning disease identification model for training and optimization to ensure that the model can learn texture features of different materials and weathering degrees. ， Among them, h i,j Z represents the bump depth of the texture at pixel (i,j); i,j Z0 represents the 3D elevation value of a pixel; Z0 represents the elevation value of the texture reference plane. ， Where, d k The distance between the k-th texture unit and the (k+1)-th texture unit; (x k ,y k ), (x k+1 ,y k+1 These are the planar coordinates of the centers of the two texture units, respectively. ， Where W is the weathering coefficient, the smaller the value, the more severe the weathering; I w The grayscale value of the weathered area texture; I i The grayscale values ​​represent the texture of the complete, unweathered area. Based on a trained deep learning-based defect identification model, and combining the boundary contours and size parameters of the defective region with the texture patterns of the surrounding intact region, the texture of the defective region is replicated: ， Among them, L min T is the loss value used for model training. pred,n T represents the texture parameter values ​​predicted by the model. true,n Here are the actual texture parameter values, λ is the regularization coefficient, and Ω(f) is the value of the texture parameter in the actual annotation. t ) is the complexity penalty term for the t-th decision tree; through iterative training, the model can accurately learn the texture distribution patterns and feature mapping relationships under different brick and stone materials and different degrees of weathering; For areas with regular defects, texture features of the surrounding complete areas are extracted and copied to the areas with regular defects to ensure that the texture direction, spacing, and bump depth are consistent with the original texture. ， Among them, T defect (x,y) represents the replicated texture parameters at the defect region (x,y), T intact (x+a,y+b) represents the texture parameters of the template corresponding to the surrounding complete area, where a and b are translation offsets to ensure the continuity and consistency between the replicated texture and the original texture. For irregularly shaped defect areas, interpolation and texture transfer techniques are used to transfer the texture features of the complete body area to the defect area, restoring the texture details of the defect area, while simulating natural weathering gradients, specifically: The basic texture mesh of the defect area is restored using bilinear interpolation. : ， Where u=x−x0 and v=y−y0 are interpolation weights, and (x0,y0), (x1,y0), (x0,y1), and (x1,y1) are the coordinates of the four surrounding complete texture pixels; Texture transfer technology is used to transfer texture features and weathering gradients from the complete area of ​​the body to the defect area: ， Among them, T source (x,y) represents the complete region texture source features, T target (x,y) represents the initial texture of the defect region, and W(x,y) represents the migration weight; the weathering gradient is simulated using the following formula: ， Among them, W center is the weathering coefficient at the center of the defect area, k is the weathering gradient coefficient, and D(x,y) is the distance from the pixel to the defect boundary, realizing a natural weathering transition from the defect center to the edge and restoring the weathering details of real bricks and stones. This represents a simulated natural weathering gradient; Weathering time-series features were extracted from hyperspectral imaging data: color difference gradation, surface hardness decay, and salt precipitation distribution of the brick and stone body. The influence weights of various environmental and material parameters on the weathering gradient were quantified using the SHAP feature contribution formula. The core feature variables that dominate the weathering evolution of brick and stone were selected, and a texture time-series weathering evolution model was constructed. ， Where T(W) represents the replicated texture parameter under the corresponding weathering coefficient, and T0 represents the original complete texture parameter of the brick / stone. η represents the texture parameters of heavily weathered material, W is the measured weathering coefficient of the body, and η is the weathering gradation adjustment coefficient. The replicated texture is compared with the texture of the complete area of ​​the original. The texture's bump depth, spacing, and direction errors are detected by the comparison software CloudCompare. At the same time, the color gradient of the texture is calibrated by combining the collected hyperspectral imaging data to make the color of the replicated texture consistent with the weathering color of the original.

6. The 3D printing-based method for authentic restoration of earthen archaeological sites by identifying and integrating aging techniques according to claim 1, characterized in that... The method for dividing the repaired area after texture replication into several independent standard components and numbering them, and then saving the repair parameters of each component in association, is as follows: After the texture is replicated, the entire brick and stone restoration area is divided into several independent standard brick and stone components according to the structural characteristics of the cultural relic and the ease of construction, and then numbered. The segmentation boundary must avoid the core texture area of ​​the cultural relic and the anchor installation position. In the virtual model of the brick and stone cultural relic body of each brick and stone component, the anchor hole position, splicing baseline and aging area are marked. At the same time, the number, texture parameters, aging parameters and size parameters are synchronously linked to the standardized fusion dataset to ensure the traceability of each brick and stone component.

7. The 3D printing-based method for authentic restoration of earthen archaeological sites by identifying and integrating aging techniques according to claim 1, characterized in that... The 3D-printed recyclable double-layer cavity restoration mold uses native soil from the surrounding area of ​​the site as raw material, specifically: Based on the outline of the repaired part and the reserved dimensions of the anchor bolt holes in the fused data, a double-layer cavity cooling mold model is designed. The inner forming cavity of the mold fits the outline of the repaired part, and holes matching the thickness of the anchor bolts are reserved at the corresponding anchor bolt positions. A 1:1 restoration mold model was constructed using 3D modeling software. The inner cavity of the mold perfectly matched the outer dimensions of the restoration part. It also had pre-set rammed layer grooves and surface texture replicas that were consistent with the original site. During manual ramming, the stratification and texture features of the original site were directly restored. The mold is designed as a modular structure with pre-installed plastic elastic demolding clips, and the joints are sealed. The soil was prepared using native soil from the surrounding area of ​​the site: the top, middle and deep layers were collected in layers to avoid compositional deviations caused by the introduction of foreign soil. The collected soil is sieved and impurities are removed. Depending on the type of disease, trace amounts of functional additives can be added. For alkali-induced diseases, oxalic acid and silane coupling agents can be added. Deionized water is added to the prepared soil to adjust it to the optimal moisture content, achieving the best forming state for ramming and ensuring that the outer side of the rammed soil remains flat.

8. The 3D printing-based method for authentic restoration of earthen archaeological sites by identifying and integrating aging techniques, as described in claim 7, is characterized in that... Layered compaction is performed, and the thickness of each layer is determined based on the texture and depth of the inner forming cavity of the mold. Specifically: S81, fill the repair material evenly to the thickness of the first layer, and use a fine brush to clean the repair material in the gaps of the mold texture to ensure that the repair material completely fills the texture grooves. S82, gently tap the surface of the repair material with a wooden tamping hammer, tapping evenly from the edge of the mold towards the center; S83, for areas with uneven surface texture, use a rubber pad to fit the texture of the inner wall of the mold, gently press the surface of the repair material to replicate the uneven details consistent with the mold texture, focus on pressing the grooves of the texture to ensure that the repair material and the mold texture are completely in contact. S84. Repeat steps S81-S83 above, filling, tamping, and replicating the texture layer by layer until the top of the mold is filled. When tamping the last layer, calibrate the overall flatness and texture continuity of the repaired part surface to ensure consistency with the virtual model. At the same time, reserve the integrity of the anchor hole positions to avoid clogging the holes during the tamping process.

9. The 3D printing-based method for authentic restoration of earthen archaeological sites by identifying and integrating aging techniques, as described in claim 8, is characterized in that... The steps for preparing rammed earth repair components, combined with low-temperature curing treatment, are as follows: After the entire repair part is compacted, the silicone sealing end cap is fastened and fixed with positioning buckles to ensure the mold is sealed. Granular dry ice is then filled through the mold filling port, with the filling amount being 80% of the cavity gap layer volume. After filling, the silicone low-temperature resistant sealing cap is fastened to prevent the dry ice from sublimating rapidly. After standing at room temperature for 2-4 hours, the free water in the soil of the repaired part is frozen by the low temperature of dry ice, and the micro-freezing heave effect of the ice fills the gaps between soil particles. Throughout the entire process, the surface temperature of the repaired parts is controlled to be no lower than -10℃ to avoid excessive low temperature causing soil cracking. After low-temperature curing is completed, open the sealing cap of the mold filling port to allow the CO2 gas in the cavity gap layer to be discharged naturally, and allow the mold and the repaired part to warm up naturally at room temperature. After the temperature recovery is complete, remove the sealing end cap and connect the repair part to the main body according to the pre-set anchor hole alignment point; The anchor rods are made of lightweight and high-strength materials, and the surface of the anchor rods is micro-abraded to enhance the adhesion between the anchor rods and the original soil of the repair soil and the original soil of the body. An environmentally friendly anti-corrosion coating is applied to the surface of the anchor rods, and inorganic materials are used for micro-grouting at the joints to enhance the adhesion between the repair body and the original body.

10. The 3D printing-based method for authentic restoration of earthen archaeological sites by identifying and integrating aging techniques according to claim 1, characterized in that... Based on the parameters output from the virtual model of the brick and stone cultural relic, brick and stone components were prepared, and CNC laser engraving technology was used to replicate the texture and perform gradient aging on the exposed outer surface of the components. The specific steps are as follows: Antique brick and stone materials that are consistent with the material of the cultural relic are selected, and individual brick and stone components are cut and prepared according to the size parameters of the virtual model of the brick and stone cultural relic. After cutting, the splicing surfaces of the brick and stone components are polished to ensure that the interlocking structure of the splicing surfaces is consistent with the virtual model, and at the same time, surface dust and burrs are cleaned. The virtual model and texture parameters of each brick and stone artifact are imported into a CNC laser engraving device, and texture engraving is performed on the exposed outer surface of the brick and stone artifact. Based on the hardness of the brick and stone material, the laser engraving parameters are adjusted to ensure that the engraving depth matches the texture depth in the virtual model of the brick and stone artifact. The brick and stone components are fixed on the carving workbench, and the brick and stone components are precisely aligned with the virtual model through the CCD vision positioning system. The process employs a layered carving method, first carving the outline and main structure of the texture, and then carving the details of the texture. The carving effect is monitored in real time during the carving process, and the carving parameters are adjusted in a timely manner if texture deviation occurs. For areas with a texture depth of 0.5mm ≤ 1.5mm, they are considered to have a large texture depth, and the number of engraving layers is increased to ensure that the three-dimensionality of the texture is consistent with the body. After the carving is completed, use a high-pressure blower to blow away the carving dust on the surface of the brick and stone components, and then use a fine brush to clean the residual dust in the texture gaps to ensure that the texture is clear and free of dust residue. At the same time, check the texture replication effect and compare it with the virtual model. Based on the aging gradient parameters output by a deep learning model, combined with the weathering degree of the artifact itself, CNC laser engraving equipment is used to perform targeted aging treatment on the exposed outer surfaces of the brick and stone components. This simulates the effects of natural weathering, wear, and erosion without altering the internal structure and splicing precision of the brick and stone components. Specifically: Different laser aging parameters are set according to the weathering level of the cultural relic; Based on the weathering gradient of the exposed outer surface of the brick and stone components, laser aging is carried out in different areas to create a natural transition from lightly weathered areas to heavily weathered areas.

11. The 3D printing-based method for authentic restoration of earthen archaeological sites by identifying and integrating aging techniques according to claim 1, characterized in that... All brick and stone components are pre-assembled to form a complete brick and stone restoration body. The specific method is as follows: Arrange the aged brick and stone components in numerical order and check the splicing surfaces and anchor bolt holes of each component; Prepare the materials for assembly; the original soil used should be the same as that used in the rammed earth repair. Following the assembly sequence of the virtual model of the brick and stone cultural relic, the brick and stone components are assembled layer by layer from bottom to top. The splicing material is evenly applied to the splicing surfaces of the brick and stone components to ensure that the splicing surfaces are completely covered without any omissions. Align adjacent brick and stone components according to the splicing baseline, so that the splicing surfaces fit together and the anchor bolt holes are aligned. Each layer is fixed with clamps, and the next layer is assembled only after the assembled material has initially solidified. After all the brick and stone components are assembled and the assembly materials have completely solidified, anchor rods are driven into the preset anchor rod holes. The anchor rods are made of lightweight and high-strength carbon fiber material, and their diameter matches the size of the holes. Before the anchor bolts are installed, the surface is micro-abraded to enhance the adhesion with the splicing materials and rammed earth, and an environmentally friendly anti-corrosion coating is applied at the same time. During implantation, the top of the anchor rod is tapped with a rubber mallet to ensure that the implantation depth of the anchor rod is consistent with the virtual model of the brick and stone cultural relic, and the length of the anchor rod extending out of the brick and stone component matches the depth of the anchor hole in the rammed earth restoration area. After implantation, inorganic materials are used to micro-grout the gap between the anchor rod and the hole to compact and fix it.

12. The 3D printing-based method for authentic restoration of earthen archaeological sites by identifying and integrating aging techniques according to claim 1, characterized in that... The rammed earth restoration components and the brick and stone restoration body were assembled on-site at the site to form a stable rammed earth-brick and stone composite structure. The specific method is as follows: Clean the surface dust of the rammed earth repair area, check the flatness of the rammed earth surface and the anchor bolt hole positions to ensure that it matches the anchor bolts and splicing surfaces of the brick and stone repair body; apply a thin, even layer of inorganic mortar material to the splicing area of ​​the rammed earth surface to enhance the adhesion. Using hoisting equipment, the assembled brick and stone restoration body is hoisted to the preset position and aligned with the rammed earth restoration area according to the splicing baseline. This ensures that the anchor rods of the brick and stone restoration body are inserted into the anchor holes in the rammed earth area, and that the splicing surfaces fit together without gaps. During the assembly process, a total station is used to monitor the assembly accuracy in real time and adjust the position of the brick and stone restoration body to ensure that the overall flatness is consistent with the original cultural relic. After assembly, inorganic materials are used for micro-grouting at the joints between the brick and stone restoration and the rammed earth, so that the joints can be naturally integrated with the surrounding structure. Non-woven fabric bandages are used to flexibly fix the connection points, and the entire structure is kept at room temperature for 7 to 14 days. After the curing period, the flexible fixing device is removed, and the exposed outer surface of the brick and stone restoration is cleaned.

13. A 3D-printed recyclable double-layer cavity repair mold for use in the method of any one of claims 1-12, characterized in that, include: The inner mold has a forming cavity that fits the outline of the rammed earth repair component. The inner mold has pre-reserved anchor holes at the corresponding anchor installation positions, and the inner wall is provided with grooves for replicating the rammed earth layering and raised textures for replicating the surface texture. The outer mold is fitted over the inner mold to provide structural rigidity for the mold. A cavity gap layer, formed in the closed gap between the inner mold and the outer mold, is used to fill the refrigerant to achieve low-temperature curing.

14. A repair structure for use in the method of any one of claims 1-12, characterized in that, include: The intermediate rammed earth layer is formed by layering and ramming the original soil around the site and then curing it at low temperature. Its outer surface is kept flat, and its inner surface is connected to the site body. Anchor rods are pre-embedded in the intermediate rammed earth layer, and one end of the anchor rods is anchored inside the site body. The outer brick and stone layer covers the outer surface of the middle rammed earth layer. The outer brick and stone layer is composed of several independent brick and stone components. The exposed outer surface of each brick and stone component is decorated with an antique texture formed by CNC laser engraving, which is integrated with the surface texture and weathering degree of the site. The intermediate rammed earth layer and the outer brick and stone layer are fixedly connected by anchor bolts and / or bonding materials to form a stable rammed earth-brick and stone composite structure.

Images