A modeling method and system of a landslide disaster chain physical model of a small watershed in a granite weathering zone

Abstract

A modeling method and system for a physical model of landslide disaster chains in small watersheds of granite weathering areas. The method includes: S1 In-situ parameter identification and inversion, acquiring UAV LiDAR, high-density electrical resistivity tomography, micro-penetration, and microseismic data, and using Bayesian-Kriging coupling inversion to obtain the three-dimensional parameter fields of each layer of the weathering crust; S2 Digital modeling and printing mix generation, constructing a similar material knowledge graph and actively learning to recommend printing mix schemes; S3 Construction based on additive-subtractive composite technology, layer by layer stacking while using wedge cutters to cut and form residual joint surfaces between layers; S4 Model integration and monitoring system deployment, connecting multiple models in series and setting up overflow weirs to form simulated channels in small watersheds. This invention achieves high-fidelity reproduction of the multi-layer gradient structure and interlayer joints of the granite weathering crust, establishes a closed-loop mapping from in-situ survey to model construction, and can simulate the entire process of a cluster landslide-ditch blocking-breakage disaster chain with high fidelity and good repeatability.

Description

Technical Field

[0001] This invention belongs to the interdisciplinary field of physical simulation of geological disasters and additive manufacturing technology. Specifically, it relates to a modeling method for a physical model of landslide disaster chain in a small watershed of granite weathering area based on a composite process of in-situ parameter inversion and 3D printing, as well as a modeling system for implementing the method. Background Technology

[0002] In the humid and hot regions of southern my country, where granite is widely distributed, weathering has created a thick, layered weathering crust profile. From top to bottom, this crust consists of multiple layers: residual soil, completely weathered, strongly weathered, and moderately weathered zones. The physical and mechanical parameters of each weathering zone exhibit a non-linear gradient, and residual joint surfaces evolving along primary joints are commonly found between the layers. Under heavy rainfall conditions, the strength of the residual soil and completely weathered zones drops sharply. Combined with the continuous hydrological replenishment of downstream slopes by upstream runoff from small watersheds, this can easily trigger cluster landslides, forming a landslide-gully blockage-breakage disaster chain, posing a serious threat to mountain infrastructure and the safety of people's lives and property.

[0003] Indoor physical model testing is an important method for studying the formation and evolution mechanisms of such disasters. Currently, several landslide physical model testing devices have been disclosed. For example, Chinese patent CN223581951U discloses a landslide physical model testing device with adjustable slip zone softening, simulating the softening effect of groundwater on the slip zone through a seepage system; Chinese patent CN121499768A discloses a landslide physical model testing device with segmented adjustable slip zone strength, using heating elements to segmentally heat jelly wax to simulate slip zone weakening; Chinese patent CN115754232B discloses a small-scale physical model experimental device for reservoir landslides with separable sliding bodies, simulating reservoir water level rise and fall conditions through a separable box. Furthermore, Chinese patent CN119007556B relates to physical model testing equipment for submarine landslides, and Chinese patent application CN202510590484.8 relates to a landslide disaster physical model testing system and method, which can adjust the slope angle and simulate rainfall and loading.

[0004] However, a comprehensive analysis of the existing technologies reveals the following common technical shortcomings:

[0005] (1) Insufficient ability to reproduce complex geological structures. The weathering crust of natural granite has a vertical multi-layered gradient structure (residual soil zone → completely weathered zone → strongly weathered zone → moderately weathered zone) and residual joint surfaces developed between layers. Existing devices and modeling methods (such as layered filling and compaction, casting, etc.) cannot simultaneously and accurately reproduce these two structural features in the physical model, resulting in poor physical and mechanical similarity between the model and the prototype, and making it difficult to guarantee the reliability of the test results.

[0006] (2) Lack of closed-loop mapping from in-situ parameters to model construction. In the existing technology, the material ratio and construction process parameters of the model are mostly determined by the researcher's experience. There is no quantitative mapping relationship between them and the actual survey parameters of the target site. The repeatability of model construction is poor, and the experimental results of different researchers and different batches are highly discrete.

[0007] (3) There is a lack of simulation methods for hydrological connectivity in small watersheds and cluster landslides. Existing devices are mostly designed for single landslides or single external force conditions (such as single rainfall or single loading), which are difficult to simulate the hydrological replenishment effect of upstream confluence on downstream slopes at the scale of small watersheds, as well as the entire process of the cluster landslide-ditch blockage-breakage disaster chain triggered by this.

[0008] Therefore, there is an urgent need to provide a modeling method and experimental system that can start from in-situ parameter identification, reproduce the complex structure of granite weathering crust with high fidelity through digital and automated means, and on this basis realize the simulation of the entire process of small watershed cluster landslide-ditch blocking-out disaster chain. Summary of the Invention

[0009] This invention aims to solve the technical problems of low accuracy in geological structure reproduction, broken parameter mapping, poor repeatability, and difficulty in simulating the entire process of disaster chain in small watersheds when constructing physical models of granite weathering crust landslides in the prior art. It provides a modeling method and modeling system based on in-situ parameter inversion and 3D printing additive-subtractive composite process.

[0010] To achieve the above objectives, the present invention provides the following technical solution:

[0011] A modeling method for a physical model of landslide hazard chains in a small watershed of a granite weathering zone includes the following steps:

[0012] Step S1: In-situ parameter identification and inversion

[0013] Survey data of a small watershed within the target granite weathering area was acquired. This data included surface morphology data from UAV LiDAR, high-density electrical resistivity profile data, and supplementary micro-penetration array data and surface microseismic data. The micro-penetration array data was obtained by deploying 5–8 shallow penetration points (depth ≤ 1.5 m) in the study area to acquire continuous penetration resistance curves, which were then calibrated and converted into cohesion and internal friction angles at each depth. The surface microseismic data was obtained by inverting surface wave dispersion to obtain the shallow shear wave velocity structure from environmental noise recorded at 6–10 short-period seismic nodes.

[0014] Based on the survey data, the three-dimensional spatial distribution of each weathering layer of the granite weathering crust and its physical and mechanical parameter fields were obtained by using the Bayesian-Kriging coupled inversion algorithm.

[0015] The physical and mechanical parameters include at least the thickness, interface morphology, bulk density, cohesion, internal friction angle, and permeability coefficient of each weathered layer.

[0016] Step S2: Digital Modeling and Printing Proportion Generation

[0017] Based on the three-dimensional spatial distribution and physical and mechanical parameter fields of each weathering layer obtained by inversion in step S1, a three-dimensional digital model containing information on multi-layer weathering crust structure, interlayer residual joint surfaces, and lateral variability of parameters within the layers is generated.

[0018] Meanwhile, the physical and mechanical parameters of each weathering layer obtained by inversion are input into the preset similar material knowledge graph and ratio generation module. Through active learning and recommendation, the 3D printing similar material ratio scheme corresponding to the target parameters of each weathering layer is output.

[0019] Step S3: 3D printing construction of weathered shell based on additive-subtractive composite process.

[0020] Input the three-dimensional digital model and proportioning scheme generated in step S2 into the 3D printing control system, and print each weathered layer from bottom to top in the model box;

[0021] The printing of each weathered layer employs the following composite process:

[0022] Additive manufacturing stage: According to the proportioning scheme of the weathered layer, a similar material slurry is extruded through the first printing nozzle, and the layers are stacked layer by layer to form the structure. The printed layers are then compacted synchronously by a compaction mechanism located behind the printing nozzle.

[0023] Subtractive processing stage: After the printing and compaction of the current weathered layer are completed, a wedge-shaped cutting tool is used to cut the surface of the current weathered layer to a controllable depth along the preset residual joint surface spatial path, thereby accurately constructing a mechanically weak surface that simulates the natural residual joint surface between layers.

[0024] Step S4: Model Integration and Monitoring System Deployment

[0025] During the construction process in step S3, sensors are pre-embedded at the designed locations to complete one or more landslide physical models with real weathering crust structural characteristics;

[0026] At least two completed physical models are arranged in series along a predetermined water flow direction, and adjustable overflow weirs are installed between adjacent models to form a series model channel simulating the hydrological connectivity conditions of a small watershed.

[0027] Further, the specific implementation of the Bayesian-Kriging coupled inversion algorithm described in step S1 is as follows: First, using the joint gradient abrupt change of resistivity and shear wave velocity, combined with topographic curvature analysis, the layered interfaces between residual soil zones, completely weathered zones, strongly weathered zones, and moderately weathered zones are automatically identified, and the thickness of each layer and the undulation morphology of the interface are output. Then, the cohesion, internal friction angle, unit weight, and permeability coefficient of each weathered layer are regarded as spatial random fields, and their spatial correlation structure (range, sill value) is obtained by fitting the experimental variogram function of the micro-penetration data. On this basis, a layered Bayesian model is constructed. The prior distribution is based on the geological statistics and penetration data of the study area, and the likelihood function simultaneously includes the apparent resistivity residual from the high-density electrical resistivity forward modeling, the residual from the shear wave velocity forward modeling, and the residual of the measured value at the penetration point. The integrated nested Laplace approximation (INLA) is used to quickly sample and solve the posterior distribution, and the mean and standard deviation of the parameters at each grid point are output.

[0028] (1)

[0029] In the formula, μ(x) is the mean of the parameters, and σ(x) is the standard deviation. For areas with a confidence level below 0.65, they are automatically marked as "encrypted printing areas" in the subsequent 3D digital model. During printing, the slice thickness of this area is reduced from the default value to 1.5 mm, and the number of compaction times is increased by 50%.

[0030] Through the above inversion, a stratigraphic parameter field with an error of less than 5% is obtained, and two output files are generated: one is a multi-attribute mesh file (VTK format) with parameter mean and standard deviation, and the other is an STL model file marked with "encrypted printing area" for subsequent differential slicing by the 3D printing system.

[0031] Further, in step S3: the nozzle diameter of the first printing nozzle is 3-8 mm, and the printing layer thickness is continuously adjustable from 0.5 to 5.0 mm; the second printing nozzle is used to extrude joint weak surface filling material in the additive manufacturing stage, and the nozzle diameter is 1-3 mm; the vibration frequency of the compaction mechanism is adjustable from 20 to 60 Hz; the tip angle of the irregular cutting tool is 15°-30°, and the cutting depth is 50%-80% of the already printed thickness of the weathered layer.

[0032] Furthermore, the spatial path of the residual joint surface mentioned in step S3 is a contour path automatically generated by a three-dimensional digital model based on the dip direction, dip angle, and density of the natural residual joint surface obtained by inversion in step S1.

[0033] Furthermore, the construction method of the similar material knowledge graph in step S2 is as follows: through no less than 500 sets of orthogonal experiments and literature data mining, a ternary graph is established with material composition (iron powder, quartz sand, barite powder, cement, clay, water), process parameters (printing layer thickness, compaction frequency, curing humidity), and mechanical parameters (bulk density, cohesion, internal friction angle, permeability coefficient) as nodes, and a graph neural network is used to learn continuous nonlinear mapping relationships. When the target mechanical parameter is within the existing data range, the graph provides high-precision interpolation results; when it exceeds the existing range, extrapolation recommendations are made based on material physicochemical constraints, and confidence levels are labeled.

[0034] Further, the active learning process described in step S2 is as follows: First, after inputting the mechanical parameter vector of the target weathered layer, the system retrieves the 5 to 10 nearest existing mix design schemes in the knowledge graph. Then, using the existing mix designs in the graph as training points, a sparse Gaussian process surrogate model is constructed, outputting the recommended mix design and its prediction variance σ. 2 If the ratio of the predicted standard deviation to the target parameter value is less than 5%, the recommended ratio is used directly. If the ratio is greater than 5%, the system automatically generates a set of candidate ratios and prompts for additional micro-sample verification tests.

[0035] Furthermore, when the parameter field obtained from the inversion in step S1 shows that the parameter difference within the same weathered layer exceeds 10%, step S2 supports the generation of continuous gradient proportions within the layer. When generating the 3D printing path, for each printing line, the linear or spline interpolation curve of the proportion as a function of distance is automatically calculated based on the boundaries of the sub-regions it passes through, and real-time valve control commands for the multi-feed system are generated to achieve continuous parameter gradient printing within the same weathered layer.

[0036] Furthermore, during the printing process in step S3, after each weathered layer is constructed, the system pauses and activates the automatic penetration detector to perform micro-penetration tests at 5-10 preset measuring points. The measured penetration resistance is calibrated and converted into cohesion or internal friction angle, which is then compared with the target value. If the deviation is less than 5%, the next layer is printed; if the deviation is between 5% and 10%, the system automatically adjusts the mix ratio or compaction number of subsequent layers to compensate; if the deviation is greater than 10%, printing is paused and an alarm is triggered. This closed-loop verification ensures that the error between the final constructed physical model and the target parameter field is stably controlled within 5%.

[0037] This invention also provides a modeling system for implementing the above-mentioned modeling method of physical modeling of landslide disaster chains in small watersheds of granite weathering areas, comprising:

[0038] In-situ parameter identification system for weathered crust: includes a drone equipped with lidar and a high-density electrical resistivity tomography instrument, as well as a data inversion module with the multi-layer thickness parameter inversion algorithm built in;

[0039] A 3D printing construction system for a multi-layered weathered crust structure includes a main model housing, a multi-axis 3D printer body, at least two printing nozzles (the first printing nozzle is used to extrude a weathered layer matrix-like material, and the second printing nozzle is used to extrude joint weak surface material), a compaction mechanism located behind the first printing nozzle, a non-shaped cutting tool for constructing interlayer residual joint surfaces and its driving robotic arm, and a control system that is communicatively connected to the data inversion module for generating the printing path and controlling the printing process.

[0040] Small watershed disaster chain simulation system: includes a model channel formed by connecting at least two physical models constructed by the 3D printing system along the direction of water flow, an adjustable triangular overflow weir set between adjacent models, an upstream confluence supply device that provides controllable flow runoff to the head of the connected channel, a rainfall simulation device erected above the connected channel, and various types of data monitoring devices buried in the model or erected on the outside.

[0041] Compared with the prior art, the present invention has the following beneficial effects:

[0042] (1) For the first time, a high-fidelity reproduction of the dual structure of "multi-layer gradient + interlayer joints" of granite weathering crust was achieved.

[0043] Existing technologies (such as prior art documents CN223581951U and CN121499768A) can only simulate homogeneous or single slip zones, failing to simultaneously represent the vertical parameter gradient of the weathering crust and the weak interlayer discontinuities. This invention, through a composite process of "additive manufacturing (3D printing layer-by-layer deposition) + subtractive manufacturing (precise cutting with wedge-shaped tools)," quantitatively reproduces these two key structural features in the same physical model for the first time. Actual measurements show that the error between the model and the prototype in key indicators such as layer thickness, density, strength, and joint weakening ratio can be controlled within 5%, significantly better than traditional manual filling methods (typically with an error of 15%–25%).

[0044] (2) A closed-loop, quantitative mapping link from "in-situ survey" to "model construction" was established.

[0045] In existing technologies, model material proportions and construction parameters are mostly determined based on experience (e.g., reference document CN115754232B uses manual filling). This invention, through a built-in multi-layer thickness parameter inversion algorithm identification system, directly converts high-density electrical resistivity tomography (EDS) and UAV data into a physical and mechanical parameter field. It also automatically outputs and prints proportioning schemes through a proportioning database, eliminating human experience errors and making model construction highly repeatable. The coefficient of variation of key monitoring data (such as landslide trigger time and peak outburst flow) from multiple repeated tests under the same working conditions can be controlled within 8%.

[0046] (3) For the first time, a physical simulation of the entire process of the disaster chain of “group landslide-ditch blocking-breach” in a small watershed was realized under indoor conditions.

[0047] Existing devices (such as prior art documents CN202510590484.8 and CN119007556B) are all designed for single landslides or single external force conditions. This invention, by connecting at least two physical models constructed using the aforementioned high-fidelity modeling method and setting up an adjustable overflow weir, simulates for the first time indoors the hydrological connectivity and recharge effect of a small watershed, as well as the entire sequence of disaster chain evolution triggered by multiple landslides, tiered damming, overtopping overflow, and final breach. This provides an unprecedented experimental platform for studying the mechanism of clustered watershed disasters.

[0048] (4) It provides a generalizable digital construction paradigm for physical models of geological bodies.

[0049] The modeling method of this invention is not limited to granite weathering crust. By adjusting the parameter inversion algorithm and the ratio database, it can be extended to the construction of physical models of other geological bodies with complex layered structures or developed structural surfaces (such as layered rock slopes, loess steep slopes, etc.). Attached Figure Description

[0050] Figure 1 This invention relates to a physical modeling device for landslide disaster chains in small watersheds of granite weathering areas.

[0051] Figure 2 This invention provides an experimental physical model of landslide disaster chains in small watersheds of granite weathering areas.

[0052] Figure 3 This is a diagram of the adjustable triangular overflow weir device of the present invention.

[0053] Figure 4 This is a flowchart of the modeling method in an embodiment of the present invention.

[0054] Figure 5 This is a distribution diagram of cohesion parameters at a certain location in an embodiment of the present invention.

[0055] Figure 6 This is a physical image of the small watershed model inside the model box in an embodiment of the present invention.

[0056] In the diagram: 1-3D printer body, 2-compaction mechanism, 3-water supply pipe, 4-compaction head of the compaction mechanism, 5-second printing nozzle, 6-first printing nozzle, 7-iron-shaped cutting tool, 8-constant pressure water tank, 9-3D printing clay substrate box, 10-3D printing cement substrate box, 11-3D printing barite powder substrate box, 12-3D printing quartz sand substrate box, 13-3D printing iron powder substrate box, 14-laptop, 15-3D substrate mixer, 16-3D printing material transfer pipe, 17-rainfall control. 18-Rainwater pipe network, 19-Suspended basket support frame, 20-Adjustable triangular overflow weir, 21-Tension sensor, 22-Sediment diversion channel, 23-Stainless steel hanging basket, 24-Model box pulley, 25-Water level sensor, 26-Box bolt, 27-Miniature earth pressure sensor, 28-Miniature pore water pressure sensor, 29-Transparent plexiglass model main box, 30-Rainfall nozzle, 31-Lifting screw, 32-Upstream fixed weir body, 33-Modible weir bottom plate, 34-Modible weir plate, 35-Downstream fixed weir body. Detailed Implementation

[0057] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are merely some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0058] Example

[0059] This embodiment provides a modeling method and system for a physical model of landslide disaster chain in a small watershed of granite weathering area, which is used to simulate the evolution process of a cluster landslide-ditch blockage-breakdown disaster chain in a small granite watershed in South China under rainstorm conditions.

[0060] I. Composition of the test system

[0061] The modeling system in this embodiment consists of three parts: an in-situ parameter identification system for weathered crust, a 3D printing construction system for multi-layered weathered crust structures, and a small watershed disaster chain simulation system.

[0062] The in-situ parameter identification system for weathering crust includes: a DZD-6A multi-functional high-density electrical resistivity meter (60 electrodes, 1m electrode spacing), three Kelida Phantom H700A UAVs (all equipped with Zenmuse L2 lidar), a data inversion module (with built-in resistivity-intensity conversion model and the granite multi-layer weathering layer thickness parameter inversion algorithm proposed in this invention), and a mix proportion generation module (with built-in orthogonal experimental mix proportion database).

[0063] The 3D printing system for constructing multi-layered weathered crust structures includes: a robotic arm 3D printer body 1, a compaction mechanism, a joint surface forming component, and a control system. The effective travel of the robotic arm is 1200mm on the X-axis, 400mm on the Y-axis, and 500mm on the Z-axis, with a positioning accuracy of ±0.1mm. The first printing nozzle 6 has a nozzle diameter of 5mm, and the second printing nozzle 5 has a nozzle diameter of 2mm. The compaction head 4 of the compaction mechanism is 200mm × 100mm in size, and its vibration frequency is infinitely adjustable from 20 to 60Hz. The irregularly shaped cutting tool 7 of the joint surface forming component has a 20° tip angle and a cutting needle height of 15mm.

[0064] The small watershed disaster chain simulation system includes: three tandem transparent plexiglass model main boxes, each 1200mm long × 1000mm wide × 1500mm high; and three adjustable triangular overflow weirs 20 (structure as follows). Figure 3 The equipment includes an upstream confluence and replenishment device (a constant pressure water tank with an effective volume of 200L and a flow rate adjustment range of 0-5L / min), a rainfall simulation device (rainfall intensity continuously adjustable from 10 to 150mm / h), and a data measurement device (water level sensor 25, miniature earth pressure cell 27, miniature pore water pressure gauge 28, tension sensor 21, stainless steel hanging basket 23, and 5 GoPro action cameras).

[0065] II. Specific Implementation Steps of the Modeling Method

[0066] Step S1: In-situ parameter identification and inversion

[0067] (1) Surface mapping was carried out using UAVs equipped with lidar in the small watershed of the target granite weathering area. Three UAVs were used to scan the area separately, with a flight path speed of 8 m / s, a lidar scanning distance of 10 m, and a safety distance of 2 m. Surface data with a planar accuracy of 0.6 mm and an elevation accuracy of 0.4 mm were acquired. A complete three-dimensional digital model of the small watershed was constructed using DJI AI Model software, saving only the surface data and constructing a three-dimensional STL model without stratigraphic parameters.

[0068] (2) Supplementary acquisition of micro-penetration array data and surface microseismic data: Six shallow micro-penetration points (1.5m deep) were set up along the main channel and the slopes on both sides of the small watershed in the study area. Electronic micro-penetrators were used to continuously penetrate at a speed of 2cm / s, and the penetration resistance curves were recorded. After indoor calibration tests, the cohesion and internal friction angle at each depth were converted. At the same time, eight short-period seismic nodes (spaced 20m apart) were set up, and environmental noise was continuously collected for 48 hours. The shallow (0-15m) shear wave velocity structure was obtained by inversion using the surface wave dispersion analysis method, with a grid resolution of 1m×1m×0.5m.

[0069] (3) A 60m long survey line was laid out along the slope of the target granite weathering area, with an electrode spacing of 1.0m. High-density electrical resistivity data was collected using a Wenner device. The collected apparent resistivity data was then converted into a format and imported into the data inversion module.

[0070] (4) Bayesian-Kriging Coupled Inversion: First, using the joint gradient abrupt change of resistivity and shear wave velocity, combined with topographic curvature analysis, the layered interfaces between residual soil zone, completely weathered zone, strongly weathered zone, and moderately weathered zone are automatically identified, and the thickness of each layer is extracted: 6.2 cm for residual soil zone (model scale 1:100, corresponding to 6.2 m on site), 8.5 cm for completely weathered zone, 12.0 cm for strongly weathered zone, and 18.0 cm for moderately weathered zone. Then, the cohesion, internal friction angle, unit weight, and permeability coefficient of each weathered layer are regarded as spatial random fields, and their spatial correlation structure (range, sill value) is obtained by fitting the experimental variogram function of micro-penetration data. A layered Bayesian model is constructed, with the prior distribution based on the geological statistics and penetration data of the study area. The likelihood function simultaneously includes the apparent resistivity residual from the high-density electrical resistivity forward modeling, the residual from the shear wave velocity forward modeling, and the residual of the measured value at the penetration point. The Integrated Nested Laplace Approximation (INLA) was used to quickly sample and solve the posterior distribution, outputting the mean and standard deviation of the parameters for each grid point (grid size 0.5m × 0.5m × 0.2m, model scale 1:100). Comparison with five sets of verification borehole sampling data showed that the error of the inversion parameters was less than 5%.

[0071] (5) Confidence marking and identification of encrypted printing areas: Calculate the confidence level of each grid point. For areas with a confidence level below 0.65 (mainly located near the interface between the fully weathered zone and the strongly weathered zone, as well as the slope toes on both sides of the gully), they are automatically marked as "encrypted printing areas" in the three-dimensional digital model.

[0072] (6) Generate output files: one is a multi-attribute mesh file (VTK format) with parameter mean and standard deviation, and the other is an STL model file with "encrypted print area" mark.

[0073] Step S2: Digital Modeling and Printing Proportion Generation

[0074] (1) Construction of a knowledge graph of similar materials: A database of similar material proportions for granite weathering crusts was established through 550 sets of orthogonal experiments in the early stage. Based on this, material components (iron powder, quartz sand, barite powder, cement, clay, water), process parameters (printing layer thickness, compaction frequency, curing humidity), and mechanical parameters (bulk density, cohesion, internal friction angle, permeability coefficient) were used as nodes. A graph neural network was used to learn the nonlinear mapping relationship between nodes to form a dynamically expandable knowledge graph. The cohesion range of the graph is 5 to 60 kPa, and the internal friction angle range is 15° to 45°, covering the parameter space of all weathering layers in the study area.

[0075] (2) Active learning of mix design recommendation: The target parameters of each weathered layer obtained from step S1 (taking residual soil zone as an example: cohesion 12.5 kPa, internal friction angle 24.0°, permeability coefficient 2.3 × 10⁻⁶) are used. -3 The system inputs the mix proportion generation module (cm / s). It retrieves the eight nearest existing mix proportion schemes from the knowledge graph, constructs a sparse Gaussian process surrogate model, and outputs the recommended mix proportions and their predicted variances. For the residual soil zone, the ratio of the predicted standard deviation to the target parameter value is 3.2%, less than 5%, so the recommended mix proportion is directly adopted. For the part of the fully weathered zone near the dense printing area, the ratio of the predicted standard deviation to the target parameter value is 6.5%, with a slightly higher predicted variance. The system automatically generates three candidate mix proportions and suggests adding micro-samples for verification. The laboratory then tests with 3D-printed samples, and after two hours, selects the scheme that best matches the target parameters (error 2.8%). The final mix proportion schemes for each weathered layer are shown in Table 1.

[0076] Table 1. Material proportioning schemes for similar materials in each weathering layer (mass percentage)

[0077]

[0078] (3) Generation of gradient mix proportions within the layer: The inversion results of step S1 show that the cohesion in the channel area (approximately 10.2 kPa) and the slope shoulder area (approximately 14.8 kPa) within the residual soil zone differ by 31%, exceeding the 10% threshold. The system automatically enables the continuous gradient printing function within the layer: When generating the 3D printing path, for each printing line, based on the boundary of the sub-region it passes through, the linear interpolation curve of the mix proportion with distance is calculated, and real-time valve control commands for the multi-feed system are generated to achieve continuous parameter gradient changes within the same weathered layer.

[0079] Step S3: 3D printing construction of weathered crust based on additive-subtractive composite process

[0080] (1) According to the proportioning scheme in Table 1, weigh out each component raw material and load it into the independent hopper of the feeding system. Import the three-dimensional digital model file generated by the in-situ parameter recognition system into the control system operation interface. The model contains the thickness data of each weathered layer and the spatial distribution data of the residual joint surface. The printing path planning module slices the model, sets the slice thickness to 3.0 mm, and generates the layer-by-layer printing trajectory and the residual joint surface forming path.

[0081] (2) The construction process begins from the bottom of the main body of the model and is printed from bottom to top in the order of moderately weathered zone → strongly weathered zone → completely weathered zone → residual soil zone. See Table 2 for specific printing process parameters.

[0082] Table 2 3D printing process parameters for each weathered layer

[0083]

[0084] (3) After each weathered layer is constructed, the system pauses and activates the automatic penetration detector. This detector is a small, self-propelled penetrator (weighing 180g, traveling at a speed of 2cm / s), which performs micro-penetration tests (1.5cm deep) at six preset measurement points (including the ordinary zone and the dense printing zone). The measured penetration resistance is converted into cohesion and internal friction angle by the calibration curve. Compared with the target value, the deviation in the ordinary zone is within 3%, and the next layer is printed. In the dense printing zone, one measurement point has a deviation of 6.5%. The system automatically adjusts the number of compaction times for the corresponding printing layer in the subsequent fully weathered zone from 5 to 7 times. Subsequent retests show that the deviation has decreased to 2.1%.

[0085] Taking weathering zone printing as an example to illustrate the composite process:

[0086] Additive manufacturing stage: The first printing nozzle 6 extrudes the weathered matrix material according to the planned trajectory, with a printing speed of 45 mm / s. The extrusion flow rate is matched with the travel speed to maintain a single-layer thickness of 5.0 mm. After each layer is printed, the compaction mechanism immediately follows to perform the compaction operation, with a vibration frequency of 30 Hz. The pressure adjustment spring preload maintains the contact pressure at approximately 5 kPa, and the compaction is performed 8 times. After calibration, the actual bulk density of the compacted material deviates from the target bulk density by 2.7%.

[0087] Subtractive Processing Stage: After the entire thickness of the moderately weathered zone has been printed and compacted, the joint surface forming component is activated. The shaped cutting tool 7 moves to the starting point of the planned residual joint surface forming path on the current layer surface, and the cutting needle azimuth angle is adjusted to 35°, consistent with the measured joint surface inclination angle on site. The cutting needle travels along the path at a speed of 1.5 cm / s, while simultaneously cutting vertically into the material, with the cutting depth controlled to 50% (approximately 9 cm) of the currently printed layer thickness (moderately weathered zone). The shaped cutting tool has a 20° tip angle, and during the cutting-out process, it forms a weak fracture surface with geometric similarity to the natural joint surface through wedge-shaped extrusion within the material. After extraction, the fracture surface undergoes moderate closure under the pressure of the overburden layer. Subsequent direct shear tests verified that the cohesion of this weak surface is reduced by approximately 42% compared to the matrix material, and the internal friction angle is reduced by approximately 28%, which highly matches the mechanical weakening characteristics of the residual joint surface measured on site.

[0088] (4) Print the strongly weathered zone, the completely weathered zone and the residual soil zone in sequence according to the same process. When printing to the middle of the completely weathered zone and the middle of the residual soil zone, stop the operation and bury two miniature earth pressure cells and two pore water pressure gauges at a depth of half the thickness of the corresponding layer. The sensor wires are led out from the reserved holes on the side wall of the box and then sealed.

[0089] The three main model boxes were all 3D printed using the above process to create multi-layered weathered shell structures. The small watershed models inside the model boxes are as follows: Figure 6 As shown.

[0090] Step S4: Model Integration and Monitoring System Deployment

[0091] (1) Arrange the three completed model main boxes (boxes A, B, and C) sequentially along the water flow direction on the test platform, with box A at the upstream end and box C at the downstream end. Install an adjustable triangular overflow weir 20 between box A and box B, and between box B and box C. The fixed weir is fixed to the downstream box inlet face by box bolts 26, and the movable weir plate 34 is embedded in the vertical sliding groove. Rotate the lifting screw 31 to adjust the elevation of the top surface of the movable weir plate of the overflow weir from box A to box B to 6.0 cm higher than the bottom plate of the channel of box A, and adjust the elevation of the top surface of the movable weir plate of the overflow weir from box B to box C to 5.0 cm higher than the bottom plate of the channel of box B.

[0092] (2) Connect the outlet of the constant pressure water tank 8 to the water distribution network 18 laid in the upstream area of ​​the tank A through the flow regulating valve. The rain nozzles 30 of the rain device are installed directly above the three main tanks of the model. The nozzles are installed at a height of 1.2m above the slope. The spray coverage area has been tested and can completely cover the slope area of ​​the three tanks.

[0093] (3) Along the longitudinal direction of the series channel, one water level sensor 25 is arranged in the middle of box A, the middle of box B, the middle of box C, and at the end of box C. The sensor probe is fixed 2.5cm above the bottom plate of the channel. A stainless steel hanging basket 23 is suspended below the end of box C. The bottom plate and side wall of the hanging basket have water permeable holes with a diameter of 2mm (smaller than the minimum particle size of similar materials of 3mm). The hanging basket is connected to a tension sensor 21 with a range of 50kg by a steel wire rope. The motion cameras are arranged as follows: one is set up on the side of box A, with the optical axis perpendicular to the long axis of the landslide channel; four are set up vertically in a 2×2 array above the blockage and breach area between box B and box C. The overlap rate of the field of view of adjacent cameras is about 65%.

[0094] III. Disaster Chain Simulation Experiment and Data Acquisition

[0095] (1) Set the test conditions: rainfall intensity 80 mm / h, upstream runoff flow rate 1.2 L / min. Start the water pump and rainfall device to begin applying hydrological load to the model.

[0096] (2) Eight minutes after the start of rainfall, the pore water pressure gauge reading embedded in the residual soil zone of box A rose from the initial -5.2 kPa to -1.8 kPa, indicating a significant decrease in matrix suction. At the 12th minute, obvious seepage points appeared at the toe of the slope of box A, and several longitudinal tension cracks appeared on the slope surface. At the 15th minute, the gate at the front end of box A was opened, and the residual soil zone and the material from the completely weathered zone moved downstream along the landslide trough, initiating the landslide.

[0097] (3) After the sliding material in box A enters the channel, it accumulates on the upstream side of the triangular overflow weir between box A and box B, forming a stable dam body at the 18th minute. The overflow weir outflow rate drops sharply from 1.2 L / min to 0.1 L / min. The water level in the channel of box A begins to rise. At the 22nd minute, the water level exceeds the top elevation of the movable weir plate of the overflow weir by 6.0 cm and begins to overflow into box B.

[0098] (4) Under the combined effects of overflow and continuous rainfall, a landslide occurred on the slope of box B at the 28th minute. The landslide material entered the channel between box B and box C, forming a second dam at the inlet of box C. The water level in the channel of box B rose rapidly, and after the water level exceeded the top surface of the overflow weir by 5.0 cm at the 35th minute, it overflowed into box C.

[0099] (5) At the 42nd minute, the dam body at the inlet of box C experienced piping failure under the continuous overflow of water flow, and the downstream face of the dam body collapsed significantly. At the 45th minute, the dam body collapsed as a whole, and the floodwater carrying a large amount of mud and sand rushed into the downstream hanging basket of box C.

[0100] (6) Throughout the experiment, the data acquisition instrument continuously recorded the data from each sensor at a sampling frequency of 20Hz. The water level sensor recorded that the water level at the end of tank C rose sharply from 4.2cm to 12.8cm at the moment of dam failure, and the peak flow rate of the breach was converted to 0.38L / s. The time history curve of the mass of the basket collected by the tension sensor showed that the cumulative increase in the mass of soil and rock transported during the breach stage was 2.86kg, and the peak erosion sediment yield rate was 0.21kg / s.

[0101] (7) The velocity field of the landslide at box A was obtained through non-contact measurement and analysis of images from four motion cameras. The maximum velocity at the leading edge of the landslide reached 0.42 m / s. The synchronous images from the four vertical cameras were reconstructed using motion recovery structures to obtain a three-dimensional geometric model of the landslide dam at five different moments during the breach. This quantitatively revealed that the dam crest elevation decreased from the initial 12.5 cm to 4.8 cm after the breach, and the dam volume decreased from 5860 cm³. 3 Reduced to 2140cm 3 The evolutionary pattern.

[0102] IV. Model Fidelity and Repeatability Verification

[0103] (1) Model fidelity verification

[0104] After construction in step S3, micro-samples were drilled from the residual soil zone, completely weathered zone, strongly weathered zone, and moderately weathered zone of model box A for indoor direct shear tests and permeability tests. The test results were compared with the target parameters obtained from the inversion in step S1, and the results are shown in Table 3.

[0105] Table 3 Comparison of measured parameters and target parameters of the model

[0106]

[0107] The above results show that the physical model constructed by the modeling method of the present invention has an error of less than 5% in the key mechanical parameters and the target site, which is significantly better than the traditional manual filling method, which usually has an error of 15% to 25%.

[0108] (2) Reproducibility verification

[0109] To verify the repeatability of this modeling method, three physical models (numbered M1, M2, and M3) were repeatedly constructed under identical modeling parameters and printing process conditions, and a disaster chain modeling experiment was conducted under the same hydrological conditions (80 mm / h rainfall and 1.2 L / min upstream flow). The key characteristic parameters of the three models are compared in Table 4.

[0110] Table 4 Comparison of Repeatability Test Results

[0111]

[0112] The above results demonstrate that the modeling method of this invention has excellent repeatability, providing a reliable model basis for quantitatively studying the evolution of disaster chains under different working conditions.

[0113] The above results demonstrate that the modeling method and system of this invention can effectively reflect the influence of different hydrological boundary conditions on the evolution of landslide disaster chains in small watersheds of granite weathering areas, and verify its effectiveness and sensitivity in simulating the entire sequence of disaster chains of clustered landslides-ditch blocking-out breaches.

[0114] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art can make various improvements and modifications without departing from the spirit and principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A modeling method for a physical model of landslide disaster chains in a small watershed of a granite weathering zone, characterized in that, Includes the following steps: Step S1: In-situ parameter identification and inversion Acquire survey data of a small watershed in the target granite weathering area. The survey data includes UAV lidar surface morphology data, high-density electrical resistivity profile data, and supplementary micro-penetrating array data and surface microseismic data. Based on the survey data, the three-dimensional spatial distribution of each weathering layer of the granite weathering crust and its physical and mechanical parameter fields were obtained by using the Bayesian-Kriging coupled inversion algorithm. The physical and mechanical parameters include at least the thickness, interface morphology, bulk density, cohesion, internal friction angle, and permeability coefficient of each weathered layer; Step S2: Digital Modeling and Printing Proportion Generation Based on the three-dimensional spatial distribution and physical and mechanical parameter fields of each weathering layer obtained by step S1, a three-dimensional digital model containing information on the multi-layer weathering crust structure, interlayer residual joint surfaces, and lateral variability of parameters within the layers is generated. The physical and mechanical parameters of each weathering layer obtained by inversion are input into the preset similar material knowledge graph and ratio generation module. Through active learning and recommendation, the 3D printing similar material ratio scheme corresponding to the target parameters of each weathering layer is output. Step S3: 3D printing construction of weathered crust based on additive-subtractive composite process Input the three-dimensional digital model and proportioning scheme generated in step S2 into the 3D printing control system, and print each weathered layer from bottom to top in the model box; The printing of each weathered layer employs the following composite process: Additive manufacturing stage: According to the proportioning scheme of the weathered layer, a similar material slurry is extruded through the first printing nozzle, and the layers are stacked layer by layer to form the structure. The printed layers are then compacted synchronously by a compaction mechanism located behind the printing nozzle. Subtractive processing stage: After completing the printing and compaction of the current weathered layer, a wedge-shaped cutting tool is used to cut the surface of the current weathered layer to a controllable depth along the preset residual joint surface spatial path, thereby accurately constructing a mechanically weak surface between the layers that simulates the natural residual joint surface. Step S4: Model Integration and Monitoring System Deployment During the construction process in step S3, sensors are pre-embedded at the designed locations to complete one or more landslide physical models with real weathering crust structural characteristics; At least two completed physical models are arranged in series along a predetermined water flow direction, and adjustable overflow weirs are installed between adjacent models to form a series model channel simulating the hydrological connectivity conditions of a small watershed.

2. The modeling method according to claim 1, characterized in that, The specific implementation of the Bayesian-Kriging coupled inversion algorithm in step S1 is as follows: By utilizing the combined gradient abrupt change of resistivity and shear wave velocity, combined with topographic curvature analysis, the layering interface between each weathering layer is automatically identified, and the thickness of each layer and the undulation morphology of the interface are output. The cohesion, internal friction angle, bulk density, and permeability coefficient of each weathering layer are regarded as spatial random fields, and their spatial correlation structure is obtained by fitting the experimental variogram function of micro-penetration data. A hierarchical Bayesian model was constructed. The prior distribution was based on the geological statistics and penetration data of the study area. The likelihood function simultaneously included the apparent resistivity residual from the high-density electrical resistivity forward modeling, the residual from the shear wave velocity forward modeling, and the residual from the measured value at the penetration point. An integrated nested Laplace approximation is used to quickly sample and solve the posterior distribution, outputting the mean and standard deviation of the parameters for each grid point; For areas with a confidence level below a preset threshold, they are automatically marked as "encrypted printing areas" in the 3D digital model.

3. The modeling method according to claim 1, characterized in that, In step S3: The nozzle diameter of the first print head is 3-8mm, and the printing layer thickness is continuously adjustable from 0.5-5.0mm. The second printing nozzle is used to extrude joint-weak surface filling material during the additive manufacturing stage, and the nozzle diameter is 1-3 mm; The vibration frequency of the compaction mechanism is adjustable from 20 to 60 Hz; The blade tip angle of the irregular cutting tool is 15° to 30°, and the cutting depth is 50% to 80% of the already printed thickness of the weathered layer.

4. The modeling method according to claim 1, characterized in that, The spatial path of the residual joint surface mentioned in step S3 is a contour path automatically generated by a three-dimensional digital model based on the dip direction, dip angle and density of the natural residual joint surface obtained by inversion in step S1.

5. The modeling method according to claim 1, characterized in that, The construction method of the similar material knowledge graph in step S2 is as follows: through no less than 500 sets of orthogonal experiments and literature data mining, a triplet graph with material composition, process parameters and mechanical parameters as nodes is established, and a graph neural network is used to learn continuous nonlinear mapping relationships; when the target mechanical parameter is within the range of existing data, the graph gives the interpolation result; when it exceeds the existing range, extrapolation recommendation is performed based on the physical and chemical constraints of the material and the confidence level is marked.

6. The modeling method according to claim 1, characterized in that, The active learning process described in step S2 is as follows: After inputting the mechanical parameter vector of the target weathered layer, the system retrieves the 5 to 10 nearest existing proportioning schemes in the knowledge graph; Using the existing proportions in the map as training points, a sparse Gaussian process surrogate model is constructed to output the recommended proportions and their prediction variance. When the ratio of the predicted standard deviation to the target parameter value is less than 5%, the recommended ratio should be used directly. When the ratio of the predicted standard deviation to the target parameter value is greater than 5%, the system automatically generates a set of candidate ratios and prompts for additional micro-sample verification tests.

7. The modeling method according to claim 1, characterized in that, When the parameter field obtained by the inversion in step S1 shows that the parameter difference within the same weathered layer exceeds 10%, step S2 supports the generation of continuous gradient ratio within the layer: when generating the 3D printing path, for each printing line, the linear or spline interpolation curve of the ratio with distance is automatically calculated according to the boundary of the sub-region it passes through, and the real-time valve control command of the multi-way feeding system is generated to realize the continuous parameter gradient printing within the same weathered layer.

8. The modeling method according to claim 1, characterized in that, During the printing process in step S3, after each weathered layer is constructed, the system pauses and activates the automatic penetration detector to perform micro-penetration tests at 5-10 preset measuring points. The measured penetration resistance is calibrated and converted into cohesion or internal friction angle, and compared with the target value. If the deviation is less than 5%, continue printing the next layer; If the deviation is between 5% and 10%, the system will automatically adjust the mix ratio or compaction number of subsequent layers to compensate. If the deviation is greater than 10%, printing will be paused and an alarm will be triggered.

9. A modeling system for implementing the modeling method described in any one of claims 1 to 8, for a physical model of a landslide hazard chain in a small watershed of a granite weathering area, characterized in that, include: In-situ parameter identification system for weathered crust: including a drone equipped with lidar, a high-density electrical resistivity tomography instrument, and a data inversion module with a built-in Bayesian-Kriging coupled inversion algorithm; A multi-layered weathering crust construction system includes a main model housing, a multi-axis printer body, at least two printing nozzles, a compaction mechanism located behind the first printing nozzle, a non-standard cutting tool for constructing residual joint surfaces between layers and its driving robotic arm, and a control system that is communicatively connected to the data inversion module for generating printing paths and controlling the construction process; wherein, the first printing nozzle is used to extrude a weathering layer matrix-like material, and the second printing nozzle is used to extrude joint weak surface material; Small watershed disaster chain simulation test system: includes a model channel formed by connecting at least two physical models constructed by the construction system in the direction of water flow, an adjustable triangular overflow weir set between adjacent models, an upstream confluence supply device that provides controllable flow runoff supply to the head end of the connected channel, a rainfall simulation device erected above the connected channel, and multiple types of data monitoring devices buried in the model or erected on the outside.

10. The modeling system according to claim 9, characterized in that, The adjustable triangular overflow weir includes a lifting screw, an upstream fixed weir body, a movable weir bottom plate, a movable weir plate, and a downstream fixed weir body. The elevation of the top surface of the movable weir plate is adjusted by rotating the lifting screw to change the overflow starting water level from the upstream model to the downstream model. The multi-type data monitoring device includes a water level sensor, a micro earth pressure sensor, a micro pore water pressure sensor, a tension sensor, and at least one motion camera.

Images