Intelligent rainfall simulation system and method for slope stability experiment

Abstract

The invention belongs to the technical field of geotechnical engineering disasters, and particularly relates to an intelligent rainfall simulation system and method for a slope stability experiment. The system comprises an intelligent rainfall simulation unit, a comprehensive monitoring unit, a real-time data fusion and adaptive control unit and a data analysis-visualization unit which are organically unified and matched with one another. According to the invention, intelligent control of the rainfall process can be realized, and the rainfall intensity, raindrop size and rainfall uniformity can be accurately adjusted; the system is high in integration level and high in automation degree, real-time multi-dimensional monitoring can be achieved, and high-precision and real-time analysis of experimental data and automatic risk early warning are achieved. Advanced monitoring means such as DIC full-field strain, optical fibers and acoustic emission are adopted, the experiment precision is remarkably improved, an accurate basis is provided for slope stability analysis and disaster prevention and control, and the method can be applied to the fields of engineering slope safety evaluation, education and public safety and has a good application prospect in the aspects of geological engineering, geotechnical engineering, colleges and scientific research and teaching, emergency drilling and science popularization and the like. And safety and disaster prevention awareness are improved.

Description

Technical Field

[0001] This invention belongs to the field of geotechnical engineering disaster technology, specifically relating to an intelligent rainfall simulation system and method for slope stability experiments. Background Technology

[0002] Landslides are common natural disasters. With global warming, landslides, debris flows, and other disasters triggered by heavy rainfall and extreme weather are becoming increasingly common, attracting widespread attention from geological hazard researchers worldwide. Rainfall and earthquakes are the two most significant factors inducing landslides. In my country, the vast majority of landslide disasters are triggered by rainfall. In-depth research into slope stability under rainfall conditions and the coupled effects of earthquakes and rainfall is of theoretical guiding significance for landslide prediction, forecasting, and the design and construction of slope engineering projects.

[0003] As a crucial structural unit in open-pit coal mine operations, the stability of slopes directly impacts the safety and economic benefits of the mine. However, extreme weather events (such as torrential rain, extreme heat, rapid cooling, and freeze-thaw cycles) make slope rock and soil structures more susceptible to instability and landslides. These disasters not only disrupt mining operations but can also trigger serious safety accidents, endangering lives and property. Therefore, researching and revealing the impact of precipitation on slope stability is of paramount importance for ensuring the safety of coal mining operations.

[0004] Under rainfall conditions, slope stability becomes more critical. Rainwater infiltration alters the physical and mechanical properties of the slope soil or rock, leading to slope instability, landslides, and other disasters. Most existing rainfall simulation systems employ simple spraying technology, resulting in poor rainfall uniformity, insufficient simulation accuracy, and difficulty in accurately simulating subtle changes during actual rainfall. Furthermore, they lack real-time monitoring and intelligent control, failing to fully reflect the true instability mechanisms of slopes during rainfall. Specifically, this includes:

[0005] (1) Isolated monitoring: Traditional systems only use displacement and humidity sensors, which cannot capture the evolution of damage inside and outside the slope. A single sensor cannot cover multi-scale signals such as surface displacement (DIC), internal strain (fiber optic), and micro-fractures (acoustic emission).

[0006] (2) Control lag: Rainfall control relies on preset programs and cannot respond to real-time deformation. There is a lack of an effective intelligent monitoring system, which cannot monitor changes in the mechanical properties of the slope after rainfall infiltration, such as pore water pressure, water content, strain changes, and crack propagation, and cannot provide accurate risk warnings. In addition, the rainfall system has a single control method and insufficient intelligence, which cannot realize automated rainfall process control and lacks a linkage mechanism with the monitoring system.

[0007] (3) Data fragmentation: Data obtained from contact sensors such as pore water pressure sensors and water content sensors are processed independently from non-contact sensors such as DIC (digital image correlation) and 3D scanning, without forming a collaborative analysis.

[0008] (4) The implementation method does not meet the actual rainfall conditions: At present, most similar schemes use traditional artificial control spraying methods for simulation, which have problems such as poor control of rainfall particle size, insufficient uniformity of rainfall intensity, and single monitoring methods. They cannot truly reflect the mechanism of gradual slope damage caused by rainfall, and cannot accurately control rainfall intensity and raindrop size, resulting in a large deviation between experimental results and actual conditions, which seriously restricts the depth and accuracy of related experimental research. Summary of the Invention

[0009] The technical problem to be solved by this invention is to provide an intelligent rainfall simulation system and method for slope stability experiments. It proposes a three-dimensional integrated intelligent system of rainfall-monitoring-control, which can realize intelligent control of the rainfall process, and the rainfall intensity, raindrop size and rainfall uniformity can be precisely adjusted. The system has a high degree of integration and automation, and can monitor in real time and in multiple dimensions, so as to realize high-precision experimental data, real-time analysis and automatic risk warning.

[0010] The technical solution adopted is as follows:

[0011] A smart rainfall simulation system for slope stability experiments includes a smart rainfall simulation unit, a comprehensive monitoring unit, a real-time data fusion and adaptive control unit, and a data analysis and visualization unit. The units are organically integrated and work together.

[0012] The intelligent rainfall simulation unit includes a water supply device, a variable frequency constant pressure device, a rainfall simulation frame device, a slope simulation platform, and a rainfall location and intensity control device.

[0013] The integrated monitoring unit includes internal and external monitoring devices. The internal monitoring devices are embedded inside the slope similarity model and include pore water pressure sensors, water content sensors, fiber optic sensors, and acoustic emission sensors. The external monitoring devices include a DIC full-field strain measurement system, a three-dimensional laser scanner, and a high-speed camera.

[0014] The real-time data fusion and adaptive control unit (i.e., intelligent control unit) includes a PLC controller, a real-time data fusion module, and an LSTM-PID combined rainfall adjustment algorithm module. It can automatically control the operation of the rainfall system according to the preset rainfall intensity and duration, and automatically adjust the rainfall intensity based on sensor data feedback.

[0015] The data analysis and visualization unit includes a data processing module and a visualization module. The data processing module synchronously collects, processes, and stores data, performs statistical analysis, strain field analysis, and permeability analysis on the collected data, and generates monitoring curves for displacement, pore water pressure, water content, and acoustic emission signals in real time. The visualization unit generates a real-time three-dimensional slope model and dynamically displays the slope deformation and crack development process.

[0016] Preferably, the rainfall simulation frame device includes a cuboid main frame, with a movable support, a connecting support, and a water supply pipeline at the top of the main frame. At least one water supply pipeline is provided, and multiple nozzle groups are installed on each water supply pipeline.

[0017] Preferably, the water supply pipeline connects the water supply device and the variable frequency constant pressure device; each water supply pipeline is equipped with at least two multi-point nozzle groups, each of which is a quick-replaceable dual-nozzle group with a programmable raindrop kinetic energy range of 0.02–0.5 J, a rainfall intensity control range of 0.1–50 mm / h, and a raindrop particle size range of 0–5 mm. The dual-nozzle groups are installed using quick-release clips, allowing for rapid replacement.

[0018] Preferably, the external monitoring device is mounted on a transfer bracket, namely an XYZ three-axis moving platform (positioning accuracy 0.05mm), with a platform travel of 2500×1800×400mm, capable of scanning a 50mm×50mm grid. It includes a Y-axis moving platform, an X-axis moving platform, and a Z-axis moving platform. The integrated components of the DIC full-field strain measurement system, the three-dimensional laser scanner, and the high-speed camera can be installed on the X-axis moving platform, and adjustment handles are provided at both ends of the X-axis moving platform.

[0019] Alternatively, the integrated component of the externally monitored DIC full-field strain measurement system, 3D laser scanner, and high-speed camera is a detachable structure, which can be installed on the slide or crossbeam of any axis of the XYZ three-axis moving platform (including the X, Y, or Z axis). When the installation position of the integrated component changes, the following coordinate transformation and calibration process ensures that the scanning grid is consistent with the field of view:

[0020] (1) Establish the correspondence between the world coordinate system ΣW and the three-axis platform coordinate systems ΣX, ΣY, and ΣZ of the mobile platform and the slope model under test;

[0021] (2) Use a checkerboard or marker points for extrinsic parameter calibration, and calculate the rigid transformation matrix T from the sensor coordinate system to ΣW. CW ;

[0022] (3) Generate a scanning path in the X–Y plane (or the Y–X plane when mounted on the Y axis) according to a 50mm×50mm grid. The Z axis is used for working distance tracking and pitch attitude compensation.

[0023] (4) Set cable drag chains and vibration damping constraints for different mounting axes to ensure that the platform positioning accuracy and stroke meet the measurement requirements;

[0024] (5) After completing the path tracking test and uniformity verification, it will be put into operation.

[0025] The above arrangement enables the integrated components to still achieve the predetermined grid scanning and synchronous imaging after the mounting axis is changed, without changing the system's functional and performance boundaries.

[0026] Preferably, at least one side of the main frame is transparent visual glass, the transfer bracket is set on the outside of the transparent visual glass of the main frame, and the bottom of the XYZ three-axis moving platform is equipped with universal pulleys.

[0027] Preferably, the real-time data fusion period of the real-time data fusion and adaptive control unit is ≤1s, and the output valve control signal error of the LSTM-PID combined rain control algorithm module is ≤±0.2mm / h.

[0028] A method for using an intelligent rainfall simulation system for slope stability experiments includes the following steps:

[0029] (1) Install the rainfall simulation frame device, connect the water supply pipeline to the water supply device and the variable frequency constant pressure device, install the external monitoring device, and bury each sensor;

[0030] (2) The monitoring equipment is calibrated, the rainfall intensity and duration are preset, the PLC controller starts the entire system, and the multi-point sprinkler group starts spraying water according to the set data; the pore water pressure sensor monitors the changes in pore water pressure on the slope and reflects the rainfall infiltration in real time; the moisture content sensor monitors the changes in soil moisture content in real time and provides data on rainfall infiltration depth and distribution; the DIC full-field strain measurement system uses a high-speed camera and image analysis software to capture the surface strain and crack propagation of the slope; fiber optic sensors are laid on the surface and inside of the slope to analyze the deep deformation characteristics of the slope through strain data; acoustic emission sensors monitor the propagation of microcracks inside the soil and rock and provide early warning of slope instability risk; the three-dimensional laser scanner establishes a three-dimensional model of the slope surface and accurately records the slope deformation.

[0031] (3) The real-time data fusion and adaptive control unit can automatically control the operation of the rainfall system according to the preset rainfall intensity and duration, and automatically adjust the rainfall intensity and duration according to the data fed back by the comprehensive monitoring unit, respond to the changes in slope status in real time, and adjust the experimental plan.

[0032] (4) The data analysis module performs in-depth analysis and visualization of the experimental data, analyzes the slope instability mechanism, and draws conclusions and suggestions.

[0033] Preferably, the LSTM-PID combined rain control algorithm module includes an LSTM feedforward predictor and a PID feedback unit, which can perform look-ahead prediction based on the number of prediction steps and obtain valve control commands based on feedback data from each sensor for control.

[0034] The prediction steps are set to τ, and the LSTM feedforward predictor uses the target rainfall intensity r as the prediction step. t (k), measured rainfall intensity r m The input vector s(k) consists of the following parameters: water supply pressure p(k), raindrop size d(k), platform pose (x(k), y(k), z(k)), tracking error e(k), error increment Δe(k), and disturbance observations. The output is a look-ahead prediction. The error prediction is further defined as follows: This represents the error in predicting the (k+τ)th time step based on the current input at time k; perturbation estimation. This represents the equivalent disturbance at time k that accounts for external factors such as pressure fluctuations, temperature changes, and partial nozzle blockage in the rainfall intensity.

[0035]

[0036] LSTM feedforward predictor Taking s(k) as input, the output is the one-step look-ahead prediction, i.e., when τ = 1.

[0037]

[0038] in, It is a prediction of the error e at the next sampling time. It is an equivalent disturbance prediction for external factors such as water supply pressure fluctuations and nozzle blockage. σ(k) is the prediction uncertainty, which is used for subsequent weight scheduling.

[0039] Preferably, the real-time control and data fusion cycle of the real-time data fusion and adaptive control unit is ≤1s, and the rainfall intensity error is ≤±0.2mm / h; when the pressure disturbance amplitude does not exceed 10% of the rated value, the system recovers to the error band within 60s.

[0040] The nozzles of a multi-point nozzle assembly can be set with particle size settings, which can be divided into large, relatively large, medium, relatively small, and small.

[0041] The real-time data fusion and adaptive control unit has a control and data fusion real-time cycle of ≤1s and a rainfall intensity error of ≤±0.2mm / h. When the pressure disturbance amplitude does not exceed 10% of the rated value, the system recovers to the error band within 60s. It achieves rapid resetting under nozzle combination, particle size setting switching, and pressure disturbance.

[0042] Preferably, the real-time data fusion and adaptive control unit supports remote monitoring and cloud data sharing.

[0043] Preferably, the pore water pressure sensor has an accuracy of ±0.1 kPa, the water content sensor has an accuracy of ±0.2%, and the high-speed camera has a resolution greater than 1000 fps.

[0044] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0045] This invention realizes a closed-loop experimental paradigm of "rainfall-response-intelligent control," overcoming three major pain points of traditional methods: fragmented monitoring, control lag, and discrepancies between rainfall implementation methods and reality. It provides a reliable tool for studying the mechanisms of rainstorm-induced landslides. Specifically, it includes:

[0046] (1) A three-dimensional integrated intelligent system for rainfall monitoring and control was proposed to form a closed-loop control and realize intelligent control of the rainfall process. The rainfall intensity, raindrop size and rainfall uniformity are precisely adjustable.

[0047] (2) Achieve high-precision, multi-sensor linkage real-time monitoring and feedback control, real-time multi-dimensional monitoring, and realize high-precision, real-time analysis of experimental data and automatic risk warning;

[0048] (3) Advanced monitoring methods such as DIC full-field strain, optical fiber, and acoustic emission are adopted to significantly improve experimental accuracy. The system has high integration and strong automation, which significantly improves experimental efficiency and accuracy. It provides comprehensive and reliable data support and provides accurate basis for slope stability analysis and disaster prevention.

[0049] (4) The system has a high degree of integration and automation, which significantly improves the efficiency and accuracy of experiments. It is highly automated and can accurately control the intensity of rainfall and the size of raindrops. At the same time, the experimental process does not require frequent human intervention.

[0050] (5) This invention can be applied to the field of engineering for the safety assessment of slopes in major projects, especially in the slope design stage of railways, highways, hydropower projects, mine tailings and so on; it can also be used in the field of education and public safety, in teaching in colleges and research institutions, as an experimental platform for majors such as geological engineering and geotechnical engineering, and in emergency drills and popular science, it can show the landslide formation process to the public through visualized data (such as three-dimensional deformation animation) to enhance disaster prevention awareness. Attached Figure Description

[0051] Figure 1 This is a schematic diagram of the intelligent rainfall simulation unit of the present invention.

[0052] Figure 2 This is a schematic diagram of the external monitoring device of the present invention.

[0053] Figure 3 This is a functional diagram of the intelligent control unit of the present invention.

[0054] In the figure, 1-DIC full-field strain measurement system, 3D laser scanner and high-speed camera integrated component; 2-XYZ three-axis moving platform; 3-Y-axis moving platform; 4-X-axis moving platform; 5-Z-axis moving platform; 11-moving bracket; 12-water supply pipeline; 13-multi-point nozzle group; 14-connecting bracket. Detailed Implementation

[0055] The accompanying drawings are for illustrative purposes only; the invention is further described below with reference to embodiments, but should not be construed as limiting the scope of the invention to the following embodiments. Various substitutions and modifications made based on ordinary technical knowledge and common practice in the art without departing from the above-described technical concept of the invention should be included within the scope of protection of the invention.

[0056] Example 1

[0057] like Figure 1 As shown, an intelligent rainfall simulation system for slope stability experiments includes an intelligent rainfall simulation unit, a comprehensive monitoring unit, a real-time data fusion and adaptive control unit, and a data analysis and visualization unit. These units are organically integrated and work together.

[0058] The intelligent rainfall simulation unit includes a water supply device, a variable frequency constant pressure device, a rainfall simulation frame device, a slope simulation platform, and a rainfall location and intensity control device. The rainfall simulation frame device includes a rectangular main frame with a movable support 11, a connecting support 14, and multiple water supply pipes 12 at the top. The movable support 11 facilitates the movement of the rainfall simulation frame device, and the connecting support 14 connects the water supply pipes 12 to the movable support 11. The water supply pipes 12 can be configured with three horizontal and three vertical lines, and each water supply pipe is equipped with 2 to 4 sets of multi-point sprinkler heads 13.

[0059] The water supply pipeline 12 connects the water supply device and the variable frequency constant pressure device; at least two multi-point nozzle groups 13 are installed on each water supply pipeline 12. Each multi-point nozzle group 13 is a double nozzle with different nozzle particle size. Each nozzle has many spray holes. The rainfall intensity control range is 0.1 to 50 mm / h, and the raindrop particle size can be controlled within the range of 0 to 5 mm.

[0060] The integrated monitoring unit includes internal and external monitoring devices. The internal monitoring devices are embedded inside the slope similarity model and include pore water pressure sensors, water content sensors, fiber optic sensors, and acoustic emission sensors (not shown in the figure, embedded in the cuboid main frame). The external monitoring devices include a DIC full-field strain measurement system, a three-dimensional laser scanner, and a high-speed camera.

[0061] like Figure 2As shown, the external monitoring device is mounted on the transfer bracket 3, and a horizontal camera track 2 is mounted on the transfer bracket 3. The integrated component 1 of the DIC full-field strain measurement system, the three-dimensional laser scanner, and the high-speed camera is installed on the camera track 2. Adjustment handles are provided at both ends of the camera track 2 to adjust the track height.

[0062] The adjustment handle 4 is an automatic adjustment handle, and the adjustment method is automatic. A pulley is installed on the side of the automatic adjustment handle that contacts the transfer bracket 3. After the data captured by the DIC full-field strain measurement system, the 3D laser scanner, and the high-speed camera are transmitted to the intelligent control unit and the data processing module, the PLC controller can send a signal to move the automatic adjustment handle up or down according to the analyzed data and the needs of real-time acquisition. Then the adjustment handle 4 will be released, controlling the pulley to move down or up, so as to achieve automatic height adjustment.

[0063] The real-time data fusion and adaptive control unit includes a PLC controller, a real-time data fusion module, and an LSTM-PID combined rainfall adjustment algorithm module. It can automatically control the operation of the rainfall system based on preset rainfall intensity and duration, and automatically adjust the rainfall intensity based on sensor data feedback. The real-time data fusion cycle of the real-time data fusion and adaptive control unit is ≤1s, and the output valve control signal error of the LSTM-PID combined rainfall adjustment algorithm module is ≤±0.2mm / h.

[0064] The data analysis and visualization unit includes a data processing module and a visualization module. The data processing module synchronously collects, processes, and stores data, performs statistical analysis, strain field analysis, and permeability analysis on the collected data, and generates monitoring curves for displacement, pore water pressure, water content, and acoustic emission signals in real time. The visualization unit generates a real-time three-dimensional slope model and dynamically displays the slope deformation and crack development process.

[0065] At least one side of the main frame is transparent glass, the transfer bracket 3 is set on the outside of the transparent glass of the main frame, and the bottom of the transfer bracket 3 is equipped with universal pulleys 5.

[0066] A method for using an intelligent rainfall simulation system for slope stability experiments includes the following steps:

[0067] (1) Install the rainfall simulation frame device, connect the water supply pipeline to the water supply device and the variable frequency constant pressure device, install the external monitoring device, and bury each sensor;

[0068] (2) The monitoring equipment is calibrated, the rainfall intensity and duration are preset, the PLC controller starts the entire system, and the multi-point sprinkler group starts spraying water according to the set data; the pore water pressure sensor monitors the changes in pore water pressure on the slope and reflects the rainfall infiltration in real time; the moisture content sensor monitors the changes in soil moisture content in real time and provides data on rainfall infiltration depth and distribution; the DIC full-field strain measurement system uses a high-speed camera and image analysis software to capture the surface strain and crack propagation of the slope; fiber optic sensors are laid on the surface and inside of the slope to analyze the deep deformation characteristics of the slope through strain data; acoustic emission sensors monitor the propagation of microcracks inside the soil and rock and provide early warning of slope instability risk; the three-dimensional laser scanner establishes a three-dimensional model of the slope surface and accurately records the slope deformation.

[0069] (3) The real-time data fusion and adaptive control unit can automatically control the operation of the rainfall system according to the preset rainfall intensity and duration, and automatically adjust the rainfall intensity and duration according to the data fed back by the comprehensive monitoring unit, respond to the changes in slope status in real time, and adjust the experimental plan.

[0070] (4) The data analysis module performs in-depth analysis and visualization of the experimental data, analyzes the slope instability mechanism, and draws conclusions and suggestions.

[0071] Preferably, the LSTM-PID combined rain control algorithm module includes an LSTM feedforward predictor and a PID feedback unit, which can perform look-ahead prediction based on the number of prediction steps and obtain valve control commands based on feedback data from each sensor for control.

[0072] The real-time data fusion and adaptive control unit supports remote monitoring and cloud data sharing.

[0073] To describe the calculation process for controlling rainfall intensity, the operation of the LSTM-PID combined rainfall control algorithm module is explained in detail; all discrete variables are sampled over a period k. s (Unit: seconds) as the time base, using This represents the discrete-time index, where the physical time is t = kT. s .

[0074] (1) Symbolic parameters:

[0075] r t (k): Target rainfall intensity, in mm / h;

[0076] r m (k): Measured rainfall intensity, in mm / h;

[0077] r(k): Rainfall intensity state, take r(k) = r m (k), unit mm / h;

[0078] e(k): Tracking error, defined as e(k) = rt (k)-r m (k), unit mm / h;

[0079] Δe(k): Error increment, defined as Δe(k) = e(k) - e(k-1), unit mm / h;

[0080] p(k): Measured water supply pressure, in kPa;

[0081] d(k): The equivalent percentage of nozzle opening or particle size setting, in %;

[0082] x(k), y(k), z(k): Platform pose, in mm;

[0083] u(k): Valve control command, i.e., the opening degree sent to the electric valve, in %;

[0084] u min ,u max : Lower and upper limits of valve position, in %;

[0085] Δu max : The maximum permissible change in valve position within a single sampling period, in %;

[0086] τ: Number of prediction steps. This implementation uses only one lookahead step, therefore τ = 1;

[0087] Error prediction from time k to time k+1, in mm / h;

[0088] Predict the equivalent perturbation at time k to time k+1, in mm / h.

[0089] σ(k): Prediction uncertainty, used for scheduling weights, dimensionless;

[0090] Φ(d): Nozzle or particle size factor, a calibrated smooth function, unit (mm / h) /

[0091] C v (u): Valve characteristic function, unit (mm / h) C′ v (u) is its derivative with respect to u;

[0092] τ r The first-order equivalent time constant of the controlled object, in seconds;

[0093] w(k): Equivalent external disturbance, converted to rainfall intensity unit mm / h;

[0094] K ff Feedforward gain, dimensionless;

[0095] γ w Disturbance compensation weight, dimensionless;

[0096] K p ,K i ,K d The proportional, integral, and derivative gains of a PID feedback circuit are dimensionless.

[0097] ε dz Error dead zone threshold, unit: mm / h;

[0098] τ d Differential filter time constant, in seconds;

[0099] I(k): The integral internal state of the PID controller, in units of (mm / h)·s;

[0100] σ0: Weighted scheduling constant, dimensionless;

[0101] α(k), β(k): The fusion weights of feedforward and feedback, satisfying α(k) + β(k) = 1;

[0102] S r (u,p,d): Sensitivity of valve position to rainfall intensity, unit (mm / h) / %.

[0103] (2) Controlled object model:

[0104] The controlled object describes the rainfall intensity response of the sprinkler system under the influence of valve position, water supply pressure, and nozzle settings. The static mapping is as follows:

[0105]

[0106] The dynamic process is described using a first-order inertial model within the sampling period:

[0107]

[0108] Local sensitivity is defined as the partial derivative of the valve position with respect to the steady-state value of the rainfall intensity:

[0109]

[0110] These equations describe the static relationship and dynamic evolution of rainfall intensity as the valve position u changes, and form the basis for subsequent control law design.

[0111] (3) LSTM feedforward predictor:

[0112] In each sampling period, the input vector s(k) is composed of the current target value, the measured value, and other states, i.e.:

[0113]

[0114] LSTM feedforward predictor Given s(k) as input, output the one-step look-ahead prediction:

[0115]

[0116] in, It is a prediction of the error e at the next sampling time. It is an equivalent disturbance prediction for external factors such as water supply pressure fluctuations and nozzle blockage. σ(k) is the prediction uncertainty, which is used for subsequent weight scheduling.

[0117] (4) Nominal inversion:

[0118] Nominal inversion maps the target rainfall intensity to an ideal threshold. Because r ss There is generally no analytical inverse solution; here, we use a single Newton update based on the valve position u(k-1) of the previous cycle:

[0119]

[0120] This operation yields the nominal valve position u. eq (k) makes the rainfall intensity approach the target r when the disturbance is ignored. t (k).

[0121] (5) Feedforward correction:

[0122] LSTM predictions give the future error and disturbance Based on sensitivity S r Calculate the feedforward correction:

[0123]

[0124] The unlimited feedforward instruction is:

[0125]

[0126] This step converts prediction errors and disturbances into valve position compensation through sensitivity normalization, enabling the system to offset possible deviations in advance.

[0127] (6) PID feedback

[0128] To suppress model errors and unknown disturbances, a PID feedback structure with dead zone, integral limiting, and derivative filtering is adopted.

[0129] Error dead zone handling:

[0130] e f (k)=sgn(e(k))max(|e(k)|-ε dz ,0);

[0131] Points Channel:

[0132] I(k)=min(max(I(k-1)+T s e f (k),I min ),I max );

[0133] Differential channel:

[0134]

[0135] Feedback instructions (without amplitude limit):

[0136]

[0137] The feedback channel corrects the valve position in real time based on the current error, ensuring system stability and eliminating steady-state error.

[0138] (7) Weighted scheduling and limit:

[0139] The prediction uncertainty σ(k) is used to schedule the weights of the feedforward and feedback:

[0140] β(k) = 1 - α(k);

[0141] The fusion resulted in an unlimited amplitude command:

[0142]

[0143] Then, physical constraints are applied. First, the upper and lower limits of the valve position (position limiting) are set:

[0144]

[0145] The final valve position is obtained by applying a rate-of-change limit:

[0146]

[0147] The limiting operator ensures that the valve movement does not exceed the physical opening range, nor does it cause water hammer effect due to excessively rapid changes.

[0148] (8) Control Laws and Object Updates:

[0149] In summary, the valve control command u(k) is obtained through feedforward, feedback, and weighted scheduling via a limiting operator, and the rainfall intensity state is updated according to the object model:

[0150]

[0151] Operator This represents a combination of positional limit and rate of change limit operations.

[0152] (9) Control process diagram

[0153] The process of calculating valve control commands and controlling rainfall intensity using known data can be summarized in the following steps:

[0154] ① Data collection: Read the target rainfall intensity r at sampling time k. t (k), measured rainfall intensity r m (k), water supply pressure p(k), nozzle setting d(k), and platform pose (x(k), y(k), z(k)), calculate error e(k) and error increment Δe(k).

[0155] ② Error prediction: The above data are combined into a vector s(k), and then passed through an LSTM feedforward predictor. Obtain the next step error Disturbance And uncertainty σ(k).

[0156] ③ Nominal inversion: Based on the target rainfall intensity r t (k) and the current valve position u(k-1) are used to calculate the nominal valve position u. eq (k), and then use the prediction error and disturbance to calculate the feedforward correction Δu ff (k).

[0157] ④PID Feedback: Perform dead-zone, integral, and derivative processing on the current error e(k) to generate feedback instructions.

[0158] ⑤ Fusion and Limiting: Calculate weights α(k) and β(k) based on the uncertainty, weight the feedforward command and the feedback command, and obtain the actual valve control command u(k) after position and rate of change limiting.

[0159] ⑥ Object Update: Apply the valve position u(k) to the nozzle system and calculate the rainfall intensity r(k+1) at the next moment according to the object dynamic model.

[0160] ⑦ Execute in a loop: Let k ← k+1, repeat steps (1)–(6) to form a closed-loop control and realize real-time adjustment of rainfall intensity.

[0161] Through the above process, the system uses known data (target rainfall intensity, measured rainfall intensity, supply pressure, nozzle settings, platform position) and prediction information to calculate valve control commands in sequence according to a series of mathematical relationships, so that the rainfall intensity of the sprinkler system stably tracks the set value.

[0162] Example 2

[0163] like Figure 3 As shown, a method for using an intelligent rainfall simulation system for slope stability experiments includes the following steps:

[0164] (1) Experimental preparation

[0165] ① Model filling and sensor installation;

[0166] ② Calibrate the monitoring equipment and set the initial parameters.

[0167] (2) Rainfall simulation

[0168] ①Activate the intelligent rainfall unit to adjust rainfall intensity and particle size in real time;

[0169] ② The monitoring system is activated to collect data such as pore water pressure, water content, strain, and acoustic emission signals in real time.

[0170] (3) Feedback adjustment

[0171] ① The intelligent control unit automatically adjusts the rainfall process based on monitoring data; the PLC controller analyzes, processes, calculates, and stores the data from the data streams fed back by various sensors through the computer integrated system, and transmits the data from each node to the cloud server, thereby forming a local database and a cloud sharing platform. On the other hand, it controls the rainfall, controls the location and amount of rainfall, and achieves linkage with the comprehensive monitoring unit. It can also be remotely controlled by setting up a remote control terminal.

[0172] ② When abnormal signals occur, the rainfall strategy will be automatically adjusted or rainfall will be suspended and a risk warning will be issued.

[0173] (4) Data Analysis

[0174] ① The data analysis unit performs in-depth analysis and visualization of the experimental data;

[0175] ②Analyze the slope instability mechanism and propose risk warning and prevention suggestions.

[0176] Comparative Examples 1 and 2 are shown in Table 1.

[0177] The system of Embodiment 1 of the present invention was compared with similar domestic systems in terms of functionality, and the results are shown in Table 1.

[0178] Table 1. Functional comparison between the system of Embodiment 1 of the present invention and the devices of Comparative Examples 1 and 2.

[0179]

[0180]

[0181] Of course, the above description is not intended to limit the present invention, and the present invention is not limited to the examples given above. Any changes, modifications, additions or substitutions made by those skilled in the art within the scope of the present invention should also fall within the protection scope of the present invention.

Claims

1. An intelligent rainfall simulation system for slope stability experiments, characterized in that, It includes an intelligent rainfall simulation unit, a comprehensive monitoring unit, a real-time data fusion and adaptive control unit, and a data analysis and visualization unit. These units are organically integrated and work together. The intelligent rainfall simulation unit includes a water supply device, a variable frequency constant pressure device, a rainfall simulation frame device, a slope simulation platform, and a rainfall location and intensity control device. The integrated monitoring unit includes internal and external monitoring devices. The internal monitoring devices are embedded inside the slope similarity model and include pore water pressure sensors, water content sensors, fiber optic sensors, and acoustic emission sensors. The external monitoring devices include a DIC full-field strain measurement system, a three-dimensional laser scanner, and a high-speed camera. The real-time data fusion and adaptive control unit includes a PLC controller, a real-time data fusion module, and an LSTM-PID combined rainfall adjustment algorithm module. It can automatically control the operation of the rainfall system according to the preset rainfall intensity and duration, and automatically adjust the rainfall intensity and duration according to the data fed back by the comprehensive monitoring unit. The data analysis and visualization unit includes a data processing module and a visualization module. The data processing module synchronously collects, processes, and stores data, performs statistical analysis, strain field analysis, and permeability analysis on the collected data, and generates monitoring curves for displacement, pore water pressure, water content, and acoustic emission signals in real time. The visualization unit generates a real-time three-dimensional slope model and dynamically displays the slope deformation and crack development process.

2. The intelligent rainfall simulation system for slope stability experiments according to claim 1, characterized in that, The rainfall simulation frame device includes a rectangular main frame, with a movable support, a connecting support, and a water supply pipeline at the top of the main frame. There is at least one water supply pipeline, and multiple nozzle groups are installed on each water supply pipeline.

3. The intelligent rainfall simulation system for slope stability experiments according to claim 2, characterized in that, The water supply pipeline connects the water supply device and the variable frequency constant pressure device; each water supply pipeline is equipped with at least two multi-point nozzle groups, each multi-point nozzle group is a quick-replaceable double nozzle group, the raindrop kinetic energy is programmable in the range of 0.02 to 0.5J, the rainfall intensity is controlled in the range of 0.1 to 50 mm / h, and the raindrop particle size is in the range of 0 to 5 mm.

4. The intelligent rainfall simulation system for slope stability experiments according to claim 2, characterized in that, The external monitoring device is mounted on the transfer bracket, namely the XYZ three-axis moving platform, which includes a Y-axis moving platform, an X-axis moving platform, and a Z-axis moving platform. The integrated components of the DIC full-field strain measurement system, the three-dimensional laser scanner, and the high-speed camera are mounted on the X-axis moving platform. Adjustment handles are provided at both ends of the X-axis moving platform. At least one side of the main frame is transparent visual glass, and the transfer bracket is set on the outside of the transparent visual glass of the main frame. Universal casters are provided at the bottom of the XYZ three-axis moving platform.

5. The intelligent rainfall simulation system for slope stability experiments according to claim 1, characterized in that, The real-time data fusion cycle of the real-time data fusion and adaptive control unit is ≤1s, and the output valve control signal error of the LSTM-PID combined rain control algorithm module is ≤±0.2mm / h.

6. The method of using the intelligent rainfall simulation system for slope stability experiments as described in any one of claims 1-5, characterized in that, Includes the following steps: (1) Install the rainfall simulation frame device, connect the water supply pipeline to the water supply device and the variable frequency constant pressure device, install the external monitoring device, and bury each sensor; (2) The monitoring equipment is calibrated, the rainfall intensity and duration are preset, the PLC controller starts the entire system, and the multi-point sprinkler group starts spraying water according to the set data; the pore water pressure sensor monitors the changes in pore water pressure on the slope and reflects the rainfall infiltration in real time; the moisture content sensor monitors the changes in soil moisture content in real time and provides data on rainfall infiltration depth and distribution; the DIC full-field strain measurement system uses a high-speed camera and image analysis software to capture the surface strain and crack propagation of the slope; fiber optic sensors are laid on the surface and inside of the slope to analyze the deep deformation characteristics of the slope through strain data; acoustic emission sensors monitor the propagation of microcracks inside the soil and rock and provide early warning of slope instability risk; the three-dimensional laser scanner establishes a three-dimensional model of the slope surface and accurately records the slope deformation. (3) The real-time data fusion and adaptive control unit can automatically control the operation of the rainfall system according to the preset rainfall intensity and duration, and automatically adjust the rainfall intensity and duration according to the data fed back by the comprehensive monitoring unit, respond to the changes in slope status in real time, and adjust the experimental plan. (4) The data analysis module performs in-depth analysis and visualization of the experimental data, analyzes the slope instability mechanism, and draws conclusions and suggestions.

7. The method of using the intelligent rainfall simulation system for slope stability experiments according to claim 6, characterized in that, The LSTM-PID combined rain control algorithm module includes an LSTM feedforward predictor and a PID feedback unit. It can perform look-ahead prediction based on the number of prediction steps and obtain valve control commands based on feedback data from various sensors for control.

8. The method of using the intelligent rainfall simulation system for slope stability experiments according to claim 7, characterized in that, The real-time control and data fusion of the real-time data fusion and adaptive control unit has a control and data fusion real-time cycle of ≤1s and a rainfall intensity error of ≤±0.2mm / h. When the pressure disturbance amplitude does not exceed 10% of the rated value, the system recovers to the error band within 60s.

9. The method of using the intelligent rainfall simulation system for slope stability experiments according to claim 6, characterized in that, The real-time data fusion and adaptive control unit supports remote monitoring and cloud data sharing.

10. The method of using the intelligent rainfall simulation system for slope stability experiments according to claim 6, characterized in that, The pore water pressure sensor has an accuracy of ±0.1 kPa, the moisture content sensor has an accuracy of ±0.2%, and the high-speed camera has a resolution of over 1000 fps.

Images