Intelligent avoidance method and system for water lock galloping vibration

Abstract

The application discloses a kind of water gate flow-induced vibration intelligent avoidance method and system, according to numerical simulation combining sensitivity analysis, measuring point and controller optimization arrangement algorithm, realize wind, rain, water and temperature, attitude and vibration sensor measuring point optimization and damper arrangement optimization, to realize water gate vibration influencing factor wind rain and water flow, and water gate state and vibration effect perception and water gate vibration inducement and response such as water gate position, vibration, deformation, attitude, temperature, again after the part of simplified fault tree is observed Markov decision process to establish light-weight intelligent avoidance algorithm, by the collaborative work of control shock absorbing damping device to realize water gate vibration intelligent avoidance.

Description

Technical Field

[0001] This invention belongs to the field of water conservancy, hydropower and waterway engineering technology, and relates to a method and system for intelligent avoidance of flow-induced vibration of sluice gates. Background Technology

[0002] In water conservancy, hydropower, water transport, and water resource optimization projects, sluice gates serve as crucial facilities for regulating upstream and downstream water levels, intercepting floods, controlling flow, discharging sediment, and facilitating vessel transport. These sluice gates play vital roles in water storage, diversion, flood control, and tide prevention. Beyond their wide range of functions, sluice gates are characterized by high efficiency, stringent safety requirements, and large numbers. Taking sluice gates and reservoirs as examples, according to the national water conservancy census, as of 2021, my country had sluice gates with a flow rate ≥5 m³ / h. 3 There are 103,575 sluice gates per second, and nearly 100,000 reservoir dams. Most reservoirs are equipped with flood discharge gates or power generation water diversion gates. In short, sluice gates play an irreplaceable role in flood control, optimal allocation of water resources, navigation safety, food security, ecological security, and social and economic development.

[0003] External excitations such as wind, rain, water flow, waves, leaks, pump station valve unit induction, gate opening and closing, impacts, and earthquakes can easily cause vibrations and even resonance. Resonance includes forced resonance caused by the external excitation frequency matching the structure's natural frequency, as well as parametric resonance, internal resonance, and combined resonance caused by the structure's dynamic characteristics. When the vibration amplitude, frequency, or phase, and their combinations, exceed a certain range, they will adversely affect the safety, durability, or controllability of the sluice gate, even causing damage, thus forming harmful vibrations. Harmful vibrations of sluice gates not only accelerate equipment damage and aging but also easily induce resonance, even endangering the normal operation of the gate, structural safety, and even the safety of people's lives and property. Especially during the flood season, if harmful vibrations prevent the sluice gate from discharging floodwater normally, it will have catastrophic consequences. Therefore, avoiding harmful vibrations of sluice gates and designing corresponding monitoring systems to regulate the vibration state of sluice gates is of great scientific and engineering significance. Of the aforementioned induced vibrations, flow-induced vibration is the most common. Although there is considerable research on flow excitation, most of these studies only analyze the causes of vibration and provide some empirical practices and suggestions for overcoming or reducing it; or they provide some monitoring system configurations or temporary dynamic testing. In summary, there is currently no complete, practical, and advanced system solution for monitoring and avoiding flow-induced vibration.

[0004] To achieve intelligent, systematic, and standardized vibration avoidance for sluice gates, this invention establishes a thermo-fluid-structure interaction dynamic mathematical model based on the sluice gate's operating environment and vibration-inducing factors. It uses probability estimation and importance sampling combined with Monte Carlo random finite element method to obtain the sluice gate's input-output sensitivity requirements, establishing a sensing and vibration avoidance system. Based on this, it optimizes the layout and pre-training samples of sensing, dynamic response, and vibration control measurement points and control devices for the sluice gate. A lightweight simplified fault tree partially observable Markov decision model (POMDP) / reinforcement learning (RL) is applied to achieve vibration avoidance during sluice gate operation. Summary of the Invention

[0005] To address the shortcomings and deficiencies of existing technologies, this invention discloses an intelligent method and system for avoiding flow-induced vibrations in sluice gates. By optimizing the sensing and control system based on numerical simulation and employing a partially observable Markov decision / reinforcement learning algorithm with a simplified fault tree, it achieves intelligent real-time perception and avoidance of harmful vibration states throughout the entire process of the sluice gate. Vibration perception involves sensing the wind, rain, and water flow loads acting on the sluice gate, as well as the sluice gate's operating status and vibration response, through corresponding sensors. Avoidance is achieved through automatic identification using reward and penalty functions in the partially observable Markov decision model / reinforcement learning, and optimized control of actuators such as gate opening and closing forces and damping facilities.

[0006] To achieve the above objectives, the technical solution of the present invention is as follows:

[0007] This invention, based on numerical simulation combined with sensitivity analysis, measurement point and controller optimization algorithms, optimizes the measurement points of wind, rain, water, temperature, attitude and vibration sensors and the arrangement of dampers. This enables the perception of factors affecting sluice gate vibration, such as wind, rain and water flow, as well as the sluice gate state and vibration effects, and the perception of sluice gate vibration causes and responses, such as sluice gate position, vibration, deformation, attitude, and temperature. Then, through a simplified fault tree partially observable Markov decision process, a lightweight intelligent avoidance algorithm is established to achieve intelligent avoidance of sluice gate vibration by controlling the coordinated work of vibration damping devices.

[0008] The beneficial effects of this invention are as follows:

[0009] This invention establishes a thermo-fluid-structure interaction dynamic mathematical model based on the operating environment and vibration-inducing factors of a sluice gate. It uses probability estimation and importance sampling combined with Monte Carlo random finite element method to obtain the input and output sensitive elements of the sluice gate and the optimized arrangement of vibration damping. Based on this, a partially observable Markov decision / reinforcement learning algorithm with a simplified fault tree is employed to coordinate the actions of the sluice gate's sensing, dynamic response, and vibration control devices, thereby achieving vibration avoidance. The initial probability of the partially observable Markov decision / reinforcement learning algorithm with a simplified fault tree is obtained based on experience and conditional probability estimation using a Copula function, resulting in a POMDP pre-trained model. The pre-trained model and response hardware and software are deployed on a real sluice gate, and after passing typical cycle testing, they are put into formal operation, realizing the intelligent, systematic, and standardized avoidance of hazardous vibrations in sluice gates. Attached Figure Description

[0010] Figure 1 This is a schematic diagram of the overall process of the intelligent avoidance method for flow-induced vibration of a sluice gate provided by the present invention.

[0011] Figure 2 This is a schematic diagram of the numerical simulation region.

[0012] Figure 3 This is a schematic diagram of the overall algorithm flow of the Markov decision model.

[0013] Figure 4 This is a flowchart of subroutine function algorithm 2.

[0014] Figure 5 This is a flowchart of subroutine function algorithm 3. Detailed Implementation

[0015] The technical solutions provided by the present invention will be described in detail below with reference to specific embodiments. It should be understood that the following specific embodiments are only used to illustrate the present invention and are not intended to limit the scope of the present invention.

[0016] This invention provides a method for intelligently avoiding flow-induced vibration in sluice gates, the process of which is as follows: Figure 1 As shown, it includes the following steps:

[0017] Step 1: Collect data on the location of the sluice gate (for analyzing local hydrogeological, meteorological and climatic conditions), structural materials (including the location data of the gate chamber, bottom plate, gate piers, retaining walls, gate metal structure, electromechanical equipment, etc.), wind and rain loads, gate opening (including combination and opening height), flow velocity and flow rate, upstream and downstream head and other operating environment data, controllable parameters of each opening and closing electromechanical equipment, as well as calculations, static and dynamic response results and failure modes for this project and similar projects.

[0018] Step 2: Establish a dynamic analysis model of the thermo-fluid-structure interaction sluice gate based on the principles of mass conservation, momentum conservation, and energy conservation. Analyze the causes of sluice gate vibration based on its structural characteristics, operating environment, and scheduling (schedules refer to gate control, such as adjusting opening size), including historical vibration data. Identify the internal and external factors causing harmful vibrations and their response characteristics. Internal factors include dynamic parameters such as sluice gate structure, materials, dynamic modulus of elasticity, dynamic stiffness, damping, and control force distribution. External factors include wind and rain vibration, water flow impact, wave impact, and gap leakage. Collect statistical data on external influencing factors and establish corresponding probabilistic models for each external factor using methods such as probability fitting and kernel estimation. Based on this, determine the (minimum and maximum) variation ranges of each external factor at corresponding time quantiles according to the sluice gate's design service life (in years), which will serve as the basis for the next step of random finite element sampling. The interval estimation of various external influencing factors can be performed using minimax model estimation and generalized extreme value distribution functions (including Frechet, Weibull, and Gumbel distributions). The model parameters can be estimated using the method of moments, maximum likelihood estimation, and probability-weighted moment estimation, with maximum likelihood estimation based on observation sequences combined with optimization algorithms being preferred. The thermo-fluid-structure interaction equations under sluice gate-fluid excitation are established based on the mechanical model of the sluice gate vibration. (Specifically, this involves the simultaneous equations of the following models.)

[0019] In this invention, a sluice gate is used as the load-bearing object, and its loads include wind and rain excitation, static and dynamic water and wave current forces, temperature load, gravity, inertial force, damping force, and supporting force. During rainfall, the interaction force mode between the gate surface and the waterline is re-examined using wetting theory, Prandtl boundary layer theory, and hydrostatic boundary layer theory, and calculation formulas for Coulomb damping force and viscous linear damping force are adopted.

[0020] For ease of explanation, corresponding mathematical models are established based on the characteristics of each region in the model, and are explained as follows:

[0021] (1) Thematic structure sluice gate model (region Ω0)

[0022] The sluice gate, as a composite structure composed of reinforced concrete and metal gates, has a dynamic model that includes dynamic conservation methods and energy conservation equations. The equilibrium equations for dynamic analysis are as follows: In the formula, [M] is the mass matrix of reinforced concrete and metal structure; [M]{ü} reflects the inertial effect, and {ü} is the second derivative matrix of nodal displacement with respect to time; [C] is the damping matrix, including the damping of the structure itself, water body, support and vibration reduction device; This reflects the damping effect. t represents time; F[t] represents the load array, which includes external forces acting on the object, such as gravity, pressure, friction, and support forces.

[0023] (2) Wind-Rain Two-Phase Flow Model (Region Ω1)

[0024] Considering the effects of wind and rain coupling on the sluice gate, a two-phase flow model is adopted. The two-phase flow model includes mass conservation, momentum conservation, and energy conservation equations, as follows:

[0025] mass conservation equation: Momentum conservation equation: Energy conservation equation:

[0026]

[0027] In the formula: k represents the phase code, where k = a represents the gas phase and k = l represents the liquid phase. ρ k Represents the k-phase density, u k Represents the k-phase velocity, g represents the gravitational acceleration, and σ k Let ω be the shear stress tensor. k For water and air phase forces, p is the pressure scalar of each phase, and e is the pressure of each phase. k is the specific thermodynamic energy, K is the thermal conductivity, T is the temperature, and Q is the internal heat source.

[0028] The above equations describe the complete two-phase flow of water and air based on the Euler model. By solving the mathematical model, we can obtain various physical and mechanical parameters in the two-phase fluid flow process, such as velocity, pressure, and temperature.

[0029] (3) The wind-wave-current coupled model (region Ω2) is used below the water surface.

[0030] Wave motion dynamic equation: Continuity equation:

[0031] In the formula: x, y, z are spatial coordinates in the Cartesian coordinate system; t is time; u, v, w are the velocity components of the fluid in the x, y, z directions; ρ is the density of the actual water body; and g is the acceleration due to gravity.

[0032] Wave motion is a form of fluid motion, therefore it must satisfy the fluid motion equations, and the water flow is assumed to be inviscid and incompressible, and in a gravitational field it also satisfies the continuity equation.

[0033] (4) Mathematical model for calculating the flow rate of a sluice gate with arbitrary gate opening under the constraints of regulations

[0034] Formula for calculating the passage volume in the free outflow state of the gate: Garbrecht formula: Duyu formula: In the formula, μ0 is the orifice flow coefficient, which can be calculated by Garbrecht's empirical formula or Du Yu's empirical formula; b is the gate width; e is the gate opening; H0 is the water level in front of the gate; This refers to the relative opening.

[0035] To obtain a definite solution, boundary conditions and initial conditions are defined at the solid-liquid-gas interface, and constraints are applied respectively. The boundary conditions are set according to the fluid-structure interaction theory, and the initial conditions are the calculation results under the steady state.

[0036] Step 3: Collect measured response data of the sluice gate. After collection, perform gross error identification to eliminate erroneous data. The identification of erroneous data is based on the instrument measurement principle and the physical meaning of the measured values, and the determination of the measured data is made according to the correlation and rationality of the data.

[0037] Step 4: For sluice gates with measured response data, the measured data is used to invert, assimilate, or correct the mechanical and thermodynamic parameters of the sluice gate, thereby ensuring the accuracy of the model and calculation parameters. For sluice gates without measured data, on-site ultrasonic testing or similar engineering experience are used to obtain calculation parameters.

[0038] Step 5: Vibration sensing and damping system optimization. Based on the above mathematical model and engineering experience, the sensing points and the layout of damping equipment and facilities are optimized.

[0039] Step 5.1: Measuring points for upstream and downstream water levels, air temperature, wind speed and direction, and rainfall intensity are arranged according to relevant specifications. Flow patterns, attitudes, and locations are arranged based on representativeness, maximum amplitude, and the principle of greatest sensitivity to vibration. The sensor arrangement is optimized using the three-dimensional effective independence method, the effective independence-driving point residual method, and the QR-MAC hybrid algorithm. The corresponding sensors include upstream and downstream radar water level gauges, anemometers, raindrop spectrometers, rain direction and intensity meters, cross-sectional flow velocity meters, surface flow video cameras, a three-dimensional vibration response instrument for the sluice gate (including acceleration, attitude, and dynamic displacement sensors), opening and closing force sensors, gate opening sensors, and free-field seismic response sensors.

[0040] Step 5.2 involves optimizing the layout of the vibration damping equipment. The vibration damping equipment includes mass-tuned dampers, frequency-tuned liquid dampers, displacement dampers, velocity dampers, electromagnetic inertial mass dampers, and AMD active controllers. The optimized layout is achieved using a method based on genetic algorithms or continuous search.

[0041] Step 6: Use the finite element method to calculate the mathematical model established in Step 2 to obtain the initial parameters or pre-training samples for the next step of the POMDP model. To reduce the computational load, not all variables are randomized. Instead, Monte Carlo random sampling is used for some physical quantities in the calculation boundary conditions that have high uncertainty, large variation, or are sensitive to vibration.

[0042] Step 6.1, Sampling of key variables

[0043] (1) Based on different combinations of upstream and downstream water level ranges, wind speed ranges, rainfall intensity ranges, gate opening ranges, and the number of gates, the dynamic response time history of the sluice gate under different operating conditions is obtained according to the uniform design combination time history. When the sluice gate amplitude increases significantly, the attitude changes abruptly, or the vibration frequency changes significantly, it is determined to be a resonance condition. Based on the sensitivity analysis results, dynamic optimization methods such as dynamic condensation are used to determine the placement positions of the accelerometer, attitude sensor, and position sensor.

[0044] (2) The dynamic and static response, gap flow and discharge flow of the sluice gate under different gate openings and upstream and downstream water level combinations were calculated using a three-dimensional river hydrodynamic wave-current coupling model of the coupled wind and rain effects of the upstream and downstream of the sluice gate. The vibration reduction effect of the sluice gate under different combinations and its observable measurement were obtained by calculating the optimized vibration reduction measures.

[0045] Step 6.2: The dynamic response of the sluice gate is numerically calculated using the domain decomposition method. For metal gates, the substructure method is used for numerical calculation. The typical regions and key processing methods are as follows:

[0046] (1) The SSTk-ε turbulence model was applied to the wind and rain area above the sluice gate to numerically simulate the three-component force coefficients of the sluice gate section under different angles of attack. The pressure and velocity distribution maps of the flow field around the section were given and analyzed. The two-phase flow theory of the Eulerian-Eulerian system was adopted, and the raindrop field was taken as the continuous medium field. For constant non-severe storms, the Reynolds average method was used to obtain the rain phase streamline diagrams under different rain phase particle sizes and different wind speeds, as well as the raindrop capture rate and impact load on the model surface. For storms, the large eddy simulation method was used to obtain the load time history data and force spectrum in the downwind, crosswind and torsional directions. Among them, the wind field adopted the Kaimal spectrum and the Lumley-Panofsky spectrum; the raindrop spectrum reproduction methods were the Rosin-Rammler preset function method and the CFD custom function method. A numerical simulation method based on two parameters of wind and rain coupling field was established by rain intensity conversion and raindrop spectrum.

[0047] (2) For the upstream and downstream water bodies, waves, and wind and rain mixed fields of the sluice gate, the Eulerian-Eulerian two-phase flow theory was adopted, and the raindrop field was used as a continuous medium field to conduct three-dimensional numerical simulation of the wind and rain motion around the sluice gate. The Reynolds average method was used to obtain the rain phase streamline diagrams under different rain phase particle sizes and wind speeds, as well as the raindrop capture rate and impact load on the model surface. The large eddy simulation method was used to obtain the load time history data and force spectrum in the downwind, crosswind, and torsional directions. The numerical simulation of the gate water motion was carried out by combining the ALE (Arbitrary Lagrange-Eulerian method) interface tracking technology with the linear mass source wave generation method; at the same time, the water-air interface tracking was established based on the VOF (Volume Fraction Method) interface capture technology. A wind-wave combined field numerical model was established based on the VOF method, and the applicability of the numerical method was verified by comparing and analyzing the experimental and numerical simulation results. There are two solutions for the flow calculation in the near-wall region and the flow calculation at low Re numbers: one is to use the wall function method, and the other is to use the k-ε model at low Re numbers.

[0048] (3) For opening and closing ropes or flexible outdoor structures, based on the EWF model and the Euler-Lagrange discrete phase model, and using the enhanced wall method combined with dynamic mesh technology, a numerical simulation method for the combined action of wind and rain on the structure was established to analyze the dynamic characteristics of the sluice gate under the action of wind and rain.

[0049] (4) In the process of establishing the overall grid for sluice gates and wind and rain loads, the method of "rigid moving area + moving grid area + static grid area" is used to first divide it into blocks, such as... Figure 2 As shown, the Newmark-β method code was then written into a user-defined function (UDF) in FLUENT to solve the vibration response of the structure. A numerical simulation method for structural vortex vibration was established by combining the dynamic mesh technology in FLUENT software.

[0050] (5) The Lagrangian and Eulerian variables are transformed through an approximate smooth δ-function. That is, the solid force density at the nodes of the finite element is distributed to the Gaussian integration points of the element through the element shape function, and then the solid force density at the Gaussian integration points in the support domain of the approximate smooth δ-function is transformed to the fluid element. Similarly, the fluid velocity on the Eulerian mesh is first transformed to the Gaussian integration points of the solid element through the δ-function, and then the velocity of the solid node is obtained through the element shape function, thereby obtaining the displacement of the solid.

[0051] Step 6.3, Overall partition calculation and boundary layer processing.

[0052] The static and dynamic response of the sluice gate under the combined effects of wind, rain, water flow, waves, and heat was analyzed using a domain decomposition method combined with a substructure approach. The selected upstream and downstream calculation ranges of the river channel were stable for the sluice gate vibration calculation results. Static boundaries were defined by flow rate, water level, and pressure, while dynamic boundary conditions used non-reflective boundaries. A specific calculation region was selected upstream and downstream of the sluice gate, encompassing a certain river length, height, and depth. The upper elevation represents the height where the sluice gate's influence from wind and rain is negligible. The depth represents the depth where the stress and deformation of the sluice gate, including the pile foundation, under various loads are negligible. The upstream and downstream regions represent lengths where the influence of the sluice gate's flow regime and flow field is negligible. The top of the flow field was assessed using symmetrical boundary conditions, equivalent to a freely sliding wall. The sluice gate cross-section and ground surface were assessed using non-slip wall conditions. A standard k-ε model or an improved model was selected, and the calculations were performed using the finite volume method coupled with the finite element method.

[0053] Step 6.4: Numerical calculation of hazardous vibration identification. Identification methods include amplitude identification, phase identification, displacement identification, velocity identification, and acceleration identification.

[0054] Step 7: Establish the POMDP model, defining the vector M = (S, A, O, T, Z, R, B), where S, A, and O represent the state vector, action vector, and implementation vector, respectively. S corresponds to the state set of the sluice gate system, including upstream and downstream water levels, wind and rain, opening degree, flow velocity, flow rate, and gap width for each gate. A corresponds to the gate opening and closing and vibration damping action set (opening and closing force, support force, sealing force, damping, etc.), i.e., variables that exert "force" or "action" on the sluice gate vibration and can be conditionally controlled by the control system. O represents the observed physical quantities acquired by monitoring instruments, such as the set of vibration acceleration vectors. T represents the probability distribution of transitions between states; Z represents the observation probability distribution; R represents the reward obtained; and B represents the initial state distribution within the state set S.

[0055] Based on step 6, after obtaining a sufficient number of samples through multiple calculations, the following probability distributions are calculated: T(s,a,s')=Pr(s'|s,a) represents the state transition probability, i.e., the probability distribution of transitioning to another state s' after performing action a in state s; Z(s,a,o)=Pr(o|s,a) represents the observation probability, i.e., the probability distribution of obtaining the observation value o after performing action a in state s; R(s,a) represents the reward obtained by performing action a in state s; B represents the initial state distribution, i.e., the distribution of the agent in the state set S at the initial moment. (Used to establish a Markov decision model)

[0056] Establish penalty and reward functions; vibrations with any of the following characteristics are considered harmful and will be subject to corresponding penalties, with the penalty magnitudes as follows. When both phenomena occur simultaneously, the penalty is set to the more severe value, and the reward to the smaller value. The results are obtained after setting a threshold and deployment. The reward scenarios and corresponding values ​​are as follows:

[0057] (1) No vibrations that could exceed the sensor sensitivity: +5;

[0058] (2) The amplitude exceeds the sensor sensitivity but is within the allowable range and is getting smaller and smaller: +4;

[0059] (3) If the value exceeds the sensor sensitivity but is within the allowable range and does not increase steadily: +2;

[0060] (4) When the vibration amplitude exceeds the sensor sensitivity and is greater than a certain value: -5;

[0061] (5) If the sensor sensitivity is exceeded and there is impact and wear between structures: -10;

[0062] (6) When resonance occurs that exceeds the sensor sensitivity: -10 (using the B-R dynamic stability criterion);

[0063] (7) When the amplitude exceeds the sensor sensitivity and resonance occurs and the amplitude increases: -15.

[0064] Step 8: Pre-training and field training of a partially observable Markov decision model.

[0065] Pre-training uses the computational samples from step 6, while practical applications use actual training and measurement samples from sensors and actuators deployed on-site. The specific training process is as follows: Figure 3 As shown, its inputs include:

[0066] β: Initial belief, the initial value of B, obtained through random finite element sample estimation or pre-training;

[0067] ε0: The target distance between μ(b0) and l(b0), obtained through the following iteration;

[0068] ξ: Target distance reduction rate, determined based on experience or trial calculation;

[0069] K: The number of sampling scenarios, determined according to different sluice gate scheduling situations;

[0070] D: Maximum depth of DESPOT, determined based on convergence;

[0071] λ: Regularization constant, determined empirically or through trial and error;

[0072] T max The maximum online planning time for each step is determined based on experience or trial calculations.

[0073] The specific iteration steps are as follows:

[0074] (1) Initialize a belief b with β;

[0075] (2) Execute (3);

[0076] (3) Execute sub-function 1: BUILDDESPORT(b) to create a new tree l;

[0077] (4) Calculate the action branch a that maximizes the value of l(b,a). * =max a∈A l(b,a);

[0078] (5) Determine whether the condition L0(b) > l(b, a) is satisfied. * If the condition is met, execute (6); otherwise execute (7).

[0079] (6) Update action branch a * For π0(b);

[0080] (7) Execute a * ;

[0081] (8) Accept the observed branch z;

[0082] (9) Add a child node b = Т(b, a) * ,z).

[0083] The descriptions of each sub-function are as follows:

[0084] Subfunction 1: BUILDDESPORT(b0) creates a new tree, including the following steps:

[0085] ① A set of K scenarios randomly sampled from the current belief b0. ;

[0086] ② Create a new DESPOTD, with node b as the root node;

[0087] ③ Initialize the upper and lower bounds U(b0), L0(b0), μ(b0), l(b0);

[0088] ④ Calculate the distance between nodes

[0089] ⑤ If the node distance The target distance is ε0 and the total running time is less than T. max If yes, then execute step 6; otherwise, execute step 9.

[0090] ⑥ Execute sub-function 2: EXPLORE(D,b0) to perform heuristic exploration of node b;

[0091] ⑦ Perform upper and lower bound backups at each node along the path, backing up the upper and lower bounds and distances of each node x on the path from b to D, μ(x), l(x), U(x);

[0092] ⑧ Calculate the distance between the upper and lower bounds ε(b0) = μ(b0) - l(b0), and then execute ⑤;

[0093] ⑨ Return the created tree l.

[0094] Sub-function 2: EXPLORE(D,b) process is as follows Figure 3 As shown, the function is to perform heuristic exploration, expand D, and shrink the upper and lower boundaries. The process is as follows:

[0095] ① If the tree height Δ(b) ≤ D and the excess uncertainty E(b) > 0, and the subfunction 3PRUNE(D,b) returns FALSE (cannot be pruned), then execute ② below; otherwise execute ⑦ below.

[0096] ② Determine if b is a leaf node of D. If yes, execute ③ below; otherwise, execute ④ below.

[0097] ③ Insert node b′ between node b and its parent node, making it the new parent node and initialize the upper and lower bounds of b′;

[0098] ④ Determine the optimal action branch a * =argmax a∈A μ(b,a);

[0099] ⑤ Observe the optimal branch

[0100] ⑥ Determine the new node b = τ(b, a) * ,z * Then execute 1;

[0101] ⑦ If the height of the tree If yes, then execute step ⑧ below; otherwise, execute step ⑨ below.

[0102] ⑧ Reset the boundaries to simplify the search: upper bound U(b) = L0(b0), lower bound u(b) = l0(b0), distance l(b) = l0(b0);

[0103] ⑨ Return to node b.

[0104] Subfunction 3: PRUNE(D,b) process is as follows Figure 4 As shown, its function is to prune from b back to the root. As the boundary is updated, new nodes may meet the pruning conditions and be pruned. The steps of subfunction 3 are as follows:

[0105] ① Initially, BLOCKED = FALSE;

[0106] ② Traverse every node x from node b to root node D. If the traversal is not finished, execute ③ below; if the traversal is finished, execute ⑨ below.

[0107] ③ Determine whether node x is blocked by any ancestor node. If it is blocked, execute ④ below; otherwise, execute ⑨ below.

[0108] ④ Reset the boundaries to simplify the search. The upper bound is U(b) = L0(b0), the lower bound is u(b) = l0(b0), and the distance is l(b) = l0(b0).

[0109] ⑤ Perform upper and lower bound backups at each node along the path, backing up the upper and lower bounds and distances of each node x on the path from b to D, μ(x), l(x), U(x);

[0110] ⑥ Set BLOCKED to TRUE;

[0111] ⑦ Execute 2;

[0112] ⑧ Execute 9;

[0113] 9. Return to BLOCKED.

[0114] Step 9, Training and Convergence Stability Analysis

[0115] The convergence condition of the partially observable Markov model training is used for judgment. When the convergence condition is met, that is, when the results of adjacent iterations are stable and the difference is within the preset range, the training can be stopped.

[0116] Step 10, Qualification Assessment

[0117] The effectiveness, robustness, and generalization ability of the POMDP model are analyzed and judged by comparing the pre-trained model with the simulation calculation results. If the conditions are met, proceed to the next step; otherwise, return to step 8.

[0118] Step 11: Deploy the hardware and software systems and the trained model on-site.

[0119] When the effectiveness, robustness, and generalization ability of the numerical simulation model meet the requirements, on-site deployment is carried out. Based on the model feedback results, the gate opening and closing force and damping facility control parameters are obtained (obtained from the training convergence of a partially observable Markov model). At this time, the vibration reduction of the sluice gate is achieved by controlling the opening and closing of the actual physical gate and the operation of the damping facility, and a manual control port is reserved for emergency situations.

[0120] Step 12, On-site simulation operation

[0121] Based on the meteorological, weather, and river hydrological characteristics of the sluice gate's location, the typical sluice gate load characteristics were analyzed, and the sluice gate was determined to have undergone at least one typical representative year of operation. During the simulated operation on-site, the actual system performance was monitored in real time, and manual methods were immediately employed to mitigate vibrations that were beyond the system's control.

[0122] Step 13, Dynamic monitoring data analysis

[0123] The system's on-site simulation effect is analyzed to assess its control capabilities and actual vibration reduction performance.

[0124] Step 14, Conformity assessment

[0125] The automatic control system is tested using different combinations of typical operating conditions and simulated extreme operating conditions over a period of more than one year to determine its actual effectiveness. If it passes the test, the system is put into formal operation; otherwise, the mathematical model of the sluice gate is modified and improved.

[0126] This invention also provides an intelligent avoidance system that can effectively reduce the impact of harmful vibrations on the safety, durability, and normal operation of sluice gates. The intelligent avoidance system includes:

[0127] The model correction / data assimilation and numerical calculation analysis subsystem is used for data assimilation between measured data and mathematical models, thereby ensuring the validity of mathematical models, calculation parameters, and boundary conditions. The assimilated numerical model is used to perform time history analysis and sensitivity analysis of the sluice gate under dynamic-static coupling, thus providing a basis for optimizing dynamic monitoring points and controllers. Simultaneously, based on material strength tolerances and stability conditions, the time history intervals of gate position, attitude, and opening / closing force are limited. The subsystem includes a computing server, a graphics server, and computing and display workstations. It implements steps 2 and 4 of the intelligent avoidance method for flow-induced vibration of sluice gates.

[0128] The sensing and control subsystem is used to sense external excitations such as dynamic and static loads on the sluice gate, as well as its own acceleration, dynamic and static displacement, and attitude responses. Corresponding sensors include a dual-light sensor for upstream and downstream flow image and video of the sluice gate, a slit jet velocity and flow rate sensor, a wind speed and direction sensor, a rain gauge, rain direction and raindrop spectrum sensors, a gate opening and closing force and attitude sensor (three-dimensional tilt sensor), a three-dimensional displacement sensor for the sluice gate and gate, a three-dimensional laser rangefinder for the gate's relative position, and upstream and downstream water level gauges and flow velocity instruments. The water level gauges are radar level gauges, and the flow velocity instruments are acoustic phased array / multibeam underwater velocity and topographic measurement instruments. This system implements steps 3 and 5 of the intelligent avoidance method for flow-induced vibration of the sluice gate.

[0129] This intelligent diagnosis, assessment, and decision support system, driven by a data-knowledge hybrid approach, incorporates intelligent signal denoising, enhancement, feature extraction, and some observable Markov decision / reinforcement learning algorithms. By fusing perceptual information and relevant knowledge, it employs machine learning and trial-and-error methods to intelligently avoid harmful vibrations in sluice gates. It implements steps 6-14 of the intelligent avoidance method for flow-induced vibrations in sluice gates.

[0130] The 3D visualization subsystem is used to construct BIM+GIS models of sluice gates and their resonance scenarios. It adopts a 3D visualization engine and augmented reality technology, and is driven by a mixture of measured data and mechanical / hydrodynamic models to realize high-fidelity display of the current state of sluice gate vibration, evolution direction and consequences of failure.

[0131] It should be noted that the above content merely illustrates the technical concept of the present invention and should not be construed as limiting the scope of protection of the present invention. For those skilled in the art, various improvements and modifications can be made without departing from the principle of the present invention, and all such improvements and modifications fall within the scope of protection of the claims of the present invention.

Claims

1. A method for intelligently avoiding flow-induced vibration in a sluice gate, characterized in that, Includes the following steps: Step 1: Collect data on the operating environment of the sluice gate and the controllable parameters of each opening and closing electromechanical equipment, as well as the calculation, static and dynamic response results and failure modes of this project and related projects; Step 2: Based on the mechanical model of sluice gate vibration, establish the thermo-fluid-structure interaction equations under sluice gate-fluid excitation. Taking the sluice gate as the object of force, its loads include wind and rain excitation, static and dynamic water and wave forces, temperature load, gravity, inertial force, damping force, and support force. Specifically, establish the following model: (1) The main structure sluice gate model The dynamic model of the sluice gate includes the dynamic conservation method and the energy conservation equation, among which the equilibrium equation for dynamic analysis is: In the formula, [M] is the mass matrix of reinforced concrete and metal structures; This reflects the inertial effect. [C] is the second derivative matrix of nodal displacement with respect to time; [C] is the damping matrix, including the damping of the structure itself, the water body, the supports, and the vibration damping devices. This reflects the damping effect. t represents time; F[t] is the load array, which includes the forces exerted on the object from the outside. (2) Wind and Rain Two-Phase Flow Model Considering the effects of wind and rain coupling on the sluice gate, a two-phase flow model is adopted. The two-phase flow model includes mass conservation, momentum conservation, and energy conservation equations, with the specific forms as follows: mass conservation equation: Momentum conservation equation: Energy conservation equation: In the formula: k represents the phase code, when k = a represents the gas phase, k = l represents the liquid phase, ρ k Represents the k-phase density, u k Represents the k-phase velocity, g represents the gravitational acceleration, and σ k Let ω be the shear stress tensor. k For water and air phase forces, p is the pressure scalar of each phase, and e is the pressure of each phase. k Where K is the specific thermodynamic energy, T is the thermal conductivity, and Q is the temperature; (3) The wind-wave-current coupled model is used below the water surface. Wave motion dynamic equation: Continuity equation: In the formula: x, y, z are spatial coordinates in the Cartesian coordinate system; t is time; u, v, w are the velocity components of the fluid in the x, y, z directions; ρ is the density of the actual water body; and g is the acceleration due to gravity. (4) Mathematical model for calculating the flow rate of a sluice gate with arbitrary gate opening under the constraints of regulations Formula for calculating the passage volume in the free outflow state of the gate: Garbrecht Official: Du Yi Official: In the formula, μ0 is the orifice flow coefficient; b is the gate width; e is the gate opening; and H0 is the water level upstream of the gate. Relative opening; Step 3: Collect actual response data of the sluice gate. After the data is collected, perform gross error judgment to eliminate erroneous data. Step 4: For sluice gates with measured response data, the measured data is used to invert, assimilate, or correct the mechanical and thermodynamic parameters of the sluice gates, thereby ensuring the correctness of the model and calculation parameters. Step 5: Optimize the sensing points and the layout of vibration reduction equipment and facilities; Step 6: Use the finite element method to calculate the mathematical model established in Step 2 to obtain the initial parameters or pre-training samples of the POMDP model for the next step; for some physical quantities in the calculation boundary conditions that have large uncertainty, large variation, or are sensitive to vibration, use the random sampling method to perform Monte Carlo random sampling. Step 7, establish the POMDP model and define the vector M = (S, A, O, T, Z, R, B), where S, A, and O represent the state vector, action vector, and implementation vector, respectively; O represents the observed physical quantity acquired by the monitoring instrument; T represents the probability distribution of mutual transitions between states; Z represents the observation probability distribution; R represents the reward obtained; and B represents the distribution in the state set S at the initial time. Based on step 6, after multiple calculations to obtain a sufficient number of samples, the following probability distributions are calculated: T(s,a,s')=Pr(s'|s,a) represents the state transition probability, that is, the probability distribution of transitioning to other states s' after performing action a in state s; Z(s,a,o)=Pr(o|s,a) represents the observation probability, that is, the probability distribution of obtaining the observation value o after performing action a in state s; these are used to establish a Markov decision model. Establish penalty and reward functions; vibrations with any of the following characteristics are considered harmful vibrations and will be subject to corresponding penalties, the magnitudes of which are as follows: When both phenomena occur simultaneously, the penalty is set to the more severe value, and the reward to the smaller value: (1) No vibrations that could exceed the sensor sensitivity: +5; (2) The amplitude exceeds the sensor sensitivity but is within the allowable range and is getting smaller and smaller: +4; (3) If the value exceeds the sensor sensitivity but is within the allowable range and does not increase steadily: +2; (4) When the vibration amplitude exceeds the sensor sensitivity and is greater than a certain value: -5; (5) If the sensor sensitivity is exceeded and there is impact and wear between structures: -10; (6) When resonance occurs that exceeds the sensor sensitivity: -10 (using the B-R dynamic stability criterion); (7) When the amplitude exceeds the sensor sensitivity and resonance occurs, and the amplitude increases: -15; Step 8: Pre-training and field training of a partially observable Markov decision model. Pre-training uses the computational samples from step 6, while actual application uses on-site deployed sensors and actuators for actual drills and measurements. Model inputs include: The inputs include: β: initial belief, the initial value of B, obtained through random finite element sample estimation or pre-training; ε0: target distance between μ(b0) and l(b0), obtained through the following iterations; ξ: target distance reduction rate, determined empirically or through trial and error; K: number of sampling scenarios, determined according to different sluice gate scheduling conditions; D: maximum depth of DESPOT, determined based on convergence; λ: regularization constant, determined empirically or through trial and error; T max The maximum online planning time for each step is determined based on experience or trial calculations. The specific iteration steps are as follows: (1) Initialize a belief b with β; (2) Execute (3); (3) Execute sub-function 1: BUILDDESPORT(b) to create a new tree l; (4) Calculate the action branch a that maximizes the value of l(b,a). * =max a∈A l(b,a); (5) Determine whether the condition L0(b) > l(b, a) is satisfied. * If the condition is met, execute (6); otherwise execute (7). (6) Update action branch a * For π0(b); (7) Execute a * ; (8) Accept the observed branch z; (9) Add a child node b = Т(b, a) * ,z); Step 9, Training and Convergence Stability Analysis The convergence condition of the partially observable Markov model training is used to determine whether training is stopped when the convergence condition is met. Step 10, Qualification Assessment The effectiveness, robustness, and generalization ability of the POMDP model are analyzed and judged by comparing the pre-trained model with the simulation calculation results; if the conditions are met, proceed to the next step; otherwise, return to step 8. Step 11: On-site deployment of hardware and software systems and the trained model. When the effectiveness, robustness and generalization ability of the numerical simulation model meet the conditions, it is deployed on site. Based on the model feedback results, the gate opening and closing force and damping facility control parameters are obtained. At this time, the vibration reduction of the sluice gate is achieved by controlling the opening and closing of the actual physical gate and the operation of the damping facility, and an emergency manual control port is reserved.

2. The intelligent avoidance method for flow-induced vibration of a sluice gate according to claim 1, characterized in that, It also includes the following steps: Step 12, On-site simulation operation Based on the meteorological, weather, and river hydrological characteristics of the sluice gate's location, analyze the typical sluice gate load characteristics and determine the sluice gate's on-site operation over a typical representative period of more than one year; during the on-site simulated operation, monitor the actual effect of the system in real time, and immediately adopt manual methods to avoid vibrations when uncontrollable vibrations occur; Step 13, Dynamic monitoring data analysis Analyze the on-site simulation effect of the system; Step 14, Conformity assessment The automatic control system is tested using different combinations of typical operating conditions and simulated extreme operating conditions for more than one year to determine its actual effectiveness. If it passes the test, the system is put into formal operation; otherwise, the mathematical model of the sluice gate is modified and improved.

3. The intelligent avoidance method for flow-induced vibration of a sluice gate according to claim 1, characterized in that, In step 1, the sluice gate operating environment data includes at least: sluice gate location, structural material data, wind and rain load, gate opening, flow velocity and flow rate, and upstream and downstream head data. The structural material data includes at least: location data of the gate chamber, bottom plate, gate pier, retaining wall, gate metal structure, and electromechanical equipment. The gate opening includes at least the gate combination and opening height.

4. The intelligent avoidance method for flow-induced vibration of a sluice gate according to claim 1, characterized in that, In step 3, the erroneous data is judged based on the instrument measurement principle and the physical meaning of the measured value, and the correlation and rationality of the data are used to determine the measured data.

5. The intelligent avoidance method for flow-induced vibration of a sluice gate according to claim 1, characterized in that, In step 4, for sluice gates without actual measured data, on-site ultrasonic testing methods are used or calculation parameters are obtained based on experience.

6. The intelligent avoidance method for flow-induced vibration of a sluice gate according to claim 1, characterized in that, Step 5 includes the following sub-steps: Step 5.1: The water level, air temperature, wind speed and direction, and rainfall intensity upstream and downstream of the sluice gate are measured by the following points according to relevant specifications. The flow state, attitude, and location are measured according to the principles of representativeness, maximum amplitude, and maximum sensitivity to vibration. The vibration sensor is arranged by using the three-dimensional effective independent method, the effective independent-driving point residual method, and the QR-MAC hybrid algorithm for optimization. Step 5.2: Optimize the layout of the vibration damping equipment. The optimization layout method adopts a genetic algorithm or a continuous search method.

7. The intelligent avoidance method for flow-induced vibration of a sluice gate according to claim 1, characterized in that, Step 6 includes the following sub-steps: Step 6.1, Sampling of key variables (1) Based on the different combinations of upstream and downstream water level ranges, wind speed ranges, rainfall intensity ranges, gate opening ranges, and the number of gates, the dynamic response time history of the sluice gate under different working conditions is obtained according to the uniform design combination time history. When the amplitude of the sluice gate increases significantly, the attitude changes abruptly, or the vibration frequency changes significantly, it is determined to be a resonance condition. Based on the sensitivity analysis results, the dynamic optimization method is used to determine the placement of the acceleration sensor, attitude sensor, and position sensor. (2) The dynamic and static response, gap flow and discharge flow of the sluice gate under different gate openings and upstream and downstream water level combinations were calculated using a three-dimensional river hydrodynamic wave-current coupled model of wind and rain action between upstream and downstream of the sluice gate. The vibration reduction effect of the sluice gate under different combinations of conditions and its observable measurement were obtained by calculating the optimized vibration reduction measures. Step 6.2: The dynamic response of the sluice gate is numerically calculated using the domain decomposition method. For metal gates, the substructure method is used for numerical calculation. The typical regions and key processing methods are as follows: (1) The SST k-ε turbulence model was applied to the wind and rain area above the sluice gate to numerically simulate the three-part force coefficients of the sluice gate section under different angles of attack. The pressure and velocity distribution diagrams of the flow field around the section were given and analyzed. The two-phase flow theory of the Euler-Euler system was adopted, and the raindrop field was taken as the continuous medium field. For constant non-storm rainstorm, the Reynolds average method was used to obtain the rain phase streamline diagrams under different rain phase particle sizes and different wind speeds, as well as the raindrop capture rate and impact load on the model surface. For storm rainstorm, the large eddy simulation method was used to obtain the load time history data and force spectrum in the downwind, crosswind and torsional directions. The wind field adopted the Kaimal spectrum and Lumley-Panofsky spectrum. The raindrop spectrum reproduction methods were Rosin-Rammler preset function method and CFD custom function method. A numerical simulation method based on two parameters of wind and rain coupling field was established by rain intensity conversion and raindrop spectrum. (2) For the upstream and downstream water bodies, waves and wind and rain mixed fields of the sluice gate, the two-phase flow theory of the Eulerian-Eulerian system is adopted, and the raindrop field is regarded as the continuous medium field to perform three-dimensional numerical simulation of the wind and rain movement around the sluice gate. The rain phase streamline diagrams under different rain phase particle sizes and different wind speeds are obtained by using the Reynolds average method, as well as the raindrop capture rate and impact load on the model surface. The load time history data and force spectrum in the downwind, crosswind and torsional directions are obtained by using the large eddy simulation method. The numerical simulation of the gate water movement is carried out by using the linear mass source wave generation method combined with the ALE interface tracking technology. At the same time, the water-air interface tracking is established based on the VOF interface capture technology. The wind-wave joint field numerical model is established based on the VOF method. The applicability of the numerical method is verified by comparing and analyzing the experimental and numerical simulation results. The flow calculation in the near wall area and the flow calculation solution at low Re numbers include the wall function method and the k-ε model at low Re numbers. (3) For opening and closing ropes or flexible outdoor structures, based on the EWF model and the Euler-Lagrange discrete phase model, and using the enhanced wall method combined with dynamic mesh technology, a numerical simulation method for the combined action of wind and rain on the structure was established to analyze the dynamic characteristics of the sluice gate under the action of wind and rain. (4) In the process of establishing the overall grid of the sluice gate and wind and rain load, the method of "rigid motion region + dynamic grid region + static grid region" is used to first divide it into blocks, and then the Newmark-β method code is written into the user-defined function of FLUENT to solve the vibration response of the structure. The structural vortex vibration numerical simulation method is established by combining the dynamic grid technology in FLUENT software. (5) The Lagrangian and Euler variables are transformed through an approximate smooth δ function, that is, the solid force density on the nodes of the finite element is distributed to the Gaussian integration points of the element through the element shape function, and then the solid force density on the Gaussian integration points in the support domain of the approximate smooth δ function is transformed to the fluid element; similarly, the fluid velocity on the Euler mesh is first transformed to the Gaussian integration points of the solid element through the δ function, and then the velocity of the solid node is obtained through the element shape function, and then the displacement of the solid is obtained; Step 6.3, Overall Partition Calculation and Boundary Layer Processing The static and dynamic response of the sluice gate under the combined effects of wind, rain, water flow, waves, and heat is analyzed using a domain decomposition combined with a substructure method. The selected upstream and downstream calculation ranges of the river channel are stable for the sluice gate vibration calculation results. The static boundary uses flow rate, water level, and pressure boundaries, while the dynamic boundary condition uses a non-reflection boundary. A certain calculation area is selected upstream and downstream of the sluice gate, with a certain river length, height, and depth. The upper elevation is the height at which the influence of wind and rain on the sluice gate is negligible, the depth is the depth at which the stress and deformation state of the sluice gate, including the pile foundation, is negligible under various loads, and the upstream and downstream areas are the lengths at which the influence of the sluice gate's flow state and flow field is negligible. The top of the flow field adopts symmetric boundary conditions, which are equivalent to free-slipping walls; the sluice gate section and the ground surface adopt no-slip wall conditions; the standard k-ε model or the improved model is selected, and the finite volume method coupled with the finite element method is used for calculation. Step 6.4: Identify hazardous vibrations through numerical calculation.

8. The intelligent avoidance method for flow-induced vibration of a sluice gate according to claim 1, characterized in that, The The BUILDDESPORT(b0) function creates a new tree and includes the following process: ① A set of K scenarios randomly sampled from the current belief b0. ② Create a new DESPOT D, with node b as the root node; ③ Initialize the upper and lower bounds U(b0), L0(b0), μ(b0), l(b0); ④ Calculate the distance between nodes ⑤ If the node distance The target distance is ε0 and the total running time is less than T. max If yes, then execute step 6; otherwise, execute step 9. ⑥ Execute branch algorithm 2: EXPLORE(D,b0) to heuristically explore node b; ⑦ Perform upper and lower bound backups at each node along the path, backing up the upper and lower bounds and distances of each node x on the path from b to D, μ(x), l(x), U(x); ⑧ Calculate the distance between the upper and lower bounds ε(b0) = μ(b0) - l(b0), and then execute ⑤; ⑨ Return the created tree l; The EXPLORE(D,b) function performs heuristic exploration, expanding D and narrowing its upper and lower boundaries. Its process is as follows: ① If the tree height Δ(b) ≤ D and the excess uncertainty E(b) > 0, and the subfunction 3PRUNE(D,b) returns FALSE (cannot be pruned), then execute ② below; otherwise execute ⑦ below. ② Determine if b is a leaf node of D. If yes, execute ③ below; otherwise, execute ④ below. ③ Insert node b′ between node b and its parent node, making it the new parent node and initialize the upper and lower bounds of b′; ④ Determine the optimal action branch a * =argmax a∈A μ(b,a); ⑤ Observe the optimal branch ⑥ Determine the new node b = τ(b, a) * ,z * Then execute 1; ⑦ If the height of the tree Then execute step ⑧ below; otherwise, execute step ⑨ below. ⑧ Reset the boundaries to simplify the search: upper bound U(b) = L0(b0), lower bound u(b) = l0(b0), distance l(b) = l0(b0); ⑨ Return to node b; The PRUNE(D,b) function prunes the node from b back to the root. As the boundary is updated, new nodes may meet the pruning conditions and be pruned. The steps of subfunction 3 are as follows: ① Initially, BLOCKED = FALSE; ② Traverse every node x from node b to root node D. If the traversal is not finished, execute ③ below; if the traversal is finished, execute ⑨ below. ③ Determine whether node x is blocked by any ancestor node. If it is blocked, execute ④ below; otherwise, execute ⑨ below. ④ Reset the boundaries to simplify the search. The upper bound is U(b) = L0(b0), the lower bound is u(b) = l0(b0), and the distance is l(b) = l0(b0). ⑤ Perform upper and lower bound backups at each node along the path, backing up the upper and lower bounds and distances of each node x on the path from b to D, μ(x), l(x), U(x); ⑥ Set BLOCKED to TRUE; ⑦ Execute 2; ⑧ Execute 9; 9. Return to BLOCKED.

9. A smart system for avoiding flow-induced vibration in a sluice gate, characterized in that, The method for intelligently avoiding flow-induced vibration of a sluice gate as described in any one of claims 1-8 includes: The model correction / data assimilation and numerical calculation analysis subsystem is used for data assimilation between measured data and mathematical models. The assimilated numerical model is used to perform time history analysis and sensitivity analysis of the sluice gate under the dynamic-static coupling effect. At the same time, based on material strength tolerance and stability conditions, the time history intervals of gate position, attitude and opening and closing force are limited. The subsystem includes a computing server, a graphics server, and computing and display workstations. The sensing and control subsystem is used to sense the external excitation of dynamic and static loads on the sluice gate and its own acceleration, dynamic and static displacement and attitude response. The corresponding sensors include a dual-light sensor for upstream and downstream flow image and video of the sluice gate, a slit jet flow velocity and flow rate sensor, a wind speed and direction sensor, a rain gauge, a rain direction and raindrop spectrum sensor, a gate opening and closing force and attitude sensor, a sluice gate and gate three-dimensional displacement sensor, a three-dimensional laser rangefinder for the relative position of the gate, and upstream and downstream water level gauges and flow velocity instruments. Among them, the water level gauge is a radar water level gauge, and the flow velocity instrument is an acoustic phased array / multibeam underwater velocity and topographic measurement instrument. Based on a data-knowledge hybrid-driven intelligent diagnosis, assessment and decision support system, it incorporates intelligent signal denoising, enhancement, feature extraction and some observable Markov decision / reinforcement learning algorithms. By fusing sensory information and relevant knowledge, it uses machine learning and trial-and-error methods to achieve intelligent avoidance of harmful vibrations in sluice gates. The 3D visualization subsystem is used to construct BIM+GIS models of sluice gates and their resonance scenarios. It adopts a 3D visualization engine and augmented reality technology, and is driven by a mixture of measured data and mechanical / hydraulic models to realize a high-fidelity display of the current state of sluice gate vibration, evolution direction and consequences of failure.

Images