Large-span bridge substructure vehicle-fire-wind real-time hybrid test platform and implementation method

Abstract

The application provides a large-span bridge substructure vehicle-fire wind force real-time hybrid test platform and implementation method, the platform comprises a physical loading and transmission system, a T-shaped wind-fire coupling simulation system, a modularized prying integrated system, an intelligent control and calculation system and a multi-dimensional high-precision monitoring system. The method realizes real-time pre-test of the boundary displacement / force of the test piece through the construction of a numerical substructure fast calculation model based on LSTM and converts the boundary displacement / force into a loading instruction to drive the through core jack to realize thermal force coupling interactive loading; the T-shaped wind-fire coupling simulation system reproduces the real vehicle fire scene under the influence of the environmental wind. The test platform supports independent loading of any component of the cable system, can execute open space self-defined fire test and closed space standard fire test, and accurately analyzes the cable failure mode, critical temperature threshold and damage evolution law under fire through the multi-dimensional high-precision monitoring system, thereby providing a high-fidelity test platform and analysis method for fire resistance design of large-span bridges.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of bridge engineering fire resistance performance testing, in particular to a large-span bridge substructure vehicle-fire-wind real-time hybrid test platform and implementation method. BACKGROUND

[0002] As the core hub of the transportation network, the fire risk of bridges has a significant destructive and chain effect. The cable system substructure of cable-supported bridges such as long-span suspension bridges and cable-stayed bridges is the core unit of the overall bridge stress, and its mechanical properties directly affect the safety of the whole bridge. The high temperature of fire causes the tensile strength of steel to decrease sharply, and the internal force redistribution of the cable system due to thermal expansion may trigger local buckling or overall instability; the bridge fire caused by the impact of oil tank trucks can make the cable temperature exceed 800℃ in a few minutes, causing the cable strand to break and the main cable to have a sudden change in sag, and then triggering the continuous collapse of the whole bridge. The coupling effect of high temperature and mechanical load further accelerates the structural failure, posing a serious threat to public safety and regional economy.

[0003] Fire tests are the core means to evaluate the fire resistance performance of bridge structures and verify the effectiveness of protection design. The mechanical behavior of the cable system under fire has strong nonlinear characteristics, such as high-temperature creep, thermal conduction differences between cable strands, etc., and numerical simulation is difficult to accurately simulate and predict its critical failure threshold and damage path. Through fire tests, the high-temperature performance degradation law of cable materials, thermal stress concentration effect in the anchorage zone, etc. can be systematically analyzed, providing direct data support for the selection of fireproof coatings or fireproof structural measures, the development of fire emergency evacuation strategies, and the design of repair schemes. In addition, fire tests have irreplaceable engineering value for improving bridge fire resistance design specifications and reducing the risk of the whole life cycle.

[0004] With the continuous growth of bridge span and load demand, the diameter of the cable is usually more than 1000mm, and the design load is up to ten thousand tons. Due to the problems of insufficient equipment carrying capacity, difficulty in reproducing wind-fire coupling scenarios, and limitations of whole-bridge catastrophe analysis methods, traditional test methods are difficult to truly reflect the damage evolution law of the cable system under complex fire scenarios. The main technical bottlenecks of existing technologies are as follows:

[0005] Firstly, the boundary constraint state of the cable system substructure free hovering multidirectional loading is distorted, and the large-scale cable has insufficient large-tonnage loading capacity. In existing cable loading tests, the cable ends are usually completely rigidly fixed by small and simple counterforce frames before being horizontally tensioned, while the actual serving cable assumes a catenary shape under the action of gravity in the bridge and has a certain inclination angle. The existing test method seriously deviates from the actual structure stress state and is prone to draw incorrect scientific conclusions. In addition, the conventional loading device is designed mainly for single component or small size cable (diameter less than 100 mm), which cannot adapt to large scale / full scale model test of cable and has insufficient loading capacity (loading limit less than 200 t), resulting in that the test results cannot reflect the nonlinear response and collaborative work characteristics of the cable system in real engineering.

[0006] Secondly, it is difficult to reproduce the real fire scene of wind-fire coupling. Real fire has highly complex dynamic characteristics: the flame in the fire field often assumes a natural gradient decreasing shape with high middle and low ends, and the flame sways under the influence of environmental wind. The dynamic change process of the non-uniform temperature field in the whole fire field space is closely related to the state of the environmental wind such as wind direction and wind speed. However, current researches are mostly limited to laboratory fixed environment, relying on uniform temperature electric furnace heating or spatial uniform temperature natural gas test furnace following standard temperature rising curve for simulation. However, these methods are difficult to restore the three-dimensional flame shape, flame dynamic spread and non-uniform spatial temperature field under the interaction effect of wind field coupling in real fire. For example, in the fire accident of a suspension bridge, the flame usually spreads upward from the lowest point of the main cable (height ≥ 10 meters), forming full-wrapped combustion of the main cable bottom, fire-facing surface and suspender, and the flame height often exceeds 10 meters. However, the existing fire test device cannot simulate the thermal impact and ablation process of the flame on the bridge cable under the action of crosswind, resulting in limited reproduction effect of real fire scene.

[0007] In addition, the existing most advanced combustion device has obvious deficiency in combustion height, which is usually limited within 1.5 meters. Therefore, to simulate the scale of real fire accident, natural oil pool fire combustion test must be used. However, this method needs to consume a large amount of fuel and material, and the combustion process will cause serious air pollution. More importantly, the development process of fire test is completely unpredictable and uncontrollable, which has extremely high risk. On the other hand, the existing combustion device also has obvious limitations in function. They lack the dynamic regulation and control ability of wind-fire coupling in real fire environment, and it is difficult to achieve precise control in the whole cycle. At the same time, these devices usually have single function and fixed site, and cannot flexibly and quickly adapt to the historical fire accident scene reconstruction demand of different sites. These problems seriously restrict the accuracy and practical application value of fire research.

[0008] Thirdly, the single scale component fire test cannot reflect the whole bridge disaster law. The traditional method relies on local component test, such as single scale main cable or suspender section, ignores the collaborative working mechanism of cable system, bridge tower and stiffening beam, and is difficult to reveal the heat transfer and failure path of the whole structure under fire and the internal force redistribution law. Due to the size effect of the scale model, the data reliability is insufficient. Although existing numerical tools such as ABAQUS and OpenSees can establish numerical models for thermal-mechanical coupling effect, the calculation efficiency is low, which cannot meet the real-time transmission demand of real-time hybrid test, and the accuracy of the calculation results is also difficult to verify through accurate and reliable full-scale or large-scale whole structure fire resistance test.

[0009] Fourthly, the existing fire resistance equipment has insufficient universality, limited loading capacity and only supports single type component, and cannot quickly switch the test demand of different cable system substructures such as suspension bridge and cable-stayed bridge. The test mode is rigid, and only the standard fire temperature curve of ISO834, HC and the like can be reproduced, and the self-defined modeling capability of the non-uniform spatial temperature field under the influence of the environment wind such as the oil tank truck, truck fire and cable short circuit induced fire is lacked. In addition, the traditional device lacks modular design, and the burning system, loading mechanism and monitoring unit cannot be flexibly reorganized according to the test demand, resulting in low test efficiency and poor data comparability.

[0010] In view of the above technical bottlenecks, it is urgent to develop a hybrid test device and method which integrates large-tonnage real boundary physical loading, efficient and safe controllable wind-fire coupling burning device and whole structure high-fidelity thermal-mechanical coupling deduction. SUMMARY

[0011] The purpose of the present application is to provide a large-span bridge substructure vehicle fire wind real-time hybrid test platform and implementation method, which breaks through the large-scale / foot-scale loading limitation of the traditional test by optimizing the reaction frame loading constraint structure and the force transmission path, modularizing the dynamic wind-fire coupling process of the fire, intelligently controlling the dynamic multi-field cooperation and fusing the numerical and physical cooperation technology, realizes the reproduction of the non-uniform temperature field of the fire and the numerical and physical cooperation analysis of the whole bridge. The device has wide universality and supports independent or cooperative fire resistance performance test of components such as main cable, stay cable, suspender and cable clamp. At the same time, it covers the dynamic fire spread in open space, the standard temperature rise in closed space and the complex wind-fire coupling environment, can reproduce the ISO834, HC standard fire curve, and can also define the non-uniform temperature field of the real fire scene such as oil tank truck explosion and cable short circuit. It is applied to the demonstration of national major bridge projects such as the Shendaotong Channel and the Shiziyang Bridge, and provides whole life cycle support for bridge fire resistance design, disaster warning and repair strategy, and helps the safety construction and long-term operation of major projects in China.

[0012] In order to achieve the above application purpose, the technical scheme adopted by the present application is specifically as follows:

[0013] The application provides a large-span bridge substructure vehicle-fire wind force real-time hybrid test platform, which comprises:

[0014] a physical loading and transmission system, a T-shaped wind-fire coupling simulation system, a modularized skid-mounted integrated system, an intelligent control and calculation system and a multi-dimensional high-precision monitoring system;

[0015] The physical loading and transmission system comprises a cable system self-balancing counterforce frame, a beam electric lifting locking system and a roller bearing transmission system, and realizes installation and positioning of a test piece and application of force load.

[0016] The T-shaped wind-fire coupling simulation system comprises a T-shaped global flame envelope combustion system and a multi-dimensional wind field simulation system, the T-shaped global flame envelope combustion system comprises a horizontal linear combustion module and a vertical annular combustion module, and the T-shaped wind-fire coupling simulation system can realize dynamic envelope of a flame in a three-dimensional space, simulate a thermal impact and ablation process of the flame on a bridge cable under the action of an environmental wind in a real fire.

[0017] The intelligent control and calculation system comprises a fire temperature field rapid generation model, an LSTM-based numerical substructure rapid calculation model, an LSTM-PID intelligent control system and an LSTM time lag compensator.

[0018] The multi-dimensional high-precision monitoring system comprises a fire field panoramic video acquisition unit, a fire field space temperature field monitoring unit and a wind speed monitoring unit.

[0019] Preferably, the cable system self-balancing counterforce frame comprises two groups of symmetrically arranged rectangular lattice frames, the two groups of rectangular lattice frames are each provided with a base at the bottom and the two bases are connected through a box-type bottom beam, and the two groups of rectangular lattice frames are connected through a main cross beam at the top; a sliding cross beam is arranged in the middle of each of the two groups of rectangular lattice frames, and the sliding cross beam is freely slidable and lockable along the height direction of the rectangular lattice frame.

[0020] The cable system self-balancing counterforce frame further comprises a free-rotation force boundary subsystem and a free-rotation anchoring subsystem, the free-rotation force boundary subsystem and the free-rotation anchoring subsystem each comprise a composite anchor ring, the composite anchor ring is installed in the sliding cross beam and freely rotates in the sliding cross beam, and the free sliding of the sliding cross beam realizes hovering tension at any angle during loading of a main cable model; the free-rotation force boundary subsystem further comprises a first through-type jack and a loading end two-stage gradient anchoring unit, the first through-type jack is fixedly installed in the composite anchor ring, and the loading end two-stage gradient anchoring unit comprises a loading end primary anchor and a loading end secondary anchor, which are arranged on the two sides of the first through-type jack; the free-rotation anchoring subsystem further comprises an anchoring end anchor.

[0021] The variable cross-section force transmission main shaft is fixedly installed on both sides of the composite anchor ring, and comprises a quadrangular platform force transmission section fixedly connected with the composite anchor ring and a cylindrical rotating section fixedly connected with the quadrangular platform force transmission section.

[0022] Preferably, the cable system self-balancing counterforce frame further comprises a sling force loading anchoring subsystem, which comprises a second through-type jack, a sling loading end anchor and a locking anchoring mechanism; the second through-type jack is embedded in the internal cavity of the box-type bottom beam, the sling loading end anchor is anchored at the end of the sling model, and the outer diameter of the sling loading end anchor is smaller than the inner diameter of the second through-type jack; the locking anchoring mechanism is arranged between the second through-type jack and the sling loading end anchor, and the inner diameter of the locking anchoring mechanism is equal to the diameter of the sling model, and the outer diameter of the locking anchoring mechanism is greater than the outer diameter of the second through-type jack.

[0023] The sling force loading anchoring subsystem further comprises a T-shaped sliding rail and a sliding rail limiter, the T-shaped sliding rail is fixedly installed on the box-type bottom beam, and the sliding rail limiter is installed on the T-shaped sliding rail; the second through-type jack moves on the T-shaped sliding rail and is fixedly limited by the sliding rail limiter.

[0024] Preferably, the horizontal beam electric lifting locking system is provided with two groups corresponding to the sliding horizontal beam, each group of the horizontal beam electric lifting locking system comprises a screw rod linkage synchronous lifting subsystem, a limiting sliding rail and an array type positioning hole group; the screw rod linkage synchronous lifting subsystem comprises a screw rod, an elevator, a motor and a base; the screw rod is provided with two symmetrically, one end of each screw rod is rotationally connected with the sliding horizontal beam through a flange, and the other end of the screw rod is connected with the elevator; the elevator is provided with two corresponding to the screw rod, the motor is linked with the two elevators through a transmission shaft; the base is fixedly installed on the top of the rectangular lattice type frame, and the elevator and the motor are fixedly installed on the base; the limiting sliding rail is integrated on the inner side of the rectangular lattice type frame, and forms a sliding pair with the sliding block at the end of the sliding horizontal beam; the array type positioning hole group is distributed at equal intervals along the height direction of the rectangular lattice type frame, and cooperates with the detachable bolt mechanism on the sliding horizontal beam to lock the position of the sliding horizontal beam.

[0025] The roller bearing transmission system comprises an adjustable roller bearing platform installed on the base, and the adjustable roller bearing platform comprises a base frame, a lifting mechanism and a top bearing beam; the lifting mechanism is arranged between the base frame and the top bearing beam, the top bearing beam moves up and down under the drive of the lifting mechanism, and the top bearing beam is equal in height to the box-type bottom beam after being lifted; a plurality of roller groups arranged at equal intervals are installed on the adjustable roller bearing platform and the box-type bottom beam, recessed clamping grooves for installing the roller groups are arranged on the top bearing beam and the box-type bottom beam, and the height of the base is lower than that of the box-type bottom beam.

[0026] Preferably, the horizontal linear combustion module comprises several frame burners arranged linearly and provided with vertical flame injection pipes, which are suspended and fixed in the middle test area of the cable system self-balancing counterforce frame by several steel wire ropes; one side wall of the frame burner is provided with a gas input interface and a self-suction air supply interface at the bottom; the top of the frame burner is provided with a multi-directional flame injection unit capable of forming a three-dimensional flame gradient pattern, which comprises several side-by-side arranged gas collectors, and the gas collectors are uniformly distributed with several flame injection pipes;

[0027] The middle part of the gas collector is communicated with the gas input interface through a plurality of equal distribution branch pipes, the upper part of the branch pipe extends into the gas collector and is provided with a reverse v-shaped diffusion cutout, and the inside of the frame burner is provided with a flow equalization assembly; air enters the frame burner from the self-suction air supply interface of the frame burner, and the flow equalization assembly forms a flow equalization air supply multi-directional flame injection unit; the reverse v-shaped diffusion cutout realizes uniform decreasing diffusion of gas from the middle part of the gas collector to both ends, and the flow equalization air formed by the flow equalization assembly synchronously ignites each gas collector to reproduce a real three-dimensional flame natural gradient pattern, and a real fire space non-uniform temperature field is presented.

[0028] The gas input interface is independently and flow-adjustably supplied with gas through a modularized pry integrated system, and the self-suction air supply interface realizes natural air supply through the negative pressure self-suction principle; the flow equalization assembly comprises an air flow equalization plate and two arc-shaped flow guide partitions arranged along the air flow direction, the self-suction air supply interface forms a rectangular gradually expanding flow channel in the inside of the frame burner, and the two arc-shaped flow guide partitions arranged along the air flow direction uniformly divide the air inlet of the frame burner into three left, middle and right paths to initially divide the air; the air flow equalization plate is arranged at the lower part of the multi-directional flame injection unit to secondarily equalize the air flow.

[0029] Preferably, the vertical ring-shaped combustion module comprises several layers of detachable ring-shaped burners, and adjacent ring-shaped burners are connected through detachable support bodies; each layer of ring-shaped burner is uniformly provided with several flame injection ports capable of adjusting the flame injection angle, the flame injection ports are supplied with gas through the symmetrical double gas inlet branch pipes arranged on the ring-shaped burner, high-energy pulse spark igniters are arranged at any two symmetrically distributed flame injection ports in each layer of ring-shaped burner, and the LSTMPID intelligent control system synchronous control circuit is used to realize synchronous ignition of each layer of flame; each double gas inlet branch pipe is independently and flow-adjustably supplied with gas through a movable pry integrated system.

[0030] The flame spout comprises an air inlet end short pipe and a flame spouting end short pipe, which are connected through an open hollow hemispherical universal hinge mechanism, the universal hinge mechanism comprises a hollow ball with an embedded flow guide channel and a hemispherical socket matched with the hollow ball, and can realize 0°-90° inclination adjustment of the flame spouting end short pipe, the air inlet end short pipe is communicated with the annular combustor through a variable diameter transition pipe; after the gas enters the annular combustor, it is accelerated into the universal hinge mechanism through the variable diameter transition pipe, and is spouted from the flame spouting end short pipe and ignited, forming a flame with adjustable spouting direction;

[0031] The annular combustor is integrated with a flange type quick release interface along its symmetry axis direction, the flame spout and the double air inlet branch pipes do not interfere with each other with the flange type quick release interface, the flange type quick release interface realizes axial separation and recombination of the annular combustor through bolt connection, after disassembly, two half annular combustor units can be formed, and each half annular combustor unit retains independent gas supply passage and ignition control, supporting true fire loading in half circle fire scene; the detachable support body is provided with an internal thread quick connector at both ends for connecting adjacent two layers of annular combustors.

[0032] Preferably, the multi-dimensional wind field simulation system comprises a plurality of high dynamic axial flow fans, which are arranged in a matrix topology in the middle test area of the cable system self-balancing counterforce frame, and each high dynamic axial flow fan is provided with an independent frequency converter;

[0033] The fire field panoramic video acquisition unit captures visible light and thermal radiation images of the three-dimensional flame gradient form fire field in real time through the fire field panoramic camera, and monitors the fire field panoramic video;

[0034] The fire field space temperature field monitoring unit comprises an infrared thermometer and a distributed thermocouple, which is used for real-time acquisition of fire field space temperature distribution data and generation of a three-dimensional temperature field model;

[0035] The wind speed monitoring unit is composed of an ultrasonic anemometer and a pressure sensor, which is used for real-time monitoring of wind speed, wind direction and wind pressure data of the multi-dimensional wind field simulation system; the data monitored by the multi-dimensional high-precision monitoring system is transmitted to the LSTM-PID intelligent control system in real time through a communication network.

[0036] Preferably, the modularized prying integrated system comprises a first prying unit and a second prying unit; the first prying unit is integrated with a gas cylinder or a fuel tank, a vaporization furnace, an explosion-proof fire extinguishing device, an explosion-proof lighting device and a gas pipeline communicated with the T-shaped global flame envelope combustion system; the second prying unit is integrated with a power distribution cabinet, a control cabinet and a frequency converter; the control cabinet is provided with a PLC controller and a communication module, which is used for realizing coordinated control of each system;

[0037] The explosion-proof fire extinguishing device comprises a linkage fire extinguishing system composed of an explosion-proof control cabinet, an explosion-proof fan, a combustible gas detector and a plurality of suspended ultra-fine powder fire extinguishers, and the explosion-proof lighting device is electrically connected with the explosion-proof control cabinet; the power distribution cabinet is used for providing power supply;

[0038] The gas pipeline comprises a main pipeline provided with a multi-channel converging interface, one end of the main pipeline is connected with a gas cylinder or a fuel tank through a high-pressure hose and an explosion-proof quick connector, the other end of the main pipeline away from the gas cylinder or the fuel tank is sequentially connected with a vaporization furnace, a gas main shut-off valve, an intelligent pressure feedback reducing valve and an intelligent valve positioner integrated with a liquid crystal screen, an input end of the intelligent valve positioner is communicated with a pressure stabilizing branch of an air compressor through a filter, and an output end of the intelligent valve positioner is communicated with a T-shaped global flame envelope combustion system through a corrugated flame arrester.

[0039] Preferably, the method for constructing the fire temperature field rapid generation model comprises:

[0040] S101, a GAN network is built, the GAN network comprises a generator and a discriminator; wherein,

[0041] The generator takes the fire source power, the normalized position coordinate and the wind speed as input conditions, extracts local fire dynamic characteristics through a double-layer spatio-temporal convolution LSTM branch, and embeds a heat diffusion equation residual term to construct a physical constraint branch; after the double-path features are fused through a full connection layer, a normalized temperature field matrix is output, and the normalized temperature field matrix is finally denormalized into an actual temperature-time curve;

[0042] The discriminator takes the generated or real temperature field matrix and the fire source parameter as joint input, extracts spatio-temporal joint features through a three-layer convolution network, embeds a gradient penalty term to constrain the smoothness of the generated data, and compresses a single node through global average pooling in an output layer, and outputs a temperature field authenticity probability, and simultaneously calculates a heat diffusion equation residual through a residual connection as a physical rationality criterion;

[0043] The heat diffusion equation residual is the difference between the generated temperature field and the predicted value of the theoretical equation, and the calculation formula is:

[0044] ;

[0045] In the formula, is the generated temperature field, is the ambient temperature, is the spatial second-order derivative of the generated temperature field, is the balance term of the heat source and the ambient temperature; is the L2 norm, used for measuring the size of the residual.

[0046] S102, generate full-bridge three-dimensional non-uniform temperature field data based on fire dynamics software simulation, extract space-time temperature distribution data, construct a training data set, verify data reliability through experiments, and compare measured and simulated temperatures to ensure that the error range is less than or equal to 5%;

[0047] S103, offline training of the fire temperature field rapid generation model based on the GAN network, and simulation based on the fire dynamics software and physical test verification data;

[0048] S104, deploy the trained fire temperature field rapid generation model to a computing server.

[0049] Preferably, the cable system bridge to be analyzed is divided into a test substructure and a numerical substructure; the real cable structure corresponding to the test piece is defined as the test substructure, including the real main cable and the corresponding suspender or sling; the stress state of the numerical substructure is calculated using the LSTM-based numerical substructure rapid calculation model;

[0050] The LSTM-based numerical substructure rapid calculation model is a multi-scale gated residual LSTM neural network prediction model, which is constructed by introducing a parallel multi-branch structure, a dynamic feature selection gate and a residual cross-layer connection based on LSTM, and is used to predict the evolution process under full-bridge fire according to the measured values of each sensor in the test and output the loading boundary force or displacement instruction in real time; The specific steps of training the LSTM-based numerical substructure rapid calculation model are as follows:

[0051] S201, build a multi-scale gated residual LSTM neural network, wherein the multi-scale gated residual LSTM neural network includes a double-branch LSTM unit, i.e., a first branch LSTM unit and a second branch LSTM unit;

[0052] S202, construct a data set for the LSTM-based numerical substructure rapid calculation model, specifically including: simulating three-dimensional non-uniform temperature field distribution in the full-bridge space under different fire scenarios using fire dynamics software, extracting temperature-time curves; establishing a thermal-mechanical coupling finite element model of the full-bridge and the test substructure based on finite element analysis software; making test pieces corresponding to the test substructure, and performing non-uniform temperature field thermal-mechanical coupling loading on the test pieces to verify the accuracy of the test piece thermal-mechanical coupling model; using the verified simulation data to train the multi-scale gated residual LSTM neural network model;

[0053] S203, offline pre-training, using a sliding time window to divide the data set, the first branch LSTM unit inputting original step data, the second branch LSTM unit inputting down-sampling data, and the dynamic feature selection gate weight being optimized through backward propagation of historical errors;

[0054] S204, online incremental learning, by real-time intercepting sensor data, only fine-tuning model down-sampling branch parameters and constraining weight change amplitude ≤0.01, dynamically adapting to fire changes;

[0055] S205, real-time prediction and control during real-time hybrid test process;

[0056] S206, cyclically iteratively updating thermal-displacement / force boundary conditions until specimen deformation or stress exceeds limit, outputting failure mode and critical threshold.

[0057] Preferably, step S201 comprises:

[0058] S2011, input layer receiving historical time series data, input features including position spatial coordinates, temperature field, boundary displacement, cable force and wind speed;

[0059] S2012, parallel double-branch LSTM unit, wherein first branch LSTM unit processes original step time series data for extracting short-time local mutation features; second branch LSTM unit processes down-sampling time series data for extracting long-time global evolution features; both first branch LSTM unit and second branch LSTM unit comprise first layer LSTM and second layer LSTM;

[0060] S2013, dynamic feature selection gate calculating branch weight through sigmoid function, fusion formula being:

[0061] ;

[0062] wherein, h final is hidden state vector after fusion of two branch LSTM units, h branch1 is output hidden state vector of first branch LSTM unit, h branch2 is output hidden state vector of second branch LSTM unit, g t means dynamic gating weight coefficient;

[0063] S2014, residual block cross-layer connection alleviating gradient vanishing, output and input of first layer LSTM and second layer LSTM being superimposed through skip connection, formula being:

[0064] ;

[0065] wherein, is hidden state vector of time step t in the i l th layer LSTM, is hidden state vector of last time step of time step t in the i l th layer LSTM, is spatial coordinate vector [x, y, z] of time step tt , is the hidden state vector of the first l layer LSTM, is the hidden state vector of the first l layer LSTM unit, l = (1, 2) ;

[0066] S2015, a mixed activation function; the forget gate and the input gate use a sigmoid function, the candidate memory cell uses a tanh function, and the output gate uses a Swish function f(x) = x·σ(x);

[0067] S2016, a fully connected output layer, outputting a boundary force prediction value or a displacement prediction value of the next time step.

[0068] Preferably, step S205 includes:

[0069] S2051, taking displacement loading as an example, converting the displacement of the next time step into a loading displacement instruction according to the displacement equivalence principle, and the formula is:

[0070]

[0071] In the formula, D load is the displacement instruction, D pred is the displacement of the next time step; L phy is the actual geometric span length of the physical specimen, L num is the equivalent span length of the numerical substructure, and for a large-scale model L phy = L num ; an experimental model with a scale ratio not less than 1:4 to the original structure is called a large-scale model.

[0072] S2052, driving the cable system self-balancing reaction frame through the LSTM-PID intelligent control system, and the control amount calculation formula is:

[0073]

[0074] In the formula, γ is a feedforward gain coefficient, which is dynamically adjusted through fuzzy logic; K p is a proportional gain coefficient, e(t) is the error between the real-time combustion state of the frame-type combustor and the target value, K i is an integral gain coefficient, and K d is a differential gain coefficient.

[0075] S2053, dynamically mapping the gating weight to the PID proportional coefficient to realize adaptive loading.

[0076] Preferably, the input layer of the LSTM-PID intelligent control system comprises temperature field data, wind speed, angle, burner valve opening and historical error, the hidden layer comprises 54-64 neurons, and the output layer outputs burner valve opening increment and fan frequency;

[0077] The generation step of the control instruction is:

[0078] S301, real-time collection of temperature field data, wind speed angle data and burner valve opening history sequence by using a sensor to construct a multi-dimensional time series input vector;

[0079] S302, inputting the input vector into a pre-trained LSTM prediction model to predict the burner valve opening increment and fan frequency at the next time, wherein the LSTM prediction model updates the weight online through a sliding time window, and the prediction error converges to <5%;

[0080] S303, taking the predicted valve opening of each burner or fan frequency as a feedforward correction, and respectively superimposing it to the feedback output of the LSTM-PID intelligent control system to generate a final control instruction;

[0081] In the horizontal linear combustion module, the control instruction calculation of the i th frame type burner is generated according to the following formula (1);

[0082] (1)

[0083] In the vertical ring combustion module, the control instruction calculation of the j th layer is generated according to the following formula (2):

[0084] (2)

[0085] The control instruction calculation of the frequency converter frequency of the control fan wind speed is generated according to the following formula (3):

[0086] (3).

[0087] Preferably, the LSTM time lag compensator main body is a neural network based on LSTM, comprising a hidden layer of 16-32 neurons, and the input layer nodes comprise historical time lag sequence Δt i , system real-time load rate, environmental temperature gradient and environmental wind speed gradient, wherein i=(1, 2,..., n), and the output layer is mapped to the next analysis step time lag prediction value Δt i+1 by a full connection layer, and specifically comprises the following steps:

[0088] S401, constructing a time lag data set, collecting a historical time lag sequence Δt iThe system real-time load rate, environmental temperature gradient and wind speed gradient three-dimensional data are constructed into an input vector, and input into the neural network based on LSTM, wherein, Δt i The delay time of the control command to the measured value of the sensor reaching the standard is referred to as a delay time;

[0089] S402, offline training and online optimization, the data samples in the preloading stage are used to train the time delay compensation model based on LSTM, and the loss function is the mean absolute error; when formally loaded, the network parameters are updated online by the optimizer every time an analysis step is completed;

[0090] S403, dynamic time delay compensation, the control command is sent in advance by Δt2 time before the second analysis step is loaded, and the actual time delay Δt is recorded real , and Δt2 refers to the time difference from the time when the command is sent to the time when it is loaded to the position;

[0091] S404, error correction and convergence, based on the deviation of Δt i-real and Δt i , the gradient descent method is used to correct the weight of the time delay compensation model based on LSTM, until the time delay error of continuous multiple analysis steps is less than or equal to 1 second; Δt i-real refers to the actual measured time delay of each analysis step, i=(1, 2,..., n).

[0092] The application also provides an implementation method of the large-span bridge substructure vehicle-fire-wind real-time hybrid test platform, which comprises the following steps:

[0093] Step 1, manufacturing and installing test pieces: the types of test pieces include main cable models, stay cable models and sling models;

[0094] Installing the main cable model or the stay cable model: synchronously lowering the two side sliding beams to the base reference surface;

[0095] Pulling the main cable model or the stay cable model through the free-rotation anchoring subsystem and the free-rotation force boundary subsystem in sequence and anchoring; then synchronously lifting the two side sliding beams to a preset height or angle;

[0096] Installing the sling model: connecting the sling model to the main cable model through the cable clamp, passing the lower end of the sling model through the second through-type jack and anchoring, and installing force sensors at the front ends of the first through-type jack and the second through-type jack;

[0097] Step 2, selecting a test mode and configuring an environment;

[0098] Mode A: open space self-defined fire test,

[0099] Inputting target fire parameters into the fire temperature field rapid generation model, generating a non-uniform temperature field and transmitting it to the LSTM-PID intelligent control system;

[0100] Mode B: closed space standard fire test

[0101] Through the quick-release interface to splice the enclosure unit to form a closed space, select the standard fire curve mode in the LSTM-PID intelligent control system, and close the multi-dimensional wind field simulation system;

[0102] Step three, preloading and time delay compensation initialization:

[0103] Start the cable system self-balancing counterforce frame and T-shaped wind-fire coupling simulation system, pre-load and collect sensor feedback data;

[0104] Use the least squares method to calculate the initial time delay Δt of each system i , input the LSTM-PID intelligent control system for feedforward compensation;

[0105] Step four, generate the temperature rise curve of the fire scene under the selected mode and perform real-time hybrid test:

[0106] Step 4.1, input the fire source power, position and wind speed into the fire temperature field fast generation model to generate a non-uniform temperature field, and transmit it to the LSTM-PID intelligent control system to analyze the T-shaped wind-fire coupling simulation system burner valve opening and fan frequency command;

[0107] If mode B is executed, input the standard target temperature rise curve, and output the burner valve opening command synchronously;

[0108] Step 4.2, start gas supply, dynamically adjust gas flow through the LSTM-PID intelligent control system, and synchronize the ignition flame;

[0109] If mode A is executed, start the multi-dimensional wind field simulation system synchronously and dynamically adjust the fan inclination angle;

[0110] Step 4.3, based on the LSTM numerical substructure fast calculation model, predict the full bridge displacement and cable force in real time, and convert it into the loading command of the first and second through-type jacks;

[0111] Step 4.4, load step by step according to the displacement / force control mode, and use the force control mode to slowly stretch to the specimen fracture when exceeding the preset limit;

[0112] Step 4.5, during the entire loading process, based on the LSTM time delay compensator, predict the time delay Δt i of the next analysis step Δt i+1 , and advance Δt i+1Send instructions to ensure that the force loading, combustion and wind field synchronization, wherein i=(1, 2, n);If mode B is executed, only predict the load and temperature time delay, ensure that the load and combustion are synchronized;

[0113] Step five, data acquisition and dynamic visualization, specifically including:

[0114] Step 5.1, real-time acquisition of specimen axial force, displacement, strain and fire temperature data, and drawing related curves in the LSTM-PID intelligent control system;

[0115] If mode A is executed, the fan speed data is synchronously collected;

[0116] Step 5.2, bind the sensor measured data and the predicted data to each grid node of the full structure three-dimensional graph through the three-dimensional animation software, and dynamically visualize the full bridge stress, deformation and temperature field distribution based on the numerical-color mapping algorithm;

[0117] Step 5.3, generate load-displacement curve and temperature-time curve, output damage evolution law and failure warning.

[0118] Compared with the prior art, the beneficial effects of the present application are:

[0119] 1, high-fidelity fire scene reproduction: the cooperative work of the horizontal linear combustion module and the vertical ring combustion module can realize the dynamic envelope of the flame in the three-dimensional space, simulate the thermal shock and ablation process of the flame on the bridge cable under the action of the environmental wind in the real fire, and avoid the difficulty that the flame height of the existing combustion equipment is difficult to break through 1.5m, and the horizontal linear combustion module and the vertical ring combustion module combined combustion equipment are designed, which can simulate the T-shaped flame envelope combustion scene of the main cable and the suspender of the suspension bridge under fire at the same time, truly restore the natural three-dimensional gradient flame licking wrapping effect of the suspension bridge cable system under the influence of the environmental wind in the fire accident, and accurately reproduce the key characteristics of the non-uniform temperature distribution in the real fire.

[0120] 2. Real boundary reconstruction loading of kiloton large-scale cable system: The free-rotating force boundary subsystem and the free-rotating anchoring subsystem rotate freely in the sliding beam through the variable cross-section force transmission main shaft, cooperating with the free sliding of the sliding beam, to realize multi-scene simulation of cable loading: arbitrary angle hovering tensioning when the main cable model is loaded alone; arbitrary angle hovering and multi-directional composite tensioning when the main cable model and the sling model are combined; vertical tensioning under the real boundary of the sling model. In addition, through the physical design of the self-balancing counterforce frame of the cable system, the composite anchor ring and the variable cross-section force transmission main shaft, the force transmission path and the stress capacity of the existing counterforce frame are optimized, which overturns the current situation that the loading capacity of the cable system is only up to hundreds of tons and the loading diameter is limited to ≤100 mm, so that the loading capacity of the whole loading system jumps to thousands of tons and the loading diameter breaks through 500 mm.

[0121] 3. Breakthrough in computing speed bottleneck: The traditional finite element method has low computing efficiency under fire scene, which is difficult to meet the demand of real-time hybrid test. The present application introduces a multi-scale gated residual neural network based on LSTM, which significantly improves the computing speed, realizes second-level thermal coupling interaction, and solves the problem of significant error accumulation of traditional methods under long-period loading. The fast generation model of fire temperature field is based on improved generative adversarial network (GAN), which can generate high-precision non-uniform temperature field distribution in real time, further improving the real-time and accuracy of the test.

[0122] 4. Real-time dynamic compensation and time delay processing: The time delay compensation mechanism can dynamically predict and compensate the time delay between the control command and the sensor measurement value through the time delay prediction model based on LSTM, ensuring the synchronization and accuracy of the test process. This mechanism can gradually reduce the time delay error until the time delay error is less than 1 second throughout the test period. During the test process, through online incremental learning and real-time data feedback, the time delay prediction model can dynamically adjust the weights to adapt to the changes in the fire scene, further improving the prediction accuracy.

[0123] 5. High-precision real-time hybrid test: Through the multi-scale gated residual LSTM network, the boundary displacement and cable force of the whole bridge are predicted in real time, and the specimen boundary force or displacement is taken as the loading instruction to realize second-level synchronization of physical loading and numerical prediction. Combined with LSTM feedforward correction and LSTM-PID intelligent control system feedback regulation, the load of the through-type jack, the gas flow of the burner and the wind speed / angle of the fan are dynamically adjusted to ensure the synchronization of loading, burning and wind field regulation. The time delay error between the control instruction and the sensor response is predicted in real time by the time delay compensator based on LSTM, and the instruction is sent in advance through feedforward compensation to ensure the synchronization of each system.

[0124] 6. Wide versatility: The platform supports independent loading of any component of the cable system, including main cables, stay cables, suspenders, and hangers, and can perform open-space custom fire tests and standard fire tests in enclosed spaces with a built-in fireproof enclosure. By adding a modular fireproof enclosure, the system can quickly switch between open-space and enclosed-space fire tests, adapting to different types of fire tests.

[0125] 7. Efficient data acquisition and analysis: The system integrates pressure sensors, infrared thermal imagers, distributed thermocouple arrays, and ultrasonic anemometers to synchronously collect axial displacement, strain / stress data of the main cable model, and fire temperature / wind speed data in real time. Through the Unity3D engine, the system dynamically visualizes the full-bridge stress, deformation, and temperature field distribution, synchronously generates load-displacement curves, temperature-time curves, and failure warning threshold panels, and outputs damage evolution laws, providing intuitive evidence for analysis.

[0126] 8. Visualization and real-time monitoring: The system binds LSTM prediction data or sensor measured data with three-dimensional model grid nodes, dynamically visualizes the stress, deformation, and temperature field distribution changes of the full-bridge under fire, and provides intuitive visual support for the test process. At the same time, the system can dynamically update load-displacement curves, temperature-time curves, and failure warning threshold panels, facilitating real-time monitoring of the test state.

[0127] 9. Improved test efficiency and data reliability: The physical loading and transmission system supports fast positioning and low-friction convenient installation, improving test preparation efficiency. Multi-source data can be transmitted to the intelligent control and calculation system in real time to generate collaborative analysis curves, improving data processing and analysis efficiency. The system supports full-scale model tests of large-diameter cables, ensuring that test results reflect the nonlinear response and collaborative work characteristics in real engineering.

[0128] 10. Safety and flexibility: The systems are designed with modularity, facilitating transportation, fixation, and rapid reorganization, improving test flexibility and efficiency. The system is equipped with explosion-proof fire extinguishing devices to ensure safety during the test and prevent industrial explosion and fire risks. The two skid-mounted units use standardized interface design to seamlessly transmit fuel, power, wind, and data, facilitating system integration and expansion.

[0129] 11. Engineering application value: The system provides a high-fidelity test platform and analysis method for large-span bridge fire resistance design, assisting in the development of bridge fire resistance design, disaster warning, and repair strategies. By accurately analyzing the cable failure mode, critical temperature threshold, and damage evolution law under fire, the system provides direct data support for the selection of fireproof coatings or structural measures, the development of fire emergency evacuation strategies, and the design of repair schemes. The system has irreplaceable engineering value in improving bridge fire resistance design specifications, reducing the risk of the whole life cycle, and assisting in the safety construction and long-term operation of major projects in China. BRIEF DESCRIPTION OF DRAWINGS

[0130] The accompanying drawings are included to provide a further understanding of the application and are incorporated in and constitute a part of this specification, illustrate embodiments of the application and are meant to explain the present application and are not intended to limit the application.

[0131] Figure 1 A structure schematic view of a large-span bridge substructure vehicle-fire wind force real-time hybrid test platform is provided for the embodiments of the present application;

[0132] Figure 2 A structure schematic view of a cross beam electric lifting locking system is provided for the embodiments of the present application;

[0133] Figure 3 A structure schematic view of an adjustable roller bearing platform is provided for the embodiments of the present application;

[0134] Figure 4 A structure schematic view of a composite anchor ring and variable cross-section force transmission main shaft is provided for the embodiments of the present application;

[0135] Figure 5 A structure schematic view of a sling force loading anchoring subsystem is provided for the embodiments of the present application;

[0136] Figure 6 A structure schematic view of a free rotation force boundary subsystem is provided for the embodiments of the present application;

[0137] Figure 7 A structure schematic view of a main cable and sling composite model is provided for the embodiments of the present application;

[0138] Figure 8 A connection schematic view of a modularized pry-mounted integrated system and a high dynamic axial flow fan is provided for the embodiments of the present application;

[0139] Figure 9 A structure schematic view of a frame type combustor is provided for the embodiments of the present application;

[0140] Figure 10 A structure schematic view of a frame type combustor is provided for the embodiments of the present application; Figure 9 An enlarged schematic view of A in the middle;

[0141] Figure 11 A structure schematic view of a vertical annular combustion module is provided for the embodiments of the present application;

[0142] Figure 12 A structure schematic view of a first pry-mounted unit is provided for the embodiments of the present application;

[0143] Figure 13 A structure schematic view of a second pry-mounted unit is provided for the embodiments of the present application;

[0144] Figure 14A structural schematic diagram of a gas pipeline provided by an embodiment of the present application is provided;

[0145] Figure 15 A multi-scale gated residual LSTM neural network architecture diagram provided by an embodiment of the present application is provided;

[0146] Figure 16 A GAN-based fire temperature field fast generation model neural network architecture diagram provided by an embodiment of the present application is provided;

[0147] Figure 17 An LSTM neural network architecture diagram in an LSTM-PID intelligent control system provided by an embodiment of the present application is provided;

[0148] Figure 18 A neural network architecture diagram of an LSTM-based time delay compensator is provided.

[0149] In the figure: 1-beam electric lifting locking system; 2-roller bearing transmission system; 3-main cable and sling composite model; 4-cable system self-balancing counterforce frame; 5-free rotation force boundary subsystem; 6-free rotation anchoring subsystem; 7-sling force loading anchoring subsystem;

[0150] 11-limiting sliding rail; 12-array type positioning hole group; 13-screw linkage synchronous lifting subsystem; 131-base; 132-screw rod; 133-motor; 134-transmission shaft; 135-reducer; 136-elevator;

[0151] 21-adjustable roller bearing platform; 22-roller group; 211-base frame; 212-top bearing beam; 213-lifting rod; 214-knob; 215-support column; 23-concave clamping groove;

[0152] 31-main cable model; 32-sling model; 33-cable clamp;

[0153] 41-main cross beam; 42-stand; 43-fixed hole; 44-upper cross beam; 45-base; 46-box type bottom beam; 47-sliding cross beam; 48-fixed seat;

[0154] 51-composite anchor ring; 52-square pyramid force transmission section; 53-cylindrical rotating section; 54-first through-type jack; 55-loading end primary anchor; 56-loading end secondary anchor;

[0155] 61-anchoring end anchor;

[0156] 71-second through-type jack; 72-T-shaped slide rail; 73-slide rail limiter; 74-sling loading end anchor; 75-separable half-ring group; 76-pole lever; 8-frame type burner;

[0157] 101-high dynamic axial flow fan; 102-frequency converter; 103-gimbal locking roller set; 201-gas collecting pipe; 202-short section stainless steel flame spraying pipe; 203-gas input interface; 204-self-suction air supply interface; 205-high-energy pulse electric spark igniter; 206-annular burner; 207-flame injection port; 208-double-inlet branch pipe; 209-hollow sphere; 210-hemispherical socket; 2111-flange quick-release interface; 2112-dismountable support; 2113-variable-diameter transition pipe; 2114-multi-way equal-branch pipe; 2115-arc-shaped flow guide baffle; 2116-inverted v-shaped diffusion cutout; 2117-air flow equalization plate; 301-gas cylinder or fuel tank; 302-vaporization furnace; 303-explosion-proof control cabinet; 304-explosion-proof fan; 305-combustible gas detector; 306-suspended ultra-fine dry powder fire extinguisher; 307-container; 308-explosion-proof louver; 309-fire extinguisher annular hoisting support; 310-touch panel; 311-LED explosion-proof lamp set; 312-explosion-proof junction box; 401-distribution cabinet; 402-control cabinet; 403-explosion-proof lighting device two; 501-main pipeline; 502-gas main shut-off valve; 503-intelligent pressure reducing valve; 504-intelligent valve positioner; 505-filter; 506-corrugated flame arrestor. DETAILED DESCRIPTION

[0158] In order to make the purpose, technical scheme and advantages of the present application clearer, the present application will be further described in detail below in combination with the drawings and examples. Of course, the specific examples described herein are only used to explain the present application, and are not used to limit the present application.

[0159] The present application provides a large-span bridge substructure vehicle-fire wind force real-time hybrid test platform, comprising:

[0160] a physical loading and transmission system, a T-shaped wind-fire coupling simulation system, a modularized skid-mounted integrated system, an intelligent control and calculation system, and a multi-dimensional high-precision monitoring system;

[0161] The physical loading and transmission system comprises a cable system self-balancing counterforce frame 4, a beam electric lifting locking system 1, and a roller bearing transmission system 2.

[0162] The T-shaped wind-fire coupling simulation system comprises a T-shaped global flame envelope combustion system and a multi-dimensional wind field simulation system. The T-shaped global flame envelope combustion system comprises a horizontal linear combustion module and a vertical annular combustion module.

[0163] The intelligent control and calculation system comprises a fire temperature field rapid generation model, an LSTM-based numerical substructure rapid calculation model, an LSTM-PID intelligent control system, and an LSTM time lag compensator.

[0164] The multi-dimensional high-precision monitoring system comprises a fire scene panoramic video acquisition unit, a fire scene space temperature field monitoring unit and a wind speed monitoring unit, and is used for monitoring flame shape, temperature field and wind field data.

[0165] The LSTM-PID intelligent control system fuses flame shape, temperature field and wind field data in real time through an LSTM model well known to those skilled in the art, dynamically optimizes gas flow and complex wind field conditions, and reproduces the three-dimensional dynamic flame envelope combustion state and non-uniform temperature field in a real fire scene.

[0166] In the embodiment, the physical loading and transmission system is suitable for multipurpose loading test of a cable system substructure with a large tonnage of 800-1000t and a large size of 500-600mm.

[0167] Referring to Figures 1-7 , the cable system stress loading system comprises a cable system self-balancing counterforce frame 4, a free-rotation force boundary subsystem 5, a free-rotation anchoring subsystem 6 and a sling force loading anchoring subsystem 7.

[0168] The cable system self-balancing counterforce frame 4 comprises two groups of symmetrically arranged rectangular lattice type frames, each group of rectangular lattice type frame comprises four steel box columns 42, is distributed in a rectangular shape, and a support is arranged between adjacent columns 42; an upper cross beam 44 is mounted on the top of the column 42 through a bolt group. The top of the two groups of rectangular lattice type frames is connected through two main cross beams 41. The bottom of each group of rectangular lattice type frame is provided with an integral base 45 formed by a box type steel platform, and the two bases 45 are connected through a box type bottom beam 46, and the height of the base 45 is lower than that of the box type bottom beam 46. The middle part of each group of rectangular lattice type frame is provided with a sliding cross beam 47, which can freely slide and lock in the height direction of the rectangular lattice type frame under the drive of the cross beam electric lifting locking system 1.

[0169] The free-rotation force boundary subsystem 5 and the free-rotation anchoring subsystem 6 are used for anchoring and loading the main cable model 31 or the stay cable model, and each of the free-rotation force boundary subsystem 5 and the free-rotation anchoring subsystem 6 comprises a composite anchor ring 51. Figure 4As shown, the composite anchor ring 51 is a square frame outside and a circular through hole inside. The composite anchor ring 51 is fixedly installed with variable cross-section force transmission main shafts on both sides; the variable cross-section force transmission main shafts include four-pyramid force transmission sections 52 welded with the composite anchor ring 51 and cylindrical rotation sections 53 welded with the four-pyramid force transmission sections 52, the four-pyramid force transmission sections 52 are uniformly distributed with radial stiffening ribs on the outer surface, the cylindrical rotation sections 53 pass through the sliding cross beam 47 and are freely rotatable around the axis of the cylindrical rotation sections 53 in the sliding cross beam 47 through the sleeve or bearing structure, so as to drive the composite anchor ring 51 to freely rotate in the sliding cross beam 47, and the free sliding of the sliding cross beam 47 can realize the suspension tensioning at any angle during the loading of the main cable model. The free rotation force boundary subsystem 5 and the free rotation anchoring subsystem 6 are both designed with the variable cross-section force transmission main shafts and the composite anchor ring 51, and can realize the free rotation and anchoring of the main cable model during the loading process, so as to simulate the boundary constraint under the actual working condition. The design of the variable cross-section force transmission main shafts can adaptively transmit the force during the loading process, and improves the flexibility and adaptability of the test device.

[0170] As shown in the figure, Figure 6 The free rotation force boundary subsystem 5 further includes a first through jack 54 and a loading end double-stage gradient anchoring unit. The first through jack 54 is fixedly installed in the circular through hole inside the composite anchor ring 51 through the flange, and the inner diameter of the circular through hole matches the outer diameter of the first through jack 54. The loading end double-stage gradient anchoring unit includes a loading end primary anchor 55 and a loading end secondary anchor 56, which are separately arranged on both sides of the first through jack 54 to anchor the main cable model 31 or the cable-stayed cable model. The outer diameter of the loading end primary anchor 55 is smaller than the inner diameter of the first through jack 54 and the inner diameter of the composite anchor ring 51 in the free rotation anchoring subsystem 6, and the outer diameter of the loading end secondary anchor 56 is 1.2-1.5 times the outer diameter of the upper flange of the first through jack 54. The steel wire bundle of the main cable model 31 or the cable-stayed cable model extends outward through the loading end primary anchor 55, and after passing through the inner cavity of the first through jack 54, the steel wire bundle is finally anchored through the loading end secondary anchor 56. The first through jack 54 drives the loading end secondary anchor 56 to realize the axial force loading of the main cable model 31 or the cable-stayed cable model, and the variable cross-section force transmission main shaft can freely rotate around the axis during the loading process. The design of the loading end double-stage gradient anchoring unit improves the reliability and safety of anchoring through the cooperation of the loading end primary anchor 55 and the loading end secondary anchor 56. The first through jack 54 drives the loading end secondary anchor 56 to realize the axial force loading of the main cable model 31 or the cable-stayed cable model, which can accurately control the size of the loading force and ensure the accuracy of the test.

[0171] The free rotation anchoring subsystem 6 further includes an anchoring end anchor 61 to anchor the anchoring end of the main cable model 31 or the cable-stayed cable model. The outer diameter of the anchoring end anchor 61 is 1.2 times the inner diameter of the composite anchor ring 51 in the free rotation anchoring subsystem 6.

[0172] like Figure 5 As shown, the cable force loading and anchoring subsystem 7 includes a second through-type jack 71, a cable loading end anchor 74, a locking and anchoring mechanism, a T-shaped slide rail 72, and a slide rail limiter 73. The second through-type jack 71 is embedded in the internal cavity of the box-shaped bottom beam 46, and the cable loading end anchor 74 is anchored to the end of the cable model 32. The inner diameter of the second through-type jack 71 is 1.1-1.2 times the outer diameter of the cable loading end anchor 74. The locking and anchoring mechanism is located between the second through-type jack 71 and the cable loading end anchor 74. Its inner diameter is equal to the diameter of the cable model 32, and its outer diameter is larger than the outer diameter of the second through-type jack 71. The locking and anchoring mechanism enables the self-locking and release of the cable model 32, ensuring the stability of the cable model 32 during loading and rapid release after unloading.

[0173] The locking and anchoring mechanism is a separable semi-ring assembly 75 connected by a lever 76 and located inside the box-shaped bottom beam 46. The lever 76 drives the separable semi-ring assembly 75 to open and close radially, achieving a two-way function of release and self-locking. After the sling loading end anchor 74 anchors the sling model 32, it passes entirely through the second through-hole jack 71. By moving the levers 76 on both sides of the opening above the box-shaped bottom beam 46, the separable semi-ring assembly 75 is closed. The annular locking surface formed by the separable semi-ring assembly 75 is engaged between the sling loading end anchor 74 and the second through-hole jack 71, preventing the loading end of the sling model 32 from being pulled out from inside the second through-hole jack 71. The second through-hole jack 71 can then load the sling model 32. By moving the lever 76 in the opposite direction, the separable semi-ring assembly 75 is separated, the loading end of the sling model 32 is released, and the sling model 32 can be freely pulled away axially. The locking and anchoring mechanism, driven by lever 76, radially opens and closes the separable semi-ring assembly 75, achieving a two-way function of releasing and self-locking the sling model 32, thus improving the reliability of the sling model 32's anchoring and the flexibility of its operation. This locking and anchoring mechanism design makes the replacement and adaptation of the sling model 32 more convenient, improving the versatility of the testing apparatus. In practice, the locking and anchoring mechanism can be manually opened and closed by entering the box-shaped bottom beam 46. When the separable semi-ring assembly 75 in the locking and anchoring mechanism is closed, it is connected and fixed by a conventional snap-fit ​​device located between the two.

[0174] The T-shaped slide rail 72 is fixedly installed on the box-shaped bottom beam 46, and the slide rail limiters 73 are provided with two and symmetrically installed at two ends of the T-shaped slide rail 72. The slide rail limiters 73 can move on the T-shaped slide rail 72 and are rigidly fixed by bolts. The second through-type jack 71 is embedded in the internal cavity of the box-shaped bottom beam 46. According to the need, the position of the second through-type jack 71 is adjusted by moving horizontally on the T-shaped slide rail 72, and then the second through-type jack 71 is rigidly connected with the two slide rail limiters 73 by bolts, and the slide rail limiters 73 are rigidly connected with the T-shaped slide rail 72 by bolts, so as to limit the horizontal position of the second through-type jack 71 on the T-shaped slide rail 72. The T-shaped slide rail 72 and the slide rail limiters 73 are arranged to facilitate the adjustment and fixation of the second through-type jack 71, improve the stability of the loading of the sling model 32 and the convenience of operation; and the horizontal position of the second through-type jack 71 is adjusted to realize the combined tensioning of the main cable and the sling composite model 3 at any angle.

[0175] The primary anchor 55 at the loading end, the secondary anchor 56 at the loading end, the anchor 61 at the anchoring end and the anchor 74 at the loading end of the sling are the same in structure and each include an anchor ring and a conical clamping piece group. The anchor ring is uniformly provided with a plurality of conical through holes in the circumferential direction, and the conical clamping piece group is composed of two equal division clamping pieces to form a complete conical ring. The outer conical surface of the clamping piece matches the inner conical surface of the conical through hole of the anchor ring, and the inner wall of the clamping piece is provided with a spiral friction groove. The force-bearing steel wire in the main cable model passes through the conical clamping piece group and the conical through hole of the anchor ring. When loaded, the force-bearing steel wire drives the conical clamping piece group to displace towards the small end of the conical through hole, the clamping piece is radially extruded and pressed against the conical through hole of the anchor ring, the spiral friction groove forms mechanical engagement with the surface of the force-bearing steel wire, and through the cooperative action of the anchor ring and the conical clamping piece group, the self-locking nonlinear anchoring of the force-bearing steel wire is realized, and the safety and reliability of the test are improved.

[0176] As Figure 2As shown, the cross beam electric lifting locking system 1 is provided with two groups corresponding to the sliding cross beam 47, each group of cross beam electric lifting locking system 1 includes a screw linkage synchronous lifting subsystem 13, a limiting sliding track 11 and an array type positioning hole group 12. The screw linkage synchronous lifting subsystem 13 includes a screw rod 132, a lifting machine 136, a motor 133 and a base 131; the screw rod 132 is symmetrically provided with two, one end of each screw rod 132 is rotatably connected with the sliding cross beam 47 through a flange, and the other end is connected with the lifting machine 136. The lifting machine 136 is provided with two corresponding to the screw rod 132, and the motor 133 drives two lifting machines 136 through a transmission shaft 134, to control the stable lifting of the sliding cross beam 47. Specifically, the motor 133 output end is connected with a speed reducer 135, to transmit power to the speed reducer 135, and the output ends of the speed reducer 135 are connected with the turbine boxes in the two lifting machines 136 through two transmission shafts 134, respectively. The motor 133 rotates to drive the turbine box to drive the screw rod 132 to move linearly, thereby driving the sliding cross beam 47 to stably lift, improving the stability of the test device and the convenience of operation. The base 131 is fixedly installed on the upper cross beam 44 of the rectangular lattice frame, and the lifting machine 136 and the motor 133 are fixedly installed on the base 131. The limiting sliding track 11 is integrated on the inner side of the middle column 42 of the rectangular lattice frame, and the cross section is T-shaped, forming a sliding pair with the slider at the end of the sliding cross beam 47. The array type positioning hole group 12 is distributed on the column 42 in the height direction with an interval of 50-100mm, a fixed seat 48 is fixedly installed on the sliding cross beam 47, a fixed hole 43 for positioning is arranged on the fixed seat 48, a detachable bolt mechanism is installed in the fixed hole 43, and the array type positioning hole group 12 cooperates with the detachable bolt mechanism to lock the position of the sliding cross beam 47. The cooperation of the limiting sliding track 11 and the array type positioning hole group 12 can accurately control the position of the sliding cross beam 47, and the position is locked through the detachable bolt mechanism, to ensure the stability of the device during the test.

[0177] Referring to Figure 3 , the roller bearing transmission system 2 includes an adjustable roller bearing platform 21 installed on the upper end surface of the base 45, the adjustable roller bearing platform 21 includes a base frame 211, a lifting mechanism and a top bearing beam 212; the lifting mechanism is arranged between the base frame 211 and the top bearing beam 212, the top bearing beam 212 moves up and down under the drive of the lifting mechanism, and the top bearing beam 212 is equal in height to the box type bottom beam 46 after being lifted; a plurality of groups of roller groups 22 are arranged at equal intervals on the adjustable roller bearing platform 21 and the box type bottom beam 46, the top bearing beam 212 and the box type bottom beam 46 are both provided with recessed clamping grooves for installing the roller groups 22, and the height of the base 45 is lower than that of the box type bottom beam 46.

[0178] The lifting mechanism can select any one of the existing technologies capable of achieving this function. In the embodiment, the lifting mechanism includes four groups of support columns 215 and lifting rods 213, which are respectively arranged at the four corners of the base frame 211; the support column 215 is a rectangular steel column fixedly installed on the base frame 211, the column body is hollow, the lifting rod 213 is a rack, one end of which is arranged in the column body, and the other end is fixedly connected with the top bearing beam 212, a gear cooperating with the rack is further arranged in the column body, and a knob 214 for driving the gear is arranged on the surface of the support column 215, and the rotation of the knob 214 can realize the lifting of the top bearing beam 212. The roller group 22 is in the shape of a sandglass, is installed on the adjustable roller bearing platform 21 and the box-shaped bottom beam 46, is a plurality of groups arranged equidistantly along the length direction of the main cable model 31, and each roller group 22 includes two rollers. The top bearing beam 212 and the box-shaped bottom beam 46 are each provided with a recessed clamping groove 23 for installing the roller group 22, and the width of the recessed clamping groove 23 matches the diameter of the roller shaft. The design of the sandglass-shaped roller group 22 can stably support and drive the main cable model, and ensure the stable movement of the main cable model in the test process.

[0179] The embodiment also provides a method for loading test using the above-mentioned large-span bridge cable system substructure multipurpose loading test device, which is suitable for the single tensioning of the main cable model 31 or the stay rope model 32, the composite tensioning of the main cable and the stay rope composite model 3 at any angle, the suspension tensioning of the stay cable model at any angle, and the vertical tensioning test of the stay rope model 32 of the stay rope arch bridge.

[0180] In the embodiment, the main cable model 31 selects the large-diameter cable test model disclosed in the Chinese invention patent with the publication number CN120064553A, a large-diameter cable test model true fire test device and test method. The stay rope model 32 directly selects a real suspender or stay rope, without the need for additional manufacturing.

[0181] The installation method of the single main cable model 31 or the stay cable model is as follows: first, the first through-type jack 54 is embedded and fixed in the composite anchor ring 51 of the free-rotation force boundary subsystem 5, the force sensor is installed at the front end of the first through-type jack 54, the force sensor is connected with the load display recorder, and the load display recorder is used for monitoring the axial force size applied by the first through-type jack 54. The loading end primary anchor 55 is installed at the loading end of the main cable model 31 or the stay cable model, after the strands pass through the loading end primary anchor 55, the anchor cable stress meter is installed on each strand, and the anchor cable stress meter is used for monitoring the stress change of each strand of the main cable model 31 or the stay cable model in the test process; and the anchoring end anchor 61 is installed at the anchoring end of the main cable model 31 or the stay cable model.

[0182] Then, the main cable model 31 or the stay cable model is hoisted onto the platform of the roller support transmission system 2. The loading end of the main cable model 31 or the stay cable model is oriented to the side of the free rotation anchoring subsystem 6. The main cable model 31 or the stay cable model is pulled, and with the assistance of multiple roller groups 22, the loading end of the main cable model 31 or the stay cable model passes through the composite anchor ring 51 of the free rotation anchoring subsystem 6 and the first through-hole jack 54 in the free rotation force boundary subsystem 5 in sequence. The loading end cable strand passes through the loading end of the first through-hole jack 54 to install the loading end secondary anchor 56.

[0183] Next, activate the electric lifting and locking system 1 for the crossbeam, raising the sliding crossbeam 47 to any preset height, and adjust the height of the sliding crossbeam 47 to perform suspension tensioning at any angle for the main cable model 31 or the stay cable model. Finally, remove the roller support transmission system 2.

[0184] The installation method of the single sling model 32 is as follows: First, a second through-hole jack 71 is embedded in the cavity inside the box-shaped bottom beam 46, and a force sensor is installed at the front end of the second through-hole jack 71. The force sensor is connected to a load display recorder to monitor the magnitude of the axial force applied by the second through-hole jack 71.

[0185] Then, install a main cable model 31 as a loading auxiliary for fixing the upper end of the sling model 32 according to the installation method of the main cable model 31; since the main cable model 31 is only an auxiliary for the sling model 32, its stress condition is not a concern, so the force sensor at the front end of the first through-hole jack 54 and the strain gauge and anchor stress gauge on the main cable model 31 can be left uninstalled.

[0186] Next, the sling model 32 and the main cable model 31 are connected by the cable clamp 33. An anchor stress gauge and sling loading end anchor 74 are installed at the loading end of the sling model 32. Finally, the sling loading end anchor 74 extends into the second through-hole jack 71 and is locked by the locking anchoring mechanism.

[0187] like Figure 7 The installation method of the main cable and sling composite model 3 shown is as follows: Install the main cable model 31 and sling model 32 in sequence according to the installation method of the main cable model 31 and sling model 32, and adjust the height of the sliding beam 47 and the horizontal position of the second through-type jack 71 to perform multi-directional joint tensioning of the main cable and sling composite model 3 at any angle.

[0188] S1. Install displacement gauges at preset positions on the main cable model, such as the mid-span of the main cable model 31 or the end of the sling model 32.

[0189] S2, control the corresponding centering jack according to the displacement control mode, each load increment is 5% of the designed breaking force, and the model stress reaches the minimum breaking force, and then the force control mode is switched to 5-10kN per level to load to the model fracture.

[0190] S3, the collected axial force, displacement and stress and strain data of the main cable model are transmitted into the computer terminal in real time, and the strain and stress state of the main cable model is monitored and drawn throughout.

[0191] S4, replace the loading end anchor, the anchoring end anchor 61 and the sling loading end anchor 74, adapt to different main cable models, and realize different loads and constraints on different cable systems.

[0192] Referring to Figures 8-14 In the embodiment, the T-shaped wind-fire coupling simulation system comprises a T-shaped global flame envelope combustion system and a multi-dimensional wind field simulation system, the T-shaped global flame envelope combustion system comprises a horizontal linear combustion module and a vertical ring-shaped combustion module; the T-shaped global flame envelope combustion system is used to simulate a T-shaped dynamic thermal erosion three-dimensional wrapped combustion mode, and the multi-dimensional wind field simulation system is used to simulate complex wind field conditions; the modularized skid-mounted integrated system is used to provide flow-adjustable fuel gas for the T-shaped global flame envelope combustion system; the T-shaped wind-fire coupling simulation system fuses flame mode, temperature field and wind field data in real time through an LSTM-PID intelligent control system, dynamically optimizes and adjusts the proportion of fuel gas and air and complex wind field conditions, and reproduces a three-dimensional dynamic flame envelope combustion state and a non-uniform temperature field in a real fire scene.

[0193] The T-shaped global flame envelope combustion system comprises a horizontal linear combustion module and a vertical ring-shaped combustion module, the horizontal linear combustion module comprises a plurality of independent linearly arranged frame burners 8, the frame burners are suspended and fixed in the middle test area of the self-balancing counterforce frame of the cable system through a plurality of steel wire ropes, one side wall of the frame burner is provided with a fuel gas input interface 203 and a self-suction air supply interface 204 at the bottom, the top of the frame burner is provided with a multidirectional flame jet unit capable of forming a three-dimensional flame gradient mode, the multidirectional flame jet unit comprises a plurality of gas collecting pipes 201 arranged side by side, a plurality of flame jet pipes are uniformly distributed along the length direction of each row of gas collecting pipes 201, the flame jet pipes are short section type stainless steel flame jet pipes 202 in the embodiment, and the short section type stainless steel flame jet pipes 202 are vertically arrayed, a high-energy pulse electric spark igniter 205 is arranged at the outer edge intersection node of each row of gas collecting pipes 201, and the high-energy pulse electric spark igniter 205 is synchronously controlled by an LSTM-PID intelligent control system synchronous control circuit to realize synchronous ignition of the flame.

[0194] The middle part of the gas collecting pipe 201 is communicated with the gas input interface 203 through a plurality of equal distribution branch pipes 2114, the upper part of the plurality of equal distribution branch pipes 2114 extends into the gas collecting pipe 201 and is provided with a reverse V-shaped diffusion cutout 2116, the inside of the frame type combustor is provided with a flow equalization assembly, air enters into the frame type combustor from the self-suction air supply interface 204 of the frame type combustor, and the flow equalization assembly forms a flow equalization air supply multi-directional flame injection unit; the reverse V-shaped diffusion cutout is used for realizing uniform decreasing diffusion of the gas from the middle part of the gas collecting pipe to both ends, cooperating with the flow equalization air to realize synchronous ignition of each gas collecting pipe and reproduce a real three-dimensional flame natural gradient form, and a real fire space non-uniform temperature field is presented;

[0195] The gas input interface 203 is independently and flow-adjustably supplied with gas through a modularized pry integrated system, and the self-suction air supply interface 204 realizes natural air supply through a negative pressure self-suction principle; the flow equalization assembly comprises an air flow equalization plate 2117 and two arc-shaped flow guide partitions 2115 arranged along the air flow direction, the self-suction air supply interface 204 forms a rectangular gradually expanding flow channel in the inside of the frame type combustor, the two arc-shaped flow guide partitions arranged along the air flow direction uniformly divide the air inlet of the frame type combustor into three left, middle and right paths to initially divide the air; the air flow equalization plate is arranged at the lower part of the multi-directional flame injection unit to secondarily equalize the air; the universal locking roller set 103 is installed at the bottom of the frame type combustor to realize free and convenient movement of the equipment.

[0196] The vertical annular combustion module comprises a plurality of layers of longitudinally connected detachable annular combustors 206, the main body of the annular combustor 206 is an annular steel pipe, adjacent annular combustors are connected through detachable support bodies 2112, each layer of annular combustor 206 uniformly distributes a plurality of flame injection ports 207 capable of adjusting a flame injection angle, the flame injection ports 207 are supplied with gas through the symmetrical double gas inlet branch pipes 208 arranged on the annular combustor 206, a high-energy pulse electric spark igniter 205 is arranged at each of the two symmetrically distributed flame injection ports in each layer of annular combustor 206, and each layer of flame is synchronously ignited through an LSTM-PID intelligent control system synchronous control circuit, and each double gas inlet branch pipe 208 is independently and flow-adjustably supplied with gas through a movable pry integrated system.

[0197] The fire spout 207 comprises an air inlet end short pipe and a fire spout end short pipe, which are connected through an open hollow hemispherical universal hinge mechanism, the universal hinge mechanism comprises a hollow ball with a flow guide channel and a hemispherical socket matched with the hollow ball, which can realize 0°-90° inclination angle adjustment of the fire spout end short pipe, the air inlet end short pipe is communicated with the annular combustor 206 through a variable diameter transition pipe 2113, the small diameter end of the variable diameter transition pipe 2113 is communicated with the air inlet end short pipe, after the gas enters the annular combustor 206, it is accelerated into the universal hinge mechanism through the variable diameter transition pipe 2113, and is spouted from the fire spout end short pipe and ignited, forming a flame with adjustable spouting direction.

[0198] The annular combustor 206 is integrated with a flange type quick release interface 2111 along its symmetry axis direction, the flange type quick release interface 2111 does not interfere with the fire spout 207 and the double air inlet branch pipe 208, the flange type quick release interface 2111 realizes axial separation and recombination of the annular combustor through bolt connection, after disassembly, two half annular combustor units can be formed, and each half annular combustor unit retains independent gas supply passage and ignition control, supporting true fire loading in half circle fire scene; the detachable support 2112 is configured with an internal thread quick connector at both ends for connecting adjacent two layers of annular combustors, by replacing the detachable support 2112 with different heights, the height of the vertical annular combustion module can be adjusted.

[0199] The modularized pry integrated system comprises a first pry unit and a second pry unit; a gas cylinder or fuel tank 301, a vaporization furnace 302, an explosion-proof fire extinguishing device, an explosion-proof lighting device one and a gas pipeline are integrated in the first pry unit, the gas cylinder or fuel tank 301 connected with the vaporization furnace 302 converts liquid fuel into gaseous fuel, the gas pipeline comprises a high-pressure pipeline, a valve and a safety device, realizing stability and safety of fuel delivery, the explosion-proof fire extinguishing device ensures safe ventilation in the first pry unit, preventing industrial explosion and timely extinguishing fire; a power distribution cabinet 401, a control cabinet 402, a frequency converter 102 and an explosion-proof lighting device two 403 are integrated in the second pry unit, the explosion-proof lighting device two 403 is electrically connected with the control cabinet 402, the power distribution cabinet 401 is used for providing power supply for the whole device, the control cabinet 402 is built-in with a PLC controller and a communication module, which is used for realizing coordinated control of each system; the two pry units adopt standardized interface design, can be quickly connected and realize seamless transmission of fuel, power, wind power and data, and the two pry units are both equipped with multi-degree-of-freedom universal locking roller groups 103, facilitating transportation and fixation.

[0200] The explosion-proof fire extinguishing device in the first pry-mounted unit comprises a linkage fire extinguishing system composed of an explosion-proof control cabinet 303, two explosion-proof fans 304, three combustible gas detectors 305, and a plurality of suspension type ultra-fine dry powder fire extinguishers 306; the explosion-proof fans 304 are symmetrically arranged at the bottom of the two sides of the inner wall of the container 307 of the first pry-mounted unit and are electrically connected with the explosion-proof control cabinet 303, and the exhaust port of the container 307 is provided with an explosion-proof louver 308; one combustible gas detector 305 is vertically suspended 50-100 cm above the flange joint of the gas pipeline, and the other two combustible gas detectors 305 are arranged in the middle of the top beam of the container above the gas cylinder or fuel tank 301, and the three combustible gas detectors 305 are all electrically connected with the explosion-proof control cabinet 303; the suspension type ultra-fine dry powder fire extinguisher 306 is suspended and fixed on the fire extinguisher annular lifting support 309 of the ceiling of the container 307, the height from the bottom of the fire extinguisher annular lifting support 309 to the bottom of the container 307 is 0.7 times the height of the container 307, the suspension type ultra-fine dry powder fire extinguisher 306 is driven by a pneumatic electromagnetic valve and is connected to the explosion-proof control cabinet 303, and the explosion-proof fan 304, the combustible gas detector 305, and the suspension type ultra-fine dry powder fire extinguisher 306 are all electrically connected with the explosion-proof control cabinet 303; the explosion-proof control cabinet 303 is connected with a touch panel 310 installed on the outer wall of the container 307, the touch panel 310 displays the gas concentration, the fan state, and the fire extinguisher pressure data in real time, when any combustible gas detector 305 detects that the gas concentration exceeds the standard, the explosion-proof control cabinet 303 starts the explosion-proof fan 304 to forcibly exhaust air, and at the same time, the pneumatic electromagnetic valve switch is triggered to activate the suspension type ultra-fine dry powder fire extinguisher 306 in the corresponding area to release dry powder, thereby realizing the automatic disposal of the whole process of explosion-proof, fire extinguishing, and risk removal.

[0201] The explosion-proof lighting device comprises an LED explosion-proof lamp group 311 with an explosion-proof grade greater than or equal to ExdIIBT4, which is symmetrically installed on the ceiling of the container 307, each explosion-proof lamp of the LED explosion-proof lamp group 311 is connected in parallel to the explosion-proof control cabinet 303 through an explosion-proof junction box 312, and the lighting switch control module of the LED explosion-proof lamp group 311 is integrated in the touch panel 310, when the combustible gas detector 305 alarms or the suspension type ultra-fine dry powder fire extinguisher 306 starts, the touch panel 310 automatically switches the lighting mode to the emergency high-brightness state.

[0202] The explosion-proof control cabinet 303, the explosion-proof fan 304, the combustible gas detector 305, the suspension type ultra-fine dry powder fire extinguisher 306, the container 307, the explosion-proof control cabinet 303, the explosion-proof louver 308, the fire extinguisher annular lifting support 309, the top beam of the container, the LED explosion-proof lamp group 311, and other structures not described in detail in the embodiment all adopt existing products or structures known to those skilled in the art, and the connection or control mode between them also adopts the existing connection or control mode known to those skilled in the art.

[0203] The number of devices in each embodiment is configured as one preferred embodiment of the present application, and other number of configurations are also within the protection scope of the present application.

[0204] The gas pipeline comprises a main pipeline 501 provided with a multi-channel converging interface, one end of the main pipeline 501 is connected to a gas cylinder or fuel tank 301 storing fuel through a high-pressure hose and an anti-explosion quick connector, the other end of the main pipeline 501 away from the gas cylinder or fuel tank 301 is sequentially connected to a vaporization furnace 302 and a gas main shutoff valve 502, and then is divided into several branch lines, each branch line is sequentially connected to an intelligent pressure reducing valve 503 with pressure feedback and an intelligent valve positioner 504 integrated with a liquid crystal screen, wherein the input end of the intelligent valve positioner 504 is connected to a pressure stabilizing branch of an existing air compressor known to those skilled in the art through a filter 505, the output end of the intelligent valve positioner 504 is connected to each gas interface of the horizontal linear combustion module and the vertical annular combustion module through a corrugated flame arrestor 506; the intelligent valve positioner 504 is connected to an LSTM-PID intelligent control system, and valve opening degree information is fed back to the LSTM-PID intelligent control system in real time, the LSTM-PID intelligent control system generates valve opening degree adjustment instructions for the next period based on preset target temperature parameters, and the instructions are returned to the intelligent valve positioner for opening degree adjustment, thereby forming a closed loop control link of dynamic matching of gas flow and temperature field, and the LSTM-PID intelligent control system can control the intelligent pressure reducing valve 503 to cut off the gas pipeline.

[0205] The fire scene panoramic video acquisition unit adopts a high-resolution camera and an infrared imaging device, captures visible light and thermal radiation images of a three-dimensional flame gradient form fire scene in real time through a fire scene panoramic camera, and monitors fire scene panoramic video;

[0206] The fire scene space temperature field monitoring unit comprises an infrared thermometer and a distributed thermocouple, is used for acquiring fire scene space temperature distribution data in real time, and generates a three-dimensional temperature field model; the infrared thermometer is erected at a position directly opposite the fire scene.

[0207] The wind speed monitoring unit is composed of an ultrasonic anemometer and a pressure sensor, is used for monitoring wind speed, wind direction and wind pressure data of a simulated wind field of a multi-dimensional wind field simulation system in real time; data monitored by the multi-dimensional high-precision monitoring system is transmitted to an LSTM-PID intelligent control system in real time through a communication network. The ultrasonic anemometer and the pressure sensor are placed at the same horizontal position 0.5 m away from the high-dynamic axial flow fan.

[0208] The multi-dimensional wind field simulation system comprises a plurality of high dynamic axial flow fans 101, which are arranged in a matrix topology, and are equipped with independent frequency converters 102 integrated in a second skid-mounted unit to realize continuous adjustment of wind speed and precise control of tilt angle; the fan array formed by the high dynamic axial flow fans 101 is controlled by a PLC intelligent dynamic control strategy in the LSTM-PID intelligent control system, which can realize simulation of wind field conditions in complex fire scenarios, including uniform wind field and gradient wind field; the multi-dimensional wind field simulation system is connected with ultrasonic wind speed and direction sensors and pressure sensors in the multi-modal fire environment monitoring system, and the specific connection mode adopts an existing connection mode well known to those skilled in the art, which is used for real-time monitoring of multi-dimensional parameters of the wind field, and feeding back monitoring data to the LSTM-PID intelligent control system, and performing multi-scale error analysis and regulation on wind field monitoring data and preset target parameters to ensure accurate reproduction of wind field conditions; the bottom of the multi-dimensional wind field simulation system is integrated with a multi-degree-of-freedom universal locking roller group 103, which supports all-terrain movement in various environments, and the roller locking mechanism ensures stability during the wind field simulation process.

[0209] The high-resolution camera and infrared imaging device, distributed thermocouple and infrared thermometer, ultrasonic anemometer and pressure sensor, high dynamic axial flow fan 101, frequency converter 102, power distribution cabinet 401, control cabinet 402, intelligent pressure reducing valve 503, intelligent valve positioner 504, touch panel 310, and LSTM-PID intelligent control system of the present embodiment adopt existing products or structures well known to those skilled in the art, and the connection or control mode between them also adopts the existing connection or control mode well known to those skilled in the art.

[0210] Other components not disclosed in detail in the present embodiment adopt existing products or structures well known to those skilled in the art, and the connection mode between them and other structures of the present embodiment also adopts the existing connection mode well known to those skilled in the art.

[0211] The LSTM-PID intelligent control system is embedded with a deep spatio-temporal sequence prediction model based on a long short-term memory (LSTM) architecture known to those skilled in the art, which receives real-time data from a multi-modal fire environment monitoring system, including holographic images of the fire scene, temperature spatio-temporal distribution, and multi-dimensional wind field parameters, wherein the multi-dimensional wind field parameters include wind speed vector, fan inclination angle, and fan turbulence intensity; the LSTM model is used for spatio-temporal feature fusion and nonlinear modeling of the data, and outputs the dynamic opening optimization value of each gas control valve and the target wind speed and angle control instruction of the high-dynamic axial flow fan of the multi-dimensional wind field simulation system; the predicted value is transmitted in real time to the PLC intelligent dynamic control system of the LSTM-PID intelligent control system, which dynamically adjusts the opening of each gas control valve and the wind speed and inclination angle of each high-dynamic axial flow fan through the control cabinet 402 according to the real-time monitoring data, to ensure accurate control and reproduction of real fire test conditions; the PLC intelligent dynamic control system and the host computer are interconnected through the OPCUA protocol, and the holographic thermal map of the fire scene, the gas valve opening deviation curve, the fan operating state matrix, and the control parameter iteration process are presented in real time and visualized.

[0212] Referring to Figures 15-17 In this embodiment, the method for constructing the fire temperature field rapid generation model comprises:

[0213] S101, a GAN network is built, which comprises a generator and a discriminator; wherein,

[0214] The generator takes the fire source power, normalized position coordinates (x, y), and wind speed as input conditions, extracts local fire dynamic features through a double-layer spatio-temporal convolution LSTM branch, and embeds a physical constraint branch by constructing a residual term of a heat diffusion equation; after the double-path features are fused through a fully connected layer, a normalized temperature field matrix is output, and finally the actual temperature-time curve is de-normalized; the two branch channels of the double-layer spatio-temporal convolution LSTM are 64 and 128 respectively, and LeakyReLU activation is used; the normalized temperature field matrix uses Sigmoid activation, and the value range is [0, 1];

[0215] The discriminator takes the generated or real temperature field matrix and the fire source parameters (power, position, wind speed) as joint input, extracts spatio-temporal joint features through a three-layer convolution network, and embeds a gradient penalty term to constrain the smoothness of the generated data; the output layer is compressed into a single node through global average pooling, and outputs the temperature field authenticity probability, while the residual connection is used to calculate the residual of the heat diffusion equation as a physical reasonableness criterion; the three-layer convolution network has channel numbers of 128, 64, and 32 respectively, and uses ReLU activation; the output layer uses Sigmoid activation;

[0216] The residual of the heat diffusion equation is the difference between the generated temperature field and the predicted value of the theoretical equation, and the calculation formula is:

[0217]

[0218] wherein, is a generated temperature field, is an ambient temperature, is a spatial second derivative of the generated temperature field, is a balance term of the heat source and the ambient temperature; is an L2 norm for measuring the size of the residual.

[0219] S102, based on a fire dynamics software (such as FDS), a three-dimensional non-uniform temperature field data of a full bridge is simulated to cover different combinations of fire source power, position and wind speed; the spatio-temporal temperature distribution data is extracted to construct a training data set, the input is the fire source parameter, and the output label is a grid temperature field matrix; the data reliability is verified through a test substructure test, the target temperature field of the fire dynamics software is reproduced, the measured and simulated temperature gradient distributions are compared, and it is ensured that the error range is less than or equal to 5%;

[0220] S103, training a fire temperature field fast generation model based on a GAN network offline, and based on the simulation of the fire dynamics software and the physical test verification data;

[0221] S104, deploying the trained fire temperature field fast generation model to a computing server, and binding a fire source parameter input interface; inputting a target as a fire source power, a position and a wind speed to generate a target temperature-time curve. The architecture of the fire temperature field fast generation model neural network based on the GAN is as shown in Figure 4 .

[0222] According to one embodiment of the present application, in step S2, the numerical substructure fast calculation model based on LSTM is a multi-scale gated residual LSTM neural network prediction model, which is constructed by introducing a parallel multi-branch structure, a dynamic feature selection gate and a residual cross-layer connection on the basis of LSTM, and is used for predicting the evolution process under the full bridge fire according to the measured values of the multi-dimensional monitoring sensors in the test substructure test and outputting the loading boundary force or displacement instruction in real time; the specific steps of training the numerical substructure fast calculation model based on LSTM are as follows:

[0223] S201, a multi-scale gated residual LSTM neural network is built, wherein the multi-scale gated residual LSTM neural network includes a double-branch LSTM unit, i.e., a first branch LSTM unit and a second branch LSTM unit, and the specific building process includes:

[0224] S2011, the input layer receives historical time series data, and the input features include position spatial coordinates, temperature field, boundary displacement, cable force and wind speed;

[0225] S2012 is a parallel dual-branch LSTM unit. The first branch LSTM unit processes the original step-size time series data with a step size of 1, 128 hidden units, and the activation function tanh, and is used to extract short-term local mutation features. The second branch LSTM unit processes downsampled time series data with a step size of 5, 64 hidden units, and the activation function tanh, and is used to extract long-term global evolution features.

[0226] S2013, the dynamic feature selection gate calculates branch weights using the sigmoid function, and the fusion formula is:

[0227]

[0228] Among them, h final h is the hidden state vector after fusing the two branch LSTM units. branch1 h is the output hidden state vector of the first branch LSTM unit. branch2 Let g be the output hidden state vector of the second branch LSTM unit. t The meaning is dynamic gating weight coefficient;

[0229] S2014, residual block cross-layer connections alleviate gradient vanishing by superimposing the outputs and inputs of the first and second LSTM layers through skip connections, as shown in the formula:

[0230] ;

[0231] in, For time step t at the 1st l The hidden state vector of the LSTM layer, The previous time step of time step t is at time step t. l The hidden state vector of the LSTM layer, Let [x, y, z] be the spatial coordinate vector at time step t. t , For time step t at the 1st l The hidden state vector of the -1st layer LSTM For the first l Layer LSTM cells, l =(1,2)

[0232] S2015 uses a hybrid activation function, with the forget gate and input gate using the sigmoid function, the candidate memory unit using the tanh function, and the output gate using the Swish function f(x)=x·σ(x);

[0233] S2016, a fully connected output layer, outputs the predicted boundary force or displacement values ​​for the next time step;

[0234] S202, constructing a data set of the LSTM-based numerical substructure rapid calculation model, specifically comprising the following steps:

[0235] S2021, simulating different fire scenes using fire dynamics software, calculating the three-dimensional non-uniform temperature field distribution of the whole bridge for different combinations of fire source power, fire source position and wind speed, and extracting the temperature-time curve of the key positions;

[0236] S2022, establishing a thermal-mechanical finite element model of the whole bridge and the substructure based on a finite element analysis software (such as ABAQUS), interpolating the temperature field output by the fire dynamics software to the finite element grid nodes according to the time step, and through the sequential coupling analysis method, first applying the temperature load to calculate the thermal expansion effect, and then superimposing the mechanical load to analyze the structural response, and outputting the displacement, cable force and strain data;

[0237] S2023, making test specimens corresponding to the experimental substructure, reproducing the target temperature field of the fire dynamics software in the laboratory, applying equivalent mechanical loads through a through jack, and comparing the measured displacement and cable force data with the simulation results of the finite element analysis software to verify the reliability of the finite element model;

[0238] S2024, processing the verified simulation data according to the input requirements of the multi-scale gated residual LSTM, constructing a time series data set, and the input features include the temperature, displacement, cable force and wind speed at the key positions, and the output features are the displacement and cable force at the next time step;

[0239] S203, offline pre-training, dividing the data set using a sliding time window, the loss function is HuberLoss (δ=1.5), the optimizer is Nadam, the initial learning rate is 0.001, and it is attenuated by 50% every 10 epochs; the first branch LSTM unit inputs the original step data, the second branch LSTM unit inputs the down-sampled data, and the dynamic feature selection gate weight is optimized through the backward propagation of historical errors;

[0240] S204, online incremental learning, by real-time intercepting sensor data, only fine-tuning the model down-sampling branch parameters and constraining the weight change amplitude ≤0.01, dynamically adapting to the changes of fire, while maintaining the stability of the model and improving the prediction accuracy;

[0241] S205, real-time prediction and control during real-time hybrid testing, the specific process is as follows:

[0242] S2051, taking displacement loading as an example, converting the displacement at the next time step into loading displacement instructions according to the displacement equivalent principle, the formula is:

[0243]

[0244] In the formula, D loadD is displacement instruction pred L is displacement of next time step phy L is actual geometric span length of physical specimen num L is numerical substructure equivalent span length, for large-scale model phy L = L num ;

[0245] S2052, drive the cable system self-balancing counterforce frame through the LSTM-PID intelligent control system, and the control quantity calculation formula is:

[0246]

[0247] In the formula, γ is a feedforward gain coefficient, which is dynamically adjusted through fuzzy logic;

[0248] K p is a proportional gain coefficient, e(t) is the error between the real-time combustion state of the frame burner and the target value, K i is an integral gain coefficient, and K d is a differential gain coefficient;

[0249] S2053, dynamically gate the weight g t to the PID proportional coefficient K p , to realize adaptive loading;

[0250] S206, cyclically update the thermal-displacement / force boundary conditions until the specimen deformation or stress is over limit, and output the failure mode and critical threshold. The architecture of the multi-scale gated residual LSTM neural network is shown in Figure 2 .

[0251] In this embodiment, the input layer of the LSTM-PID intelligent control system includes temperature field data, wind speed, angle, burner valve opening and historical error, the hidden layer includes 54-64 neurons, and the output layer outputs the burner valve opening and the fan frequency;

[0252] Finally, the control instruction is issued to the burner and the fan frequency converter, and the wind-fire coupling loading is triggered synchronously, to realize dynamic reproduction of the fire scene;

[0253] The generation steps of the control instruction are:

[0254] S301, real-time acquisition of temperature field data, wind speed angle data and burner valve opening history sequence by using a sensor, to construct a multi-dimensional time series input vector;

[0255] S302, input the input vector into the pre-trained LSTM prediction model to predict the burner valve opening and fan frequency at the next time, wherein the LSTM prediction model updates the weight online through a sliding time window, and the prediction error converges to <5%;

[0256] S303 uses the predicted burner valve opening increment or fan frequency as a feedforward correction, which is then superimposed on the feedback output of the LSTM-PID intelligent control system to generate the final control command.

[0257] In the horizontal linear combustion module, the control command for the i-th frame burner is calculated and generated according to the following formula (1);

[0258] (1)

[0259] In the formula: α i The dynamic weighting coefficient for the i-th frame burner is determined by the LSTM model embedded in the LSTM-PID intelligent control system, and can be dynamically adjusted by those skilled in the art based on real-time operating conditions; u gas,i(t) For the system control command of the i-th frame burner at time step t;

[0260] It is an existing PID control equation. K p For proportional gain coefficient, e i ( t ) represents the error between the real-time combustion state and the target value of the i-th frame burner. K i For integral gain coefficient, K d Here, t is the differential gain coefficient, and t is the combustion time.

[0261] LSTM gas,i (T i ,F i W) represents the correction term for the i-th frame burner output by the LSTM model, and the input is the temperature T of the i-th frame burner. i Flame Form F i The real-time data, W is the fan speed; where: the total number of frame burners is selected according to the size of the test specimen, in this embodiment i=(1,2,3);

[0262] The control commands for the j-th layer of the vertical annular combustion module are calculated and generated according to formula (2):

[0263] (2)

[0264] Where: β j u represents the dynamic weighting coefficient of the j-th layer of the vertical annular combustion module. gas,j (t) represents the system control command for the j-th layer of the vertical annular combustion module at time step t; e j(t) is the error of the real-time combustion state of the vertical annular combustion module j layer and the target value; LSTM gas,j (T i ,F i ,W) is the correction term of the LSTM model output of the vertical annular combustion module j layer, and the input is the real-time data of the temperature T j , flame shape F j of the vertical annular combustion module j layer; wherein: the number of layers of the vertical annular combustion module is selected according to the size of the test specimen, and in the embodiment, j=(1, 2, 3);

[0265] The control instruction calculation of the frequency of the frequency converter controlling the fan speed is generated according to the following formula (3): (3)

[0266] In the formula, γ is the dynamic weight coefficient of the frequency of the frequency converter, f fan (t) is the system control instruction of the frequency of the frequency converter; LSTM fan (W, θ) is the correction term of the LSTM model output of the frequency of the frequency converter, and the input is the fan speed and the fan inclination angle θ;

[0267] Referring to Figure 18 , the LSTM time lag compensator main body is a neural network based on LSTM, containing a hidden layer of 16-32 neurons, and the input layer nodes contain a historical time lag sequence Δt i , a system real-time load rate, an environmental temperature gradient and an environmental wind speed gradient, wherein i=(1, 2,..., n), and the output layer is mapped to a next analysis step time lag prediction value Δt i+1 through a fully connected layer, and specifically includes the following steps:

[0268] S401, constructing a time lag data set, collecting a historical time lag sequence Δt i , a system real-time load rate, an environmental temperature gradient and an environmental wind speed gradient are constructed into an input vector, and input into a neural network based on LSTM, wherein Δt i refers to the delay time from the issuance of the control command to the real-time value of the sensor meeting the standard;

[0269] S402, offline training and online optimization, training the LSTM-based time lag compensation model using the preloaded data samples, and the loss function is the mean absolute error; when formally loaded, the network parameters are updated online through the optimizer every time an analysis step is completed;

[0270] S403, dynamic time lag compensation, sending the control instruction in advance by Δt2 time before the second analysis step is loaded, and recording the actual time lag Δt real , and Δt2 refers to the time difference from the issuance of the first analysis step instruction to the loading in place;

[0271] S404, error correction and convergence, based on Δt i-real deviation from Δt i The gradient descent method is used to correct the time delay compensation model weight based on LSTM until the time delay error of consecutive multiple analysis steps is ≤1 second; Δt i-real refers to the actual measured time delay of each analysis step, i=(1, 2, …, n).

[0272] The embodiment also provides an implementation method of the large-span bridge substructure vehicle-fire-wind real-time hybrid test platform, including the following steps:

[0273] Step one, making and installing test pieces: the types of test pieces include main cable models, stay cable models, and sling models;

[0274] Installing the main cable model or the stay cable model: synchronously lower the two side sliding beams to the base reference surface;

[0275] Pull the main cable model or the stay cable model through the free-rotation anchoring subsystem and the free-rotation force boundary subsystem in sequence and anchor; then synchronously lift the two side sliding beams to a preset height or angle;

[0276] Installing the sling model: connect the sling model to the main cable model through the cable clamp, pass the lower end of the sling model through the second through-type jack and anchor, and install force sensors at the front ends of the first through-type jack and the second through-type jack;

[0277] Step two, selecting a test mode and configuring an environment;

[0278] Mode A: open space self-defined fire test;

[0279] Input target fire parameters into the fire temperature field fast generation model, generate a non-uniform temperature field, and transmit it to the LSTM-PID intelligent control system;

[0280] Mode B: closed space standard fire test

[0281] Splice the enclosing units through the quick-release interface to form a closed space, select the standard fire curve mode in the LSTM-PID intelligent control system, and close the multi-dimensional wind field simulation system;

[0282] Step three, preloading and time delay compensation initialization:

[0283] Start the cable system self-balancing counterforce frame and the T-shaped wind-fire coupling simulation system, pre-load, and collect feedback data of each sensor;

[0284] Calculate the initial time delay Δt i of each system by using the least square method, and input it into the LSTM-PID intelligent control system for feedforward compensation;

[0285] Step four, generate the selected mode under the fire scene temperature rise curve and perform real-time hybrid test:

[0286] Step 4.1, input the fire source power, location and wind speed to the fire temperature field fast generation model to generate non-uniform temperature field, and transmit to the LSTM-PID intelligent control system to analyze as the burner valve opening and fan frequency command;

[0287] If mode B is executed, input the standard target temperature rise curve, and output the burner valve opening command synchronously;

[0288] Step 4.2, start gas supply, dynamically adjust gas flow through LSTM-PID intelligent control system, and start pilot flame synchronously;

[0289] If mode A is executed, start the multi-dimensional wind field simulation system and dynamically adjust the fan inclination angle synchronously;

[0290] Step 4.3, based on the LSTM numerical substructure fast calculation model, real-time prediction of full bridge displacement and cable force, converted into loading instructions of the first and second through-type jacks 54 and 71;

[0291] Step 4.4, load step by step according to displacement / force control mode, use force control mode to slowly stretch to specimen fracture when exceeding the preset limit;

[0292] Step 4.5, during the whole loading process, based on LSTM time delay compensator, each analysis step Δt i Predict the time delay Δt i+1 Of the next analysis step, and send instructions in advance Δt i+1 To ensure that force loading, combustion and wind field are synchronized, where i=(1,2,......,n); If mode B is executed, only predict the loading and temperature time delay, and ensure that loading and combustion are synchronized;

[0293] Step five, data acquisition and dynamic visualization, specifically including:

[0294] Step 5.1, real-time acquisition of specimen axial force, displacement, strain and fire field temperature data, and drawing of related curves in LSTM-PID intelligent control system;

[0295] If mode A is executed, wind speed data of the fan is collected synchronously;

[0296] Step 5.2, bind the sensor measured data and predicted data to each grid node of the full structure three-dimensional graph through three-dimensional animation software, and dynamically visualize the full bridge stress, deformation and temperature field distribution based on numerical-color mapping algorithm;

[0297] Step 5.3, generate load-displacement curve, temperature-time curve, output damage evolution law and failure warning.

[0298] The above merely describes preferred embodiments of the present application and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A large-span bridge substructure vehicle-fire-wind real-time hybrid test platform, characterized in that, The platform comprises a physical loading and transmission system, a T-shaped wind-fire coupling simulation system, a modularized skid-mounted integrated system, an intelligent control and calculation system, and a multi-dimensional high-precision monitoring system. The physical loading and transmission system comprises a cable system self-balancing counterforce frame, a beam electric lifting locking system, and a roller bearing transmission system. The T-shaped wind-fire coupling simulation system comprises a T-shaped global flame envelope combustion system and a multi-dimensional wind field simulation system. The intelligent control and calculation system comprises a fire temperature field rapid generation model, an LSTM-based numerical substructure rapid calculation model, an LSTM-PID intelligent control system, and an LSTM time delay compensator. The LSTM-PID intelligent control system comprises a temperature field data, wind speed, angle, burner valve opening, and historical error in the input layer, 54-64 neurons in the hidden layer, and a burner valve opening increment and high dynamic axial flow fan frequency of the T-shaped global flame envelope combustion system in the output layer. The multi-dimensional high-precision monitoring system comprises a fire field panoramic video acquisition unit, a fire field space temperature field monitoring unit, and a wind speed monitoring unit.

2. The real-time hybrid test platform for vehicle-fire-wind force of a long-span bridge substructure according to claim 1, wherein the cable system self-balancing counterforce frame comprises two groups of symmetrically arranged rectangular lattice frames, the two groups of rectangular lattice frames are each provided with a base at the bottom and the two bases are connected through a box-type bottom beam, the two groups of rectangular lattice frames are connected through a main cross beam at the top, and the two groups of rectangular lattice frames are each provided with a sliding cross beam at the middle, which is freely slidable and lockable along the height direction of the rectangular lattice frame. The cable system self-balancing counterforce frame further comprises a free-rotation force boundary subsystem and a free-rotation anchoring subsystem, the free-rotation force boundary subsystem and the free-rotation anchoring subsystem each comprise a composite anchor ring, the composite anchor ring is installed in the sliding cross beam and freely rotates in the sliding cross beam, and the free sliding of the sliding cross beam realizes the suspension tension at any angle during the loading of the main cable model; the free-rotation force boundary subsystem further comprises a first through-type jack and a loading end two-stage gradient anchoring unit, the first through-type jack is fixedly installed in the composite anchor ring, and the loading end two-stage gradient anchoring unit comprises a loading end primary anchor and a loading end secondary anchor, which are arranged on the two sides of the first through-type jack; the free-rotation anchoring subsystem further comprises an anchoring end anchor. The two sides of the composite anchor ring are each fixedly installed with a variable cross-section transmission main shaft, the variable cross-section transmission main shaft comprises a quadrangular frustum transmission section fixedly connected with the composite anchor ring and a cylindrical rotation section fixedly connected with the quadrangular frustum transmission section, and the cylindrical rotation section penetrates into the sliding cross beam and freely rotates in the sliding cross beam.

3. The real-time hybrid test platform for vehicle-fire-wind force of a long-span bridge substructure according to claim 2, wherein ​ The cable system self-balancing counterforce frame further comprises a sling force loading anchoring subsystem, the sling force loading anchoring subsystem comprising a second through-jack, a sling loading end anchor and a locking anchoring mechanism; the second through-jack is embedded in the internal cavity of the box-shaped bottom beam, the sling loading end anchor is anchored to the end of the sling model, and the outer diameter of the sling loading end anchor is smaller than the inner diameter of the second through-jack; the locking anchoring mechanism is arranged between the second through-jack and the sling loading end anchor, and the inner diameter of the locking anchoring mechanism is equal to the diameter of the sling model, and the outer diameter of the locking anchoring mechanism is greater than the outer diameter of the second through-jack; The sling force loading anchoring subsystem further comprises a T-shaped sliding rail and a sliding rail limiter, the T-shaped sliding rail is fixedly installed on the box-shaped bottom beam, and the sliding rail limiter is installed on the T-shaped sliding rail; the second through-jack moves on the T-shaped sliding rail and is fixedly limited by the sliding rail limiter.

4. The long-span bridge substructure vehicle-fire-wind force real-time hybrid test platform according to claim 3, wherein The horizontal beam electric lifting locking system is provided with two groups corresponding to the sliding horizontal beam, each group of the horizontal beam electric lifting locking system comprises a screw rod linkage synchronous lifting subsystem, a limiting sliding rail and an array type positioning hole group; the screw rod linkage synchronous lifting subsystem comprises a screw rod, a lifting machine, a motor and a base; the screw rod is provided with two symmetrically, one end of each screw rod is rotationally connected with the sliding horizontal beam through a flange, and the other end of the screw rod is connected with the lifting machine; the lifting machine is provided with two corresponding to the screw rod, the motor is connected with the two lifting machines through a transmission shaft; the base is fixedly installed on the top of the rectangular lattice type frame, and the lifting machine and the motor are fixedly installed on the base; the limiting sliding rail is integrated on the inner side of the rectangular lattice type frame, and forms a sliding pair with the sliding block at the end of the sliding horizontal beam; the array type positioning hole group is distributed at equal intervals along the height direction of the rectangular lattice type frame, and cooperates with the detachable bolt mechanism on the sliding horizontal beam to lock the position of the sliding horizontal beam; The roller bearing transmission system comprises an adjustable roller bearing platform installed on the base, the adjustable roller bearing platform comprises a base frame, a lifting mechanism and a top bearing beam; the lifting mechanism is arranged between the base frame and the top bearing beam, the top bearing beam moves up and down under the drive of the lifting mechanism, and the top bearing beam is equal in height to the box-shaped bottom beam after being lifted; a plurality of roller groups are arranged at equal intervals on the adjustable roller bearing platform and the box-shaped bottom beam, the top bearing beam and the box-shaped bottom beam are each provided with a concave clamping groove for installing the roller groups, and the height of the base is lower than that of the box-shaped bottom beam.

5. The long-span bridge substructure vehicle-fire-wind force real-time hybrid test platform according to claim 4, wherein The horizontal linear combustion module comprises a plurality of frame burners arranged in a linear manner and provided with vertical flame injection pipes, the frame burners are suspended and fixed in the middle test area of the cable system self-balancing counterforce frame through a plurality of steel wire ropes, a gas input interface and a self-suction air supply interface are arranged at the bottom of one side wall of the frame burner, a multi-directional flame injection unit capable of forming a three-dimensional flame gradient mode is arranged at the top of the frame burner, the multi-directional flame injection unit comprises a plurality of gas collecting pipes arranged side by side, and a plurality of flame injection pipes are uniformly distributed on the gas collecting pipes; The middle part of the gas collecting pipe is communicated with the gas input interface through a plurality of equal distribution branch pipes, the upper part of the equal distribution branch pipe extends into the gas collecting pipe and is provided with a reverse v-shaped diffusion cutout, a flow equalization assembly is arranged in the frame burner, air enters the frame burner from the self-suction air supply interface of the frame burner, and the flow equalization assembly forms a flow equalization air supply multi-directional flame injection unit; the reverse v-shaped diffusion cutout is used for realizing uniform decreasing diffusion distribution of gas from the middle part of the gas collecting pipe to both ends, the flow equalization air supply formed by the flow equalization assembly is used for synchronously igniting each gas collecting pipe to reproduce a real three-dimensional flame natural gradient mode, and a real fire space non-uniform temperature field is presented; The gas input interface is supplied with gas through a gas pipeline of the independent and flow-adjustable modularized pry integrated system, and the self-suction air supply interface realizes natural air supply through the negative pressure self-suction principle; the flow equalization assembly comprises an air flow equalization plate and two arc-shaped flow guide partitions arranged in the air flow direction, the self-suction air supply interface forms a rectangular gradually expanding flow channel in the frame burner, and the two arc-shaped flow guide partitions arranged in the air flow direction uniformly divide the air inlet of the frame burner into three left, middle and right paths to initially distribute the air; the air flow equalization plate is arranged at the lower part of the multi-directional flame injection unit to secondarily distribute the air.

6. The long-span bridge substructure vehicle fire wind force real-time hybrid test platform according to claim 5, wherein The vertical annular combustion module comprises a plurality of layers of annular burners which are detachable, adjacent annular burners are connected through detachable support bodies, a plurality of flame injection ports capable of adjusting a flame injection angle are uniformly arranged on each layer of annular burners, the flame injection ports are supplied with gas through symmetrical double gas inlet branch pipes arranged on the annular burners, a high-energy pulse electric spark igniter is arranged at any two symmetrically distributed flame injection ports in each layer of annular burners, synchronous ignition of flames of each layer is realized through a LSTM-PID intelligent control system synchronous control circuit, and each double gas inlet branch pipe is supplied with gas through an independent and flow-adjustable gas pipeline in a movable pry integrated system. The fire spout comprises an air inlet end short pipe and a fire spouting end short pipe, the air inlet end short pipe and the fire spouting end short pipe are connected through an open hollow hemispherical universal hinge mechanism, the universal hinge mechanism comprises a hollow ball body with an embedded flow guide channel and a hemispherical socket matched with the hollow ball body, and can realize 0°-90° inclination angle adjustment of the fire spouting end short pipe, the air inlet end short pipe is communicated with the annular combustor through a variable diameter transition pipe; after the gas enters the annular combustor, the gas is accelerated into the universal hinge mechanism through the variable diameter transition pipe, and is spouted from the fire spouting end short pipe and ignited, forming a flame with adjustable spouting direction; The annular combustor is integrated with a flange type quick release interface along the symmetry axis direction of the annular combustor, the fire spout and the double air inlet branch pipes do not interfere with each other with the flange type quick release interface, the flange type quick release interface realizes axial separation and recombination of the annular combustor through bolt connection, after disassembly, two half annular combustor units can be formed, and each half annular combustor unit retains independent gas supply passage and ignition control, and supports true fire loading in a half circle fire scene; the detachable support body is provided with an internal thread quick connector at both ends for connecting adjacent two layers of annular combustors; the multi-dimensional wind field simulation system comprises a plurality of high dynamic axial flow fans, and the high dynamic axial flow fans are arranged in a matrix topology form in the middle test area of the cable system self-balancing counterforce frame.

7. The real-time hybrid test platform for vehicle fire wind force on substructure of long-span bridge according to any one of claims 1-6, characterized in that, The modularized prying integrated system comprises a first prying unit and a second prying unit; the first prying unit is internally integrated with a gas cylinder or a fuel tank, a vaporizing furnace, an explosion-proof fire extinguishing device, an explosion-proof lighting device, and a gas pipeline communicated with the T-shaped global flame envelope combustion system; the second prying unit is internally integrated with a power distribution cabinet, a control cabinet, and a frequency converter; the control cabinet is internally provided with a PLC controller and a communication module, and is used for realizing coordinated control of each system; The explosion-proof fire extinguishing device comprises a linkage fire extinguishing system composed of an explosion-proof control cabinet, an explosion-proof fan, a combustible gas detector, and a plurality of suspended ultra-fine dry powder fire extinguishers, and the explosion-proof lighting device is electrically connected with the explosion-proof control cabinet; the power distribution cabinet is used for providing power supply; The gas pipeline comprises a main pipeline provided with a multi-channel converging interface, one end of the main pipeline is connected with the gas cylinder or the fuel tank through a high-pressure hose and an explosion-proof quick connector, the other end of the main pipeline away from the gas cylinder or the fuel tank is sequentially connected with the vaporizing furnace, a gas main shut-off valve, an intelligent pressure reducing valve with pressure feedback, and an intelligent valve positioner integrated with a liquid crystal screen, an input end of the intelligent valve positioner is connected with a pressure stabilizing branch of an air compressor through a filter, and an output end of the intelligent valve positioner is connected with the T-shaped global flame envelope combustion system through a corrugated flame arrester.

8. The long-span bridge substructure vehicle-fire-wind real-time hybrid test platform according to any one of claims 1-6, characterized in that, The construction method of the fire temperature field rapid generation model comprises: S101, a GAN network is built, the GAN network comprises a generator and a discriminator; wherein, The generator takes the fire source power, normalized position coordinates and wind speed as input conditions, extracts local fire dynamic characteristics through a double-layer spatio-temporal convolution LSTM branch, and embeds a physical constraint branch by constructing a residual term of the heat diffusion equation; after the double-path features are fused through a fully connected layer, a normalized temperature field matrix is output, and finally it is denormalized to an actual temperature-time curve; The discriminator takes the generated or real temperature field matrix and fire source parameters as joint input, extracts spatio-temporal joint features through a three-layer convolution network, and embeds a gradient penalty term to constrain the smoothness of the generated data; the output layer is compressed into a single node through global average pooling, and outputs the temperature field authenticity probability, while the residual connection is used to calculate the residual of the heat diffusion equation as a physical reasonableness criterion; The residual of the heat diffusion equation is the difference between the generated temperature field and the predicted value of the theoretical equation, and the calculation formula is: ； wherein is the generated temperature field, is the ambient temperature, is the spatial second derivative of the generated temperature field, is the balance term for the heat source and the ambient temperature; is the L2 norm, used to measure the size of the residual; S102, generate full-bridge three-dimensional non-uniform temperature field data based on fire dynamics software simulation, extract spatio-temporal temperature distribution data, construct a training data set, verify the reliability of the data through experiments, and compare the measured and simulated temperatures to ensure that the error range is ≤5%; S103, offline training of the fire temperature field fast generation model based on the GAN network, and simulation and physical test verification data based on the fire dynamics software; S104, deploy the trained fire temperature field fast generation model to a computing server.

9. The real-time hybrid test platform for the substructure of a long-span bridge under vehicle fire wind of any one of claims 1-6, characterized in that: The LSTM-based numerical substructure fast calculation model is a multi-scale gated residual LSTM neural network prediction model, which is constructed by introducing a parallel multi-branch structure, a dynamic feature selection gate and a residual cross-layer connection based on LSTM, and is used to predict the evolution process under full-bridge fire according to the measured values of each sensor in the test and output the loading boundary force or displacement instruction in real time; the specific steps for training the LSTM-based numerical substructure fast calculation model are as follows: S201, build a multi-scale gated residual LSTM neural network, wherein the multi-scale gated residual LSTM neural network includes a double-branch LSTM unit, i.e., a first branch LSTM unit and a second branch LSTM unit; S202, construct a data set for the LSTM-based numerical substructure fast calculation model, specifically including: simulating the three-dimensional non-uniform temperature field distribution in the full-bridge space under different fire scenarios using fire dynamics software; establishing a thermal-mechanical coupling finite element model of the full-bridge and the test substructure based on finite element analysis software; making a test specimen corresponding to the real cable structure, and verifying the accuracy of the thermal-mechanical coupling model of the test specimen by loading the non-uniform temperature field; and using the verified simulation data to train the multi-scale gated residual LSTM neural network model; S203, offline pre-training, the data set is divided by a sliding time window, the first branch LSTM unit inputs the original step data, the second branch LSTM unit inputs the down-sampled data, and the dynamic feature selection gate weight is optimized through backward propagation of historical errors. S204, online incremental learning, by real-time intercepting sensor data, only fine-tuning model down-sampling branch parameters and constraining weight change amplitude ≤0.01, dynamically adapting to fire changes; S205, real-time prediction and control during real-time hybrid test; S206, cyclic iteration updating thermal-displacement / force boundary conditions until specimen deformation or stress exceeds limit, outputting failure mode and critical threshold.

10. The long-span bridge substructure vehicle-fire wind force real-time hybrid test platform of claim 9, wherein, Step S201 includes: S2011, input layer receiving historical time series data, input features including position space coordinates, temperature field, boundary displacement, cable force and wind speed; S2012, parallel double-branch LSTM unit, wherein the first branch LSTM unit processes original step time series data for extracting short-time local mutation features; the second branch LSTM unit processes down-sampled time series data for extracting long-time global evolution features; S2013, dynamic feature selection gate calculating branch weight through sigmoid function, fusion formula being: ； wherein h final is the hidden state vector after fusion of the two branch LSTM units, h branch1 is the output hidden state vector of the first branch LSTM unit, h branch2 is the output hidden state vector of the second branch LSTM unit, g t has the meaning of a dynamic gating weight coefficient; S2014, residual block cross-layer connection relieving gradient vanishing, output of the first branch LSTM unit and the second branch LSTM unit being superimposed with input through jump connection, formula being: ； wherein, is the hidden state vector of the layer LSTM at time step t in the l layer LSTM, is the hidden state vector of the layer LSTM at the previous time step t-1 in the l layer LSTM, is the spatial coordinate vector [x, y, z] at time step t, t , is the hidden state vector of the layer-1 LSTM at time step t in the l layer-1 LSTM, is the layer-1 LSTM unit at time step t, l is the layer-1 LSTM unit at time step t-1, l = (1, 2); S2015, hybrid activation function; S2016, fully connected output layer, outputting boundary force prediction value or displacement prediction value of next time step.

11. The long-span bridge substructure vehicle-fire wind force real-time hybrid test platform of claim 10, wherein, Step S205 includes: S2051, taking displacement loading as an example, converting displacement of next time step into loading displacement instruction according to displacement equivalence principle, formula being: ； where D load is the displacement command, D pred is the displacement at the next time step; L phy is the actual geometric span of the physical specimen, L num is the numerical substructure equivalent span, for large scale models L phy = L num ; an experimental model with a scale ratio not less than 1:4 to the original structure is called a large scale model; S2052, driving cable system self-balancing counterforce frame through LSTM-PID intelligent control system, control quantity calculation formula being: ； where γ is a feed-forward gain coefficient, dynamically adjusted by fuzzy logic; K p is a proportional gain coefficient, e(t) is the error between the real-time combustion state of the frame combustor and the target value, K i is an integral gain coefficient, K d is a differential gain coefficient, and t is the combustion time. S2053, dynamic gating weight mapping to PID proportionality coefficient, realizing adaptive loading.

12. The long-span bridge substructure vehicle-fire-wind real-time hybrid test platform according to claim 6, characterized in that, the generation step of the control instruction of the LSTM-PID intelligent control system is: S301, using sensors to collect temperature field data, wind speed angle data and historical sequence of burner valve opening degree in real time, constructing input vector of multi-dimensional time series; S302, inputting the input vector into a pre-trained LSTM prediction model to predict burner valve opening degree increment and fan frequency of next time, wherein the LSTM prediction model updates weight online through sliding time window, and prediction error converges to <5%; S303, taking the predicted valve opening degree of each burner or fan frequency as a feedforward correction, respectively superimposing to feedback output of the LSTM-PID intelligent control system to generate final control instruction; in the horizontal linear combustion module, control instruction calculation of the i th frame-type burner is generated according to the following formula (1); （1） In the formula: a i is the dynamic weight coefficient of the ith frame combustor, which is determined by the LSTM model embedded in the LSTM-PID intelligent control system; u gas,i(t) is the system control instruction of the ith frame combustor at time step t; K p is the proportional gain coefficient, e i ( t ) is the error between the real-time combustion state of the ith frame combustor and the target value, K i is the integral gain coefficient, K d is the differential gain coefficient; LSTM gas,i (T i ,F i ,W) is the correction term of the i-th frame combustor output by the LSTM model, the input is the real-time data of the temperature T i , flame shape F i of the i-th frame combustor, and W is the fan speed; in the vertical ring-shaped combustion module, control instruction calculation of the j th layer is generated according to the following formula (2): （2） In the formula: β j is the dynamic weight coefficient of the jth layer of the vertical annular combustion module, u gas,j (t) is the system control instruction of the jth layer of the vertical annular combustion module at the t time step; e j (t) is the error between the real-time combustion state and the target value of the jth layer of the vertical annular combustion module; LSTM gas,j (T i ,F i ,W) is the correction term of the jth layer of the vertical annular combustion module output by the LSTM model, and the input is the real-time data of the temperature T j and the flame shape F j of the jth layer of the vertical annular combustion module; control instruction calculation of frequency of the frequency converter controlling the wind speed is generated according to the following formula (3): （3） where γ is a dynamic weight coefficient of the frequency converter frequency, f fan (t) is the system control instruction of the frequency converter frequency; LSTM fan (W, θ) is the correction term of the frequency converter frequency output by the LSTM model, and the input is the fan wind speed and fan inclination angle θ.

13. The long-span bridge substructure vehicle-fire-wind real-time hybrid test platform according to claim 12, characterized in that, The LSTM time lag compensator body is an LSTM-based neural network, containing a hidden layer of 16-32 neurons, and the input layer nodes contain historical time lag sequences Δt i , system real-time load rate, environmental temperature gradient, and environmental wind speed gradient, where i=(1, 2, …, n), and the output layer is mapped to the next analysis step time lag prediction value Δt through a fully connected layer i+1 , specifically comprising the following steps: S401, construct time delay dataset, collect historical time delay sequence Δt i The system real-time load rate, environmental temperature gradient and wind speed gradient three-dimensional data are constructed into an input vector, which is input into the neural network based on LSTM, wherein Δt i The delay time refers to the delay time from the control command to the sensor measured value reaching the standard. S402, offline training and online optimization, using pre-loading stage data samples to train the LSTM-based time delay compensation model, and the loss function is the mean absolute error; when formally loaded, the network parameters are updated online by the optimizer after each analysis step; S403, dynamic time lag compensation, send control command in advance △t2 time before the second analysis step loading, and record the actual time lag △t real , △t2 refers to the time difference from the first analysis step to load the command to the position. S404, error correction and convergence, based on Δt i-real with the deviation of Δt i , the LSTM-based time delay compensation model weight is corrected by using gradient descent method until the time delay error of continuous multiple analysis steps is ≤1 second; Δt i-real refers to the actual measured time delay of each analysis step, i=(1, 2, …, n).

14. The implementation method of the large-span bridge substructure vehicle-fire-wind real-time hybrid test platform, using the large-span bridge substructure vehicle-fire-wind real-time hybrid test platform according to claims 1-13, characterized in that, The method comprises the following steps: Step 1: making and installing test pieces: the types of test pieces include main cable models, stay cable models and sling models; Install the main cable model or the stay cable model: synchronously lower the two sides of the sliding beam to the base reference surface, and pull the main cable model or the stay cable model through the free-rotation anchoring subsystem and the free-rotation force boundary subsystem in sequence and anchor; then synchronously lift the two sides of the sliding beam to a preset height or angle; Install the sling model: connect the sling model to the main cable model through the cable clamp, and pass the lower end of the sling model through the second through-type jack and anchor, and install force sensors at the front ends of the first through-type jack and the second through-type jack; Step 2: select the test mode and configure the environment; Mode A: open space self-defined fire test, Input the target fire parameters into the fire temperature field fast generation model to generate a non-uniform temperature field and transmit it to the LSTM-PID intelligent control system; Mode B: closed space standard fire test Assemble the enclosing units through the quick-release interface to form a closed space, select the standard fire curve mode in the LSTM-PID intelligent control system, and close the multi-dimensional wind field simulation system; Step 3: pre-load and time delay compensation initialization Start the cable system self-balancing counterforce frame and the T-shaped wind-fire coupling simulation system, pre-load and collect feedback data of each sensor; The least square method is used to calculate initial time delay Δt of each system i , and the input is fed forward to compensate for the LSTM-PID intelligent control system; Step 4: generate the temperature rise curve of the fire scene in the selected mode and execute the real-time hybrid test: Step 4.1, input the fire source power, position and wind speed into the fire temperature field fast generation model to generate a non-uniform temperature field, and transmit it to the LSTM-PID intelligent control system to analyze the burner valve opening degree and the fan frequency command; If mode B is executed, input the standard target temperature rise curve, and synchronously output the burner valve opening degree command; Step 4.2, start the gas supply, and dynamically adjust the gas flow through the LSTM-PID intelligent control system, and synchronously ignite the flame; If mode A is executed, synchronously start the multi-dimensional wind field simulation system and dynamically adjust the fan inclination angle; Step 4.3, the LSTM-based numerical substructure fast calculation model predicts the full-bridge displacement and cable force in real time, and converts it into the loading command of the first through-type jack and the second through-type jack; Step 4.4, load step by step according to the displacement / force control mode, and use the force control mode to slowly stretch to the test piece fracture when the preset limit is exceeded; Step 4.5, LSTM time-lag compensator based on the whole loading process, each analysis step Δt i Predict the next analysis step time-lag Δt i+1 And advance Δt i+1 Send instructions to ensure that the force loading, combustion and wind field are synchronized, where i = (1, 2, …, n); if mode B is executed, only the loading and temperature time-lags are predicted, and it is sufficient to ensure that the loading and combustion are synchronized. Step 5: data acquisition and dynamic visualization, specifically including: Step 5.1, real-time acquisition of test piece axial force, displacement, strain and fire field temperature data, and drawing of related curves in the LSTM-PID intelligent control system; If mode A is executed, synchronously collect the fan wind speed data; Step 5.2, bind the sensor measured data and the predicted data to the full-structure three-dimensional graph nodes through three-dimensional animation software, and dynamically visualize the full-bridge stress, deformation and temperature field distribution based on the numerical-color mapping algorithm; Step 5.3, generate load-displacement curve, temperature-time curve, output damage evolution law and failure warning.

Images