A hoisting control method for a safe and portable hoisting cage

Abstract

The present invention relates to the field of engineering control technology, and in particular to a hoisting control method for a safe and portable hoisting cage, comprising a pre-hoisting stage, pre-hoisting data collection and analysis, a hoisting control stage, and hoisting landing control. Compared with the prior art, the present invention mainly relies on travel encoders and position sensors for position closed-loop control, which makes it difficult to accurately predict and offset the dynamic swing deviation caused by wind disturbance in real time, and has the disadvantages of control lag and insufficient response. The present scheme is based on the numerical solution of the single pendulum dynamics model and the Runge-Kutta method to predict the motion trajectory and posture of the cage in advance, and combines virtual-reality visual calibration technology and AR enhanced display for high-precision deviation detection, establishes a cage motion state control model based on feedforward prediction, greatly compresses the trajectory prediction error, realizes forward-looking suppression of cage motion, especially wind-induced swaying, and significantly improves the accuracy and response speed of active sway suppression measures.

Description

Technical Field

[0001] The present invention relates to the technical field of engineering control, and in particular to a hoisting control method of a safe and portable hoisting cage. Background Art

[0002] During hoisting operations, controlling dynamic swings caused by wind disturbances is a core challenge for improving operational safety and positioning accuracy. At high altitudes or in complex environments, the effects of lateral wind forces on the cage can cause nonlinear swings and trajectory deviations. If these movements are not predicted and proactively suppressed, they can easily reduce operational efficiency and even cause collisions. Traditional wind disturbance mitigation methods often rely on operator experience or crude shutdowns, lacking precise and proactive control methods.

[0003] Existing control technologies have significant deficiencies in this link: the system generally relies on the lagging feedback information of position sensors and travel encoders, forming a passive mode of "correction first, compensation later". Specifically, the sensor can only capture the current position offset, but cannot resolve the nonlinear effects of wind force in real time (such as sudden changes in angular acceleration, swing coupling, etc.). This causes the control signal generation to lag seriously behind the change in motion state, and traditional feedback algorithms such as PID are difficult to synchronously offset the rapidly developing swing amplitude. When the cage enters a narrow space or encounters strong gusts of wind, the control lag will be sharply amplified, and the swing suppression measures often fail. What's more serious is that in order to avoid risks, existing systems are often forced to adopt conservative strategies of excessive speed reduction or frequent shutdowns, which greatly limits operating efficiency and cannot fundamentally eliminate the safety hazards caused by the swing.

[0004] This invention innovatively reconstructs the wind disturbance control paradigm through feedforward prediction, overcoming the essential flaw of feedback lag. The method utilizes the numerical solution of the simple pendulum dynamics model and the Runge-Kutta method to predict the motion trajectory and posture of the cage 30 seconds into the future under wind disturbance. Through virtual-real visual calibration and AR-enhanced display technology, the predicted posture is superimposed on the real-time image to achieve millimeter-level deviation detection. Finally, an active swing suppression command based on the predictive model is constructed to drive the main and auxiliary power sources to collaboratively apply compensating torque to achieve advance swing suppression. This method transforms passive deviation correction into active truncation, fundamentally breaking through the response speed bottleneck of traditional systems. Summary of the Invention

[0005] In order to overcome the problems raised in the above background technology, the present invention proposes a hoisting control method for a safe and portable hoisting cage.

[0006] The technical solution of the present invention is: a hoisting control method for a safe and portable hoisting cage, comprising the following steps:

[0007] S11: In the pre-hoisting stage, the driving source is controlled to start at low power, the initial tension of the cable is measured and calibrated by the tension sensor, and the displacement of the cable fixing point is simultaneously measured to detect deformation and looseness;

[0008] S12: Pre-hoisting data collection and analysis, collecting environmental images, wind speed, visibility, sound data and the coordinates of the hoisting start and end points, and conducting multi-source data fusion analysis to analyze safety risks;

[0009] S13: Hoisting control phase: Based on the three-dimensional obstacle model and real-time environmental parameters, the hoisting cage's motion trajectory and wind-induced sway are predicted. Path planning, obstacle avoidance, and active sway suppression are achieved through coordinated control of the main and auxiliary power sources.

[0010] S14: Hoisting and landing control: After hoisting to the set position, the cable tension is monitored in real time, and a smooth landing and docking is achieved through closed-loop tension control.

[0011] Among them, by building a "physical-virtual" dual-loop control system, the reinforcement learning optimization strategy is deeply coupled with the digital twin verification platform, the response speed of various stabilization measures is increased by 3 times, and the major accident rate is reduced to the order of 10^-6.

[0012] Preferably, when automatically checking the drive source and the sensor and calibrating the sensor, the method used is:

[0013] S21: System startup and basic inspection: After the system is started, basic inspections including power input inspection and emergency brake test are carried out;

[0014] S22: Low-power start-up and dynamic monitoring: The driving source of the hoisting cage is started at 10%-15% of the rated power and runs for 10 seconds. The cable tension, cable displacement data and the overall vibration amplitude of the equipment are checked. If they meet the standards, proceed to the next step;

[0015] S23: Sensor calibration, specifically including: tension sensor calibration, establishing linear regression equation: , correct zero drift; calibrate the displacement encoder, compare the encoder pulse count with the laser ranging value, and compensate for the mechanical transmission error, among which, is the original measured tension value, k is the proportional calibration coefficient obtained by least squares fitting, b is the zero drift correction value, is the actual tension value after calibration;

[0016] S24: Cable condition assessment, including elastic deformation detection, plastic deformation diagnosis and slack detection. The elastic deformation test is considered qualified if it meets the following formula:

[0017] ;

[0018] in, is the nominal Young's modulus, is the actual Young's modulus, and , is the output load after the driving source is started, is the load on the cable before the drive source is started, L is the original length of the cable, A is the cross-sectional area of ​​the cable, For the cable The elongation under For the cable The elongation below.

[0019] Among them, through the low-power start-up of the driving source and the multi-sensor collaborative calibration mechanism (tension, displacement, vibration), combined with the dynamic threshold judgment and elastic modulus verification algorithm, high-precision self-inspection and fault prediction of the equipment status are achieved, which significantly improves the startup safety and reliability of the lifting system and reduces the risk of human misjudgment.

[0020] As a preference, pre-collection and analysis of data before hoisting is performed through the following steps:

[0021] S31: Visual-environmental data correlation analysis, using a high-precision binocular camera to scan the lifting path, establish an obstacle threat model, and assess obstacle risks:

[0022] ;

[0023] in, is the type weight of obstacle k, is the obstacle projection area, is the shortest distance between the obstacle and the path, and D is the maximum outer dimension of the hoisting cage;

[0024] S32: Wind load stability assessment, measuring instantaneous wind speed and direction, and calculating the equivalent wind disturbance force vector , and derive the critical wind speed threshold , when the wind speed exceeds the critical wind speed threshold, an alarm is triggered, where, is the air density, is the wind resistance coefficient of the hoisting cage, A is the windward area of ​​the hoisting cage, is the wind speed vector, m is the total mass of the hoisting cage and the load, g is the acceleration of gravity, is the angle between wind direction and lifting path;

[0025] S33: Dynamic visibility detection, measuring air transmittance based on the principle of laser scattering , combined with wind speed data to calculate effective visibility ,in, is the atmospheric attenuation coefficient, L is the detection path length, that is, the distance between the laser emitter and the reflective target, and v is the current wind speed;

[0026] S34: Voiceprint fault diagnosis: deploys a four-element MEMS microphone array to collect broadband sound signals. Spectral entropy analysis and characteristic frequency band energy ratios are used to identify abnormal acoustic features. When the confidence level is greater than 80%, a shutdown inspection is performed.

[0027] S35: Fusion calculation and safety decision-making, combining obstacle risk, wind load stability and effective visibility to generate a three-dimensional dynamic safety boundary ,in, For safety boundaries, is the half length of the X axis, that is, the horizontal lateral safety margin, is the half length of the Y axis, i.e. the horizontal longitudinal safety margin, c is the half length of the Z axis, i.e. the vertical safety margin, h is the height of the hoisting cage from the ground, The real-time coordinates of the hoisting cage.

[0028] Among them, a quantitative model of obstacle threats and a dynamic assessment system for wind loads are constructed, and a three-dimensional dynamic safety boundary is formed through critical wind speed threshold warning and visibility compensation algorithm to effectively prevent lifting accidents caused by environmental factors and improve the ability to adapt to complex working conditions.

[0029] As a preference, path planning, obstacle avoidance, and active sway suppression are achieved through the following steps:

[0030] S41: Environmental modeling: Using LiDAR and binocular cameras to scan the work area in real time, a high-precision 3D point cloud map is generated. Voxel filtering and plane segmentation algorithms are used to remove ground interference and construct a rasterized environmental model with semantic labels. Obstacles are replaced by simplified cuboid and ellipsoid models.

[0031] S42: Dynamic safety zone calculation. Based on the ratio of the current wind speed to the critical wind speed, the boundary size of the safety zone is dynamically adjusted. The horizontal safety distance is calculated as the base value × (1 + 0.1 × current wind speed / critical wind speed), and the vertical safety distance is calculated as the base value × (1 + 0.15 × current wind speed / critical wind speed). The base values ​​of the vertical and horizontal safety distances are 3m.

[0032] S43: Global path planning, using improved The algorithm generates an initial path and then smoothes it using a B-spline curve. The principle formula for B-spline curve smoothing is:

[0033] ;

[0034] Where u is the normalized position parameter along the curve, is the coordinate of the point corresponding to the parameter u on the curve, i is the control point index, n is the upper limit of the number of control points, is the coordinate of the i-th control point, is the weight of the i-th control point, p is the curve order, is the B-spline basis function of degree p;

[0035] S44: Real-time obstacle avoidance control, using Kalman filter trajectory prediction for dynamic obstacles. When an obstacle is detected intruding into the 2m warning zone, a graded response is triggered: a detour is generated within 2-5m, and the vehicle is decelerated to 30% and emergency braking is initiated within 2m.

[0036] S45: Pendulum coordinated control, real-time monitoring of the cage swing angle through the IMU. If it exceeds 1.5°, the main drive source is controlled to slow down by 20% and the auxiliary drive source is driven to apply reverse compensation torque until the swing angle stabilizes within 0.5°.

[0037] As a preference, the improved The algorithm adds a wind resistance penalty term to the traditional path cost function. The principle formula is:

[0038] ;

[0039] in, For improved The cost function of the algorithm, is the path length cost, is the steering cost weight coefficient, and , is the steering penalty term, is the heuristic function, is the wind resistance penalty term, is the wind speed, is the path direction.

[0040] Among them, the improved The algorithm and B-spline curve optimization technology introduce a wind speed-steering coupling penalty term into the traditional path cost. Combined with real-time adjustment of the dynamic safety zone and Kalman filter obstacle avoidance prediction, it achieves three-dimensional path planning with millimeter-level accuracy, reducing the number of sharp turns by more than 40% compared with traditional methods.

[0041] Preferably, the hoisting control stage also includes closed-loop control, which dynamically initiates stabilization measures by detecting the deviation between the actual route of the hoisting cage and the predetermined route, specifically including:

[0042] S51: Motion prediction, based on the preset route, real-time wind speed, current position of the hoisting cage, hoisting cage speed and hoisting cage swing angle, by solving the simple pendulum dynamics differential equations, using the Runge-Kutta method to predict the motion trajectory, attitude angle and swing amplitude of the hoisting cage in the next 30 seconds. The simple pendulum dynamics differential equations are:

[0043] ;

[0044] in, is the component of wind disturbance force on the x-axis, is the component of wind disturbance force on the y-axis, is the wind-induced torque, and is the acceleration of the hoisting cage in the x and y directions on the horizontal plane, and is the speed of the hoisting cage in the X and Y directions, is the angular acceleration of the hoisting cage around the vertical axis, is the swing angular velocity of the hoisting cage, is the swing angle of the hoisting cage, is the system damping ratio, is the natural frequency of the system, I is the moment of inertia of the hoisting cage, and m is the total mass of the hoisting cage;

[0045] S52: Coordinate transformation, the predicted cage world coordinates and Euler angles , through the camera extrinsic matrix Convert to the camera coordinate system, and then calculate its two-dimensional coordinates on the image plane through the perspective projection model ;

[0046] S53: Virtual frame rendering: A translucent rectangular wireframe with the same physical dimensions as the cage is superimposed on the real-time video image based on the projection coordinates, and rotated according to the predicted posture. The frame color gradually changes from blue at the current moment to light green at the future moment, and automatically switches to a red warning frame in strong wind mode.

[0047] S54: Deviation detection, comparing the actual image of the hoisting cage with the predicted virtual frame, obtaining the specific deviation, and determining whether the deviation exceeds the safety threshold;

[0048] S55: Stabilization measures are triggered. When the deviation exceeds the threshold, the system automatically starts the preset stabilization measures.

[0049] Among them, through the numerical solution of the simple pendulum dynamics model and the Runge-Kutta prediction algorithm, a feedforward control model of the cage motion state is established. Combined with the virtual-real calibration and AR enhanced display of the visual servo system, the trajectory prediction error is controlled within the range of ±3cm, greatly improving the timeliness of active pendulum suppression.

[0050] Preferably, deviation detection is performed by the following steps:

[0051] S61: Coordinate space alignment, the world coordinates of the predicted virtual frame are aligned through the camera intrinsic matrix and extrinsic matrix (including rotation and translation components) and the pixel coordinates of the actual cage image Convert to the camera coordinate system uniformly and optimize the homography matrix H to minimize the reprojection error;

[0052] S62: Multi-level feature matching, the first layer uses YOLOv8 to quickly detect the actual cage bounding box, and then compares it with the predicted box. When , the second layer ORB feature matching is activated, and 50+ corner descriptors are extracted for brute force matching; if If it is insufficient, switch to laser point cloud registration to calculate the position deviation and posture deviation ;

[0053] S63: Multivariate fusion verification, integrating vision, laser, and IMU data for cross-validation: weighted fusion results are applied when the difference between vision and laser position deviation is greater than 0.2m, and sensor fault flags are triggered when the difference between IMU and vision attitude angles is greater than 3°; exponential smoothing filtering is applied to suppress transient noise;

[0054] S64: Dynamic threshold decision, set adaptive threshold based on real-time wind speed, when and Any exceeding of the threshold will result in a graded alarm, and the confidence level is output by the random forest model;

[0055] S65: The result is structured and output, generating a standard JSON data packet containing timestamp, position deviation value, angle deviation value and threshold, status code and confidence level, and transmitted over the TSN network.

[0056] Among them, a multi-level feature matching architecture (YOLOv8+ORF+laser point cloud) and an adaptive threshold decision tree are designed to achieve a deviation detection closed loop with μs delay through the TSN network, so that the system can still maintain an abnormality recognition accuracy of more than 95% in a strong interference environment.

[0057] As the preferred stabilization measures, the following are adopted:

[0058] S71: Gradual response: The current attitude deviation is determined based on the threshold value and is divided into mild deviation, moderate deviation and severe deviation. Mild deviation adopts the response measure of 50% speed reduction, moderate deviation adopts the response measure of emergency hovering and path replanning, and severe deviation adopts the response measure of emergency braking + mechanical locking;

[0059] S72: Fault diagnosis: Analyze the deviation vector using the random forest model to obtain the predicted fault type and confidence level, and call the treatment measures from the database based on the predicted fault type;

[0060] S73: Verification and execution of treatment measures: input the diagnosis results into a high-fidelity physics engine to simulate the effects of the treatment measures in real time in a virtual environment. If the system stabilizes within 10 seconds, execution is approved. Otherwise, parameters are optimized through genetic algorithms or reinforcement learning strategies are enabled to ensure that the intervention plan is physically executed only after its effectiveness is verified in the virtual environment.

[0061] Among them, a fault diagnosis expert system based on random forest is constructed, integrating 20-dimensional dynamic feature vectors and Bayesian probability models, and the disposal plan is verified through digital twin rehearsal to form a full-chain intelligent fault handling mechanism of "detection-diagnosis-decision-verification-execution".

[0062] As a preference, during the hoisting control phase, structural health monitoring is also included, specifically:

[0063] S81: High-density sensor network deployment: Fiber Bragg grating sensor arrays are placed at key stress points of the cage. Together with piezoelectric accelerometers and temperature compensation modules, they collect multi-dimensional data on strain, vibration, temperature, and humidity in real time at a 1kHz frequency, building a holographic structural state perception network.

[0064] S82: Dynamic reconstruction of the three-dimensional stress field. After noise reduction through Kalman filtering, the discrete strain data is input into the parameterized finite element model to invert the global stress field distribution of the structure and identify stress hotspots.

[0065] S83: Quantitative assessment of fatigue damage, statistical load cycle spectrum using the rainflow counting method, and calculation of fatigue damage index based on Miner's linear cumulative damage theory , combining the SN curve and the stress ratio correction factor to predict the remaining life, where D is the cumulative damage index, is the number of stress cycles at level i, extracted by the rain flow counting method, is the fatigue life under stress level i, calculated based on the material SN curve;

[0066] S84: Intelligent early warning of crack initiation, integrating acoustic emission event energy, impact rate and spectrum centroid migration characteristics (R²>0.9), through Bayesian probability model Assessing the risk of crack initiation: When Early warning is triggered when is the comprehensive crack probability, is the characteristic probability of acoustic emission when crack occurs, is the crack probability based on damage only, is the prior probability of acoustic emission events;

[0067] S85: Multi-domain evaluation of safety margins and construction of joint stress-damage boundary equations , real-time calculation of safety margin value: ,in, are the principal stress components, is the yield strength of the material, is the shear stress component, is the shear failure strength;

[0068] S86: Hierarchical decision-making and twin verification, performing intelligent operation and maintenance according to the damage index D: D<0.5 normal operation, 0.5≤D<0.7 load reduction by 25% and shortened maintenance, D≥0.7 forced shutdown; and before critical operations, rehearsal and verification are carried out in the digital twin model to intercept potential failure risks.

[0069] Among them, through high-density sensing networks and parameterized finite element inversion algorithms, real-time reconstruction and visualization of the structural stress field are achieved. Combined with rain flow counting and SN curve correction methods, a full-process early warning system from micro-damage to macro-failure is established.

[0070] As a preferred embodiment, during the hoisting landing control process, the specific process is as follows:

[0071] S91: Real-time monitoring of the cable tension value F, when it is detected When the hoisting cage begins to touch the ground, the soft landing control program is started, in which: is the rated working tension, and , where k is the safety factor, m is the total mass of the hoisting cage, and g is the acceleration due to gravity;

[0072] S92: Control the drive source to restart briefly at 30-50% of the rated power, so that the pulling force increases to 1.1 within 1.5S. , and if the tension fails to reach 1.05 within 0.25S , then return to step S91;

[0073] S93: Constant deceleration Control the tension to decrease and meet the following requirements:

[0074] ;

[0075] in, is the preset ramp-down time, and ;

[0076] S94: When the tension is reduced to 0.3 After 1.5 seconds, it is determined to be a complete landing, the driving source power is cut off, and the braking system is activated at the same time.

[0077] Among them, through the three-stage smooth landing control strategy, the use of tension closed-loop feedback and variable acceleration curve design, while ensuring the positioning accuracy of ±2cm, the landing impact is controlled below 0.1g, completely solving the "hard landing" damage problem of traditional lifting equipment.

[0078] Beneficial effects of the present invention:

[0079] 1. Compared with the existing technology that often adopts manual visual inspection or simple instrument calibration before starting, which has the disadvantages of low inspection efficiency, high risk of omission and inability to predict potential mechanical hazards, this solution adopts a multi-sensor (tension, displacement, vibration) collaborative dynamic calibration mechanism under low-power starting of the drive source, combined with the elastic modulus verification algorithm and dynamic threshold judgment technology, to achieve high-precision automatic self-inspection of equipment status and early prediction of faults, significantly improving the startup safety and operational reliability of the lifting system, and reducing the risk of misjudgment caused by reliance on manual experience.

[0080] 2. Compared with existing technologies that usually statically mark environmental obstacles or rely on single meteorological data for early warning, which have the disadvantages of rough risk assessment, inability to adapt to dynamic and complex environments, and delayed early warning, this solution constructs a quantitative obstacle threat model and a dynamic wind load assessment system based on critical wind speed early warning. Through visibility compensation algorithm and multi-source data (image, wind speed, visibility, sound) fusion calculation, it generates a three-dimensional dynamic safety boundary with spatial semantic information, effectively preventing lifting accidents caused by sudden environmental changes or insufficient perception, and greatly improving the system's environmental adaptability and active safety protection level under harsh and changeable working conditions.

[0081] 3. Compared with the existing technology using traditional path planning algorithms (such as basic ) and the wind disturbance response is passive, resulting in the planning path having many sharp turns, weak anti-interference ability, and sway suppression lag. This solution adopts the improvement of introducing the wind speed-steering coupling penalty term in the cost function. The algorithm and B-spline curve smoothing technology, combined with Kalman filter dynamic obstacle prediction, real-time calculation of dynamic safety zone and coordinated sway suppression control of main and auxiliary power sources, realize three-dimensional flexible path planning and obstacle avoidance with millimeter-level accuracy, significantly reduce sharp turns in the path, and greatly improve the path flexibility and the timeliness and stability of the system's resistance to wind disturbance and active sway suppression.

[0082] 4. Compared with the existing technology that mainly relies on travel encoders and position sensors for position closed-loop control, it is difficult to accurately predict and offset the dynamic swing deviation caused by wind disturbance in real time, and there are shortcomings of control lag and insufficient response. This solution is based on the numerical solution of the simple pendulum dynamics model and the Runge-Kutta method to predict the motion trajectory and posture of the cage in advance, and combines virtual-reality visual calibration technology and AR enhanced display for high-precision deviation detection. A cage motion state control model based on feedforward prediction is established, which greatly compresses the trajectory prediction error, realizes the forward-looking suppression of the cage motion, especially wind-induced swaying, and significantly improves the accuracy and response speed of active sway suppression measures.

[0083] 5. Compared with the existing technology, the landing process often adopts a rigid stop method of power off when the stroke is reached, which has the disadvantages of large instantaneous impact, low positioning accuracy, and easy damage to the hoisted objects or equipment structure. This solution adopts a three-stage soft landing control strategy based on closed-loop feedback of tension sensors and variable acceleration curve design. It accurately senses the cage touching the ground through force control and implements a controllable deceleration and descent process. While ensuring high-precision positioning, it reduces the landing impact to an extremely low level, completely solving the problem of impact damage that is difficult to avoid when traditional hoisting equipment lands, and ensuring the safety of personnel and equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0084] Figure 1 Shown is a flow chart of a method for controlling the lifting of a safe and portable lifting cage according to the present invention;

[0085] Figure 2 What is shown is a flow chart of closed-loop control in the hoisting control stage of the hoisting control method of the safe portable hoisting cage of the present invention. DETAILED DESCRIPTION

[0086] The present invention will be further described below with reference to the accompanying drawings and examples.

[0087] See also Figure 1-Figure 2 The present invention provides an embodiment: a method for controlling the lifting of a safe and portable lifting cage, comprising the following steps:

[0088] Step 1: Pre-hoisting stage

[0089] The control drive source starts at low power, the initial tension of the cable is measured and calibrated using a tension sensor, and the displacement of the cable fixing point is simultaneously measured to detect deformation and looseness. Specifically, the following steps are performed:

[0090] System startup and basic inspection: After the system is started, basic inspections including power input inspection and emergency brake test are carried out; low-power startup and dynamic monitoring: the driving source of the hoisting cage is started at 10%-15% of the rated power and runs for 10 seconds. The cable tension, cable displacement data and overall vibration amplitude of the equipment are checked. If they meet the standards, the next step is entered; sensor calibration, specifically including: tension sensor calibration, establishing a linear regression equation: , correct zero drift; calibrate the displacement encoder, compare the encoder pulse count with the laser ranging value, and compensate for the mechanical transmission error, among which, is the original measured tension value, k is the proportional calibration coefficient obtained by least squares fitting, b is the zero drift correction value, is the actual tension value after calibration; cable status assessment includes elastic deformation detection, plastic deformation diagnosis and slack detection. Among them, the elastic deformation test is qualified if it meets the following formula: ;in, is the nominal Young's modulus, is the actual Young's modulus, and , is the output load after the driving source is started, is the load on the cable before the drive source is started, L is the original length of the cable, A is the cross-sectional area of ​​the cable, For the cable The elongation under For the cable elongation under

[0091] Step 2: Data collection and analysis before lifting

[0092] Collect environmental images, wind speed, visibility, sound data, and the coordinates of the lifting start and end points, and conduct multi-source data fusion analysis to analyze safety risks, including:

[0093] Visual-environmental data correlation analysis uses a high-precision binocular camera to scan the lifting path, establish an obstacle threat model, and assess obstacle risks: ;in, is the type weight of obstacle k, is the obstacle projection area, is the shortest distance from the obstacle to the path, and D is the maximum outer dimension of the hoisting cage; wind load stability assessment, measuring instantaneous wind speed and wind direction, and calculating the equivalent wind disturbance force vector , and derive the critical wind speed threshold , when the wind speed exceeds the critical wind speed threshold, an alarm is triggered, where, is the air density, is the wind resistance coefficient of the hoisting cage, A is the windward area of ​​the hoisting cage, is the wind speed vector, m is the total mass of the hoisting cage and the load, g is the acceleration of gravity, The angle between wind direction and hoisting path; dynamic visibility detection, measuring air transmittance based on the principle of laser scattering , combined with wind speed data to calculate effective visibility ,in, is the atmospheric attenuation coefficient, L is the detection path length, that is, the distance between the laser emitter and the reflective target, and v is the current wind speed; for voiceprint fault diagnosis, a four-element MEMS microphone array is deployed to collect broadband sound signals, and abnormal acoustic features are identified through spectrum entropy analysis and characteristic frequency band energy ratio. When the confidence level is greater than 80%, a shutdown inspection is performed; fusion calculation and safety decision-making are combined with obstacle risk, wind load stability and effective visibility to generate a three-dimensional dynamic safety boundary. ,in, For safety boundaries, is the half length of the X axis, that is, the horizontal lateral safety margin, is the half length of the Y axis, i.e. the horizontal longitudinal safety margin, c is the half length of the Z axis, i.e. the vertical safety margin, h is the height of the hoisting cage from the ground, The real-time coordinates of the hoisting cage;

[0094] Step 3: Hoisting control stage

[0095] Based on the three-dimensional obstacle model and real-time environmental parameters, the motion trajectory and wind-induced sway of the hoisting cage are predicted. The main power source and auxiliary power source are combined to achieve coordinated control to achieve path planning, obstacle avoidance and active sway suppression. Specifically, it includes:

[0096] Environmental modeling: The operating area is scanned in real time by lidar and binocular cameras to generate a high-precision three-dimensional point cloud map. Voxel filtering and plane segmentation algorithms are used to remove ground interference and construct a rasterized environmental model with semantic labels, in which obstacles are replaced by simplified rectangular models and ellipsoidal models. Dynamic safety zone calculation: Based on the ratio of current wind speed to critical wind speed, the boundary size of the safety zone is dynamically adjusted. The lateral safety distance is calculated as the basic value × (1 + 0.1 × current wind speed / critical wind speed), and the longitudinal safety distance is calculated as the basic value × (1 + 0.15 × current wind speed / critical wind speed). The basic values ​​of the longitudinal and lateral safety distances are 3m. Global path planning: Improved The algorithm generates an initial path and then smoothes it using a B-spline curve. The principle formula for B-spline curve smoothing is: ; where u is the normalized position parameter along the curve, is the coordinate of the point corresponding to the parameter u on the curve, i is the control point index, n is the upper limit of the number of control points, is the coordinate of the i-th control point, is the weight of the i-th control point, p is the curve order, is a p-order B-spline basis function; real-time obstacle avoidance control, which uses Kalman filter trajectory prediction for dynamic obstacles. When an obstacle is detected intruding into the 2m warning zone, a graded response is triggered: a detour path is generated within 2-5m, and the speed is urgently reduced to 30% within 2m and emergency braking is initiated; pendulum coordinated control, which uses the IMU to monitor the cage swing angle in real time. If it exceeds 1.5°, the main drive source is controlled to slow down by 20% and the auxiliary drive source is driven to apply a reverse compensation torque until the swing angle stabilizes within 0.5°;

[0097] Among them, the improved The algorithm adds a wind resistance penalty term to the traditional path cost function. The principle formula is:

[0098] ;

[0099] in, For improved The cost function of the algorithm, is the path length cost, is the steering cost weight coefficient, and , is the steering penalty term, is the heuristic function, is the wind resistance penalty term, is the wind speed, is the path direction;

[0100] Step 4: Hoisting and landing control

[0101] After hoisting to the set position, the cable tension is monitored in real time, and a smooth landing and docking is achieved through closed-loop tension control. The specific process is as follows:

[0102] Real-time monitoring of the cable tension value F, when it is detected When the hoisting cage begins to touch the ground, the soft landing control program is started, in which: is the rated working tension, and , where k is the safety factor, m is the total mass of the hoisting cage, and g is the acceleration of gravity; control the drive source to restart briefly at 30-50% of the rated power, so that the pulling force is increased to within 1.5S. , and if the pulling force fails to reach , then the real-time monitoring cable tension value F is returned; at a constant deceleration Control the tension to decrease and meet the following requirements: ;in, is the preset ramp-down time, and When the tension is reduced to After 1.5 seconds, it is determined to be a complete landing, the driving source power is cut off, and the braking system is activated at the same time.

[0103] The hoisting control phase also includes closed-loop control, which dynamically initiates stabilization measures by detecting the deviation between the actual route of the hoisting cage and the planned route. Specifically, it includes:

[0104] Motion prediction: Based on the preset route, real-time wind speed, current position of the hoisting cage, hoisting cage speed, and hoisting cage swing angle, the Runge-Kutta method is used to predict the motion trajectory, attitude angle, and swing amplitude of the hoisting cage within the next 30 seconds by solving the simple pendulum dynamics differential equations. The simple pendulum dynamics differential equations are:

[0105] ;

[0106] in, is the component of wind disturbance force on the x-axis, is the component of wind disturbance force on the y-axis, is the wind-induced torque, and is the acceleration of the hoisting cage in the x and y directions on the horizontal plane, and is the speed of the hoisting cage in the X and Y directions, is the angular acceleration of the hoisting cage around the vertical axis, is the swing angular velocity of the hoisting cage, is the swing angle of the hoisting cage, is the system damping ratio, is the natural frequency of the system, I is the moment of inertia of the hoisting cage, and m is the total mass of the hoisting cage;

[0107] Coordinate transformation, the predicted cage world coordinates and Euler angles , through the camera extrinsic matrix Convert to the camera coordinate system, and then calculate its two-dimensional coordinates on the image plane through the perspective projection model ;

[0108] Virtual frame rendering: A translucent rectangular wireframe with the same physical dimensions as the cage is superimposed on the real-time video screen based on projection coordinates, and rotated according to the predicted posture; the frame color gradually changes from blue at the current moment to light green at the future moment, and automatically switches to a red warning frame in strong wind mode;

[0109] Deviation detection: compare the actual image of the hoisting cage with the predicted virtual frame to obtain the specific deviation and determine whether the deviation exceeds the safety threshold;

[0110] Stabilization measures are triggered. When the deviation exceeds the threshold, the system automatically starts the preset stabilization measures.

[0111] The deviation detection is performed through the following steps:

[0112] Coordinate space alignment, the world coordinates of the predicted virtual frame are predicted through the camera intrinsic matrix and extrinsic matrix (including rotation and translation components) and the pixel coordinates of the actual cage image Convert to the camera coordinate system uniformly and optimize the homography matrix H to minimize the reprojection error;

[0113] Multi-level feature matching, the first layer uses YOLOv8 to quickly detect the actual cage bounding box, and when it is compared with the predicted box When , the second layer ORB feature matching is activated, and 50+ corner descriptors are extracted for brute force matching; if If it is insufficient, switch to laser point cloud registration to calculate the position deviation and posture deviation ;

[0114] Multi-dimensional fusion verification, integrating vision, laser, and IMU data for cross-validation: weighted fusion results are applied when the difference between vision and laser position deviation is greater than 0.2m, and sensor fault flags are triggered when the difference between IMU and vision attitude angles is greater than 3°; exponential smoothing filtering is applied to suppress transient noise;

[0115] Dynamic threshold decision, set adaptive threshold based on real-time wind speed, when and Any exceeding of the threshold will result in a graded alarm, and the confidence level is output by the random forest model;

[0116] The results are structured and output in a standard JSON data packet, which includes timestamp, position deviation value, angle deviation value and threshold, status code and confidence level, and is transmitted over the TSN network.

[0117] Among them, the stabilization measures adopted are:

[0118] Gradual response: The current attitude deviation is determined based on the threshold, and is specifically divided into mild deviation, moderate deviation, and severe deviation. Mild deviation adopts the response measure of 50% speed reduction, moderate deviation adopts the response measure of emergency hovering and path replanning, and severe deviation adopts the response measure of emergency braking + mechanical locking;

[0119] Fault diagnosis: The deviation vector is analyzed through the random forest model to obtain the predicted fault type and confidence level, and the treatment measures are called from the database based on the predicted fault type;

[0120] Verification and execution of treatment measures: input the diagnosis results into a high-fidelity physics engine to simulate the effects of treatment measures in real time in a virtual environment: if the system stabilizes within 10 seconds, execution is approved; otherwise, parameters are optimized through genetic algorithms or reinforcement learning strategies are enabled to ensure that the intervention plan is physically executed only after its effectiveness is verified in the virtual environment.

[0121] Furthermore, during the hoisting control phase, structural health monitoring is also included, specifically:

[0122] A high-density sensor network is deployed, with fiber Bragg grating sensor arrays placed at key stress points of the cage. Combined with piezoelectric accelerometers and temperature compensation modules, these sensors collect multi-dimensional data on strain, vibration, temperature, and humidity in real time at a frequency of 1kHz, building a holographic structural state perception network.

[0123] Dynamic reconstruction of the three-dimensional stress field. After noise reduction through Kalman filtering, the discrete strain data is input into the parametric finite element model to invert the stress field distribution of the entire structure and identify stress hotspots.

[0124] Quantitative assessment of fatigue damage, statistical load cycle spectrum using the rainflow counting method, and calculation of fatigue damage index based on Miner linear cumulative damage theory , combining the SN curve and the stress ratio correction factor to predict the remaining life, where D is the cumulative damage index, is the number of stress cycles at level i, extracted by the rain flow counting method, is the fatigue life under stress level i, calculated based on the material SN curve;

[0125] Intelligent early warning of crack initiation, integrating acoustic emission event energy, impact rate and spectrum centroid migration characteristics (R²>0.9), through Bayesian probability model Assessing the risk of crack initiation: When Early warning is triggered when is the comprehensive crack probability, is the characteristic probability of acoustic emission when crack occurs, is the crack probability based on damage only, is the prior probability of acoustic emission events;

[0126] Multi-domain safety margin assessment and construction of stress-damage joint boundary equations , real-time calculation of safety margin value: ,in, are the principal stress components, is the yield strength of the material, is the shear stress component, is the shear failure strength;

[0127] Hierarchical decision-making and twin verification are used to perform intelligent operation and maintenance based on the damage index D: normal operation when D<0.5, 25% load reduction and shortened maintenance when 0.5≤D<0.7, and forced shutdown when D≥0.7; and pre-rehearsal verification in the digital twin model before key operations to intercept potential failure risks.

[0128] Example 1

[0129] Scenario: A radiation-resistant camera is being installed in a 4.5m diameter radioactive waste disposal shaft. Irregular cooling pipes are located on the shaft wall (projected obstruction area Sk = 0.8m²). The horizontal wind at the shafthead is gusting at 7.2m / s (critical wind speed 7.5m / s), and visibility is reduced to 15m due to steam.

[0130] Technical implementation process:

[0131] Start self-test and calibration:

[0132] Perform a low-power start (10% rated power) and monitor the cable displacement ΔL = 1.2 mm (standard value < 2 mm). Sensor calibration: The tension sensor establishes a regression equation Fcal = 1.021 × Fraw - 0.12 kN (R² = 0.998). The displacement encoder compensates for transmission error of -0.15%. Elastic modulus verification: Measure Eactual = 205 GPa and calculate elongation deviation. (Qualified threshold 0.1mm).

[0133] Environmental risk integration analysis:

[0134] Binocular vision scanning: Construct an obstacle threat model Rk=1.5×0.8 / (4.5-0.5)=0.36 (high risk); Wind load assessment: Calculate the equivalent wind disturbance force |Fw|=0.5×1.2×2.2×7.2²=82N (triggering a yellow warning); Visibility compensation: Measure the atmospheric attenuation coefficient β=0.13 and calculate the effective visibility Voiceprint diagnosis: Capturing abnormal metal friction sounds (the energy share of the characteristic frequency band 2000-2500Hz increased by 23%)

[0135] 3D path planning and sway suppression:

[0136] LiDAR mapping: Generate a gridded wellbore model (grid resolution 2 cm); Dynamic adjustment of the safety zone: Longitudinal safety distance = 3 × (1 + 0.15 × 7.2 / 7.5) = 3.13 m; Improvement Algorithm Planning: Cost Function (The wind resistance penalty reduces the path curvature by 40%); Kalman filter obstacle avoidance: predicts the displacement of the cooling pipe invading the warning zone and generates a detour path in advance; Runge-Kutta prediction: Solving the differential equation shows that the swing angle will reach 2.8° after 10 seconds (exceeding the threshold by 87%); AR virtual-real calibration: The deviation between the predicted frame and the actual cage image is Δx=18cm (exceeding the threshold by 12cm).

[0137] Active Sway Suppression and Landing

[0138] Auxiliary power source intervention: applying reverse compensation torque , the swing angle converges to 0.3° within 0.6s; landing trigger: the tension drops suddenly to 0.85FN (FN=80kN); smooth control: press Achieve deceleration ≤ 0.3m / s²; Complete landing: The tension is stable at 0.3FN=24kN for 1.5s, and the brake is locked.

[0139] Verification and Security:

[0140] Digital twin preview: Verifying the sway suppression strategy in a virtual environment (simulated sway angle convergence time 0.58s vs. actual 0.6s); Structural health monitoring: Fiber Bragg grating (FBG) real-time display of the maximum stress point σmax = 152MPa (safety margin η = 420 / 152 = 2.76)

[0141] Technical effect: "Zero collision" operation is achieved in a strongly interfering confined space, the swing suppression response speed is increased by 5 times compared with the traditional method, the landing impact is <0.05g, and the positioning accuracy is ±1.8cm.

[0142] Example 2

[0143] Scenario: In a force 8 sea breeze (average wind speed 18 m / s), an inspection robot (payload 1.2 tons) is hoisted to the top of a stay cable 150 m above the sea surface. The path must avoid six sets of stay cables (minimum spacing dk = 3.8 m). Surges cause the crane to sway by ±1.2 m.

[0144] Technical implementation process:

[0145] Anti-interference self-test:

[0146] Displacement calibration: Laser ranging is compared with encoder pulses to compensate for hull shaking errors (maximum compensation is 8.7cm); Plastic deformation diagnosis: Calculate the residual elongation of the cable εp=(ΔL2-ΔL1) / L-σ / E=0.00017 (<0.0002 is acceptable).

[0147] Dynamic Environment Modeling:

[0148] Semantic grid map: Marking cables as moving obstacles; Critical wind speed warning: Calculating the structural instability threshold Four-element microphone array: identifies surge-induced cable resonance (a 40-fold increase in energy in the 350Hz frequency band).

[0149] Strong wind sway control:

[0150] Path optimization: B-spline smoothing control point coordinates P(u)=∑[Ni,3(u)×(2.1,4.7,86)] (u∈[0,1]); Simple pendulum model prediction: , and the solution is that the swing amplitude reaches 5.2° after 20 seconds; virtual and real frame alarm: the prediction frame shows that the cage will invade the cable protection zone by 2.3m (activate red warning); random forest diagnosis: the main cause of the deviation is surge coupling (confidence level 91%).

[0151] Fault handling and landing

[0152] Level 3 response: trigger emergency hovering + path replanning; twin verification: genetic algorithm optimizes the anti-sway parameters (compensation torque from ); Variable acceleration landing: Design the descent curve to meet Impact control: After landing, the vibration spectrum analysis showed that the main frequency was 12Hz (less than the equipment's natural frequency of 16Hz).

[0153] Structural health closed loop:

[0154] Stress field reconstruction: Kalman filter fusion 120 channels of FBG data to reconstruct the stress concentration area of ​​the lifting lug (maximum stress of the lifting cage ); Fatigue assessment: Rainflow counting results in a load spectrum of Δσ = 35-180 MPa, cumulative damage D = ∑(ni / Ni) = 0.41; safety margin solution is 1.37:

[0155] Technical effect: It successfully avoided all moving obstacles in level 9 winds and waves, reduced the energy consumption of swing suppression by 35% through digital twin optimization, achieved a landing positioning error of ±2.1cm, and achieved a structural fatigue life prediction accuracy of 92%.

[0156] The embodiments of the present invention are described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the above embodiments. Various changes can be made within the scope of knowledge of those skilled in the art without departing from the spirit of the present invention.

Claims

1. A method for controlling the lifting of a safe and portable lifting cage, characterized in that: The following steps are involved: S11: In the pre-hoisting stage, the driving source is controlled to start at low power, the initial tension of the cable is measured and calibrated by the tension sensor, and the displacement of the cable fixing point is simultaneously measured to detect deformation and looseness; S12: Pre-hoisting data collection and analysis, collecting environmental images, wind speed, visibility, sound data and the coordinates of the hoisting start and end points, and conducting multi-source data fusion analysis to analyze safety risks; S13: Hoisting control phase: Based on the three-dimensional obstacle model and real-time environmental parameters, the hoisting cage's motion trajectory and wind-induced sway are predicted. Path planning, obstacle avoidance, and active sway suppression are achieved through coordinated control of the main and auxiliary power sources. S14: Hoisting and landing control: After hoisting to the set position, the cable tension is monitored in real time and a smooth landing and docking is achieved through closed-loop tension control; The hoisting control phase also includes closed-loop control, which detects deviations between the actual route of the hoisting cage and the planned route and dynamically initiates stabilization measures, including: S51: Motion prediction, based on the preset route, real-time wind speed, current position of the hoisting cage, hoisting cage speed and hoisting cage swing angle, by solving the simple pendulum dynamics differential equations and using the Runge-Kutta method to predict the motion trajectory, attitude angle and swing amplitude and damping ratio of the hoisting cage in the next 30 seconds; S52: Coordinate transformation, the predicted cage world coordinates and Euler angles , through the camera extrinsic matrix Convert to the camera coordinate system, and then calculate its two-dimensional coordinates on the image plane through the perspective projection model ; S53: Virtual frame rendering: A translucent rectangular wireframe with the same physical dimensions as the cage is superimposed on the real-time video image based on the projection coordinates, and rotated according to the predicted posture. The frame color gradually changes from blue at the current moment to light green at the future moment, and automatically switches to a red warning frame in strong wind mode. S54: Deviation detection, comparing the actual image of the hoisting cage with the predicted virtual frame, obtaining the specific deviation, and determining whether the deviation exceeds the safety threshold; S55: Stabilization measures are triggered. When the deviation exceeds the threshold, the system automatically starts the preset stabilization measures.

2. A method for controlling the lifting of a safe and portable lifting cage according to claim 1, characterized in that: When performing automatic inspection of the drive source and sensor and calibrating the sensor, the method used is: S21: System startup and basic inspection: After the system is started, basic inspections including power input inspection and emergency brake test are carried out; S22: Low-power start-up and dynamic monitoring: The driving source of the hoisting cage is started at 10%-15% of the rated power and runs for 10 seconds. The cable tension, cable displacement data and the overall vibration amplitude of the equipment are checked. If they meet the standards, proceed to the next step; S23: Sensor calibration, specifically including: tension sensor calibration, establishing a linear regression equation, and correcting zero drift; displacement encoder calibration, comparing encoder pulse counts with laser ranging values, and compensating for mechanical transmission errors; S24: Cable condition assessment, including elastic deformation detection, plastic deformation diagnosis and slack detection.

3. The method for controlling the lifting of a safe and portable lifting cage according to claim 1, wherein: Perform pre-hoisting data collection and analysis through the following steps: S31: Visual-environmental data correlation analysis: Scan the lifting path with a high-precision binocular camera, establish an obstacle threat model, and assess obstacle risks; S32: Wind load stability assessment, measuring instantaneous wind speed and direction, calculating the equivalent wind disturbance force vector, and deriving the critical wind speed threshold. When the wind speed exceeds the critical wind speed threshold, an alarm is triggered; S33: Dynamic visibility detection, which measures air transmittance based on the principle of laser scattering and calculates effective visibility based on wind speed data; S34: Voiceprint fault diagnosis: deploys a four-element MEMS microphone array to collect broadband sound signals. Spectral entropy analysis and characteristic frequency band energy ratios are used to identify abnormal acoustic features. When the confidence level is greater than 80%, a shutdown inspection is performed. S35: Fusion calculation and safety decision-making, combining obstacle risk, wind load stability and effective visibility to generate a three-dimensional dynamic safety boundary.

4. The method for controlling the lifting of a safe and portable lifting cage according to claim 1, wherein: Path planning, obstacle avoidance, and active sway suppression are achieved through the following steps: S41: Environmental modeling: Using LiDAR and binocular cameras to scan the work area in real time, a high-precision 3D point cloud map is generated. Voxel filtering and plane segmentation algorithms are used to remove ground interference and construct a rasterized environmental model with semantic labels. Obstacles are replaced by simplified cuboid and ellipsoid models. S42: Dynamic safety zone calculation, based on the ratio of the current wind speed to the critical wind speed, dynamically adjust the boundary size of the safety zone; S43: Global path planning, using improved Algorithm,generates the initial path, which is then smoothed by B-spline curve; S44: Real-time obstacle avoidance control, using Kalman filter trajectory prediction for dynamic obstacles. When an obstacle is detected intruding into the 2m warning zone, a graded response is triggered: a detour is generated within 2-5m, and the vehicle is decelerated to 30% and emergency braking is initiated within 2m. S45: Pendulum coordinated control, real-time monitoring of the cage swing angle through the IMU. If it exceeds 1.5°, the main drive source is controlled to slow down by 20% and the auxiliary drive source is driven to apply reverse compensation torque until the swing angle stabilizes within 0.5°.

5. A method for controlling the lifting of a safe and portable lifting cage according to claim 4, characterized in that: Improved The algorithm adds a wind resistance penalty term to the traditional path cost function.

6. A method for controlling the hoisting of a safe and portable hoisting cage according to claim 5, characterized in that: Deviation detection is performed through the following steps: S61: Coordinate space alignment, the world coordinates of the predicted virtual frame are predicted through the camera intrinsic matrix and extrinsic matrix and the pixel coordinates of the actual cage image Convert to the camera coordinate system uniformly and optimize the homography matrix H to minimize the reprojection error; S62: Multi-level feature matching, the first layer uses YOLOv8 to quickly detect the actual cage bounding box, and then compares it with the predicted box. When , the second layer ORB feature matching is activated, and 50+ corner descriptors are extracted for brute force matching; if If it is insufficient, switch to laser point cloud registration to calculate the position deviation and posture deviation ; S63: Multivariate fusion verification, integrating vision, laser, and IMU data for cross-validation: weighted fusion results are applied when the difference between vision and laser position deviation is greater than 0.2m, and sensor fault flags are triggered when the difference between IMU and vision attitude angles is greater than 3°; exponential smoothing filtering is applied to suppress transient noise; S64: Dynamic threshold decision, set adaptive threshold based on real-time wind speed, when and Any exceeding of the threshold will result in a graded alarm, and the confidence level is output by the random forest model; S65: The result is structured and output, generating a standard JSON data packet containing timestamp, position deviation value, angle deviation value and threshold, status code and confidence level, and transmitted over the TSN network.

7. A method for controlling the hoisting of a safe and portable hoisting cage according to claim 6, characterized in that: The stabilization measures adopted are: S71: Gradual response: The current attitude deviation is determined based on the threshold value and is divided into mild deviation, moderate deviation and severe deviation. Mild deviation adopts the response measure of 50% speed reduction, moderate deviation adopts the response measure of emergency hovering and path replanning, and severe deviation adopts the response measure of emergency braking + mechanical locking; S72: Fault diagnosis: Analyze the deviation vector using the random forest model to obtain the predicted fault type and confidence level, and call the treatment measures from the database based on the predicted fault type; S73: Verification and execution of treatment measures: input the diagnosis results into a high-fidelity physics engine to simulate the effects of the treatment measures in real time in a virtual environment. If the system stabilizes within 10 seconds, execution is approved. Otherwise, parameters are optimized through genetic algorithms or reinforcement learning strategies are enabled to ensure that the intervention plan is physically executed only after its effectiveness is verified in the virtual environment.

8. The method for controlling the lifting of a safe and portable lifting cage according to claim 1, characterized in that: During the hoisting control phase, structural health monitoring is also included, specifically: S81: High-density sensor network deployment: Fiber Bragg grating sensor arrays are placed at key stress points of the cage. Together with piezoelectric accelerometers and temperature compensation modules, they collect multi-dimensional data on strain, vibration, temperature, and humidity in real time at a 1kHz frequency, building a holographic structural state perception network. S82: Dynamic reconstruction of the three-dimensional stress field. After noise reduction through Kalman filtering, the discrete strain data is input into the parameterized finite element model to invert the global stress field distribution of the structure and identify stress hotspots. S83: Quantitative assessment of fatigue damage, using the rainflow counting method to calculate the load cycle spectrum, calculating the fatigue damage index based on Miner's linear cumulative damage theory, and predicting the remaining life by combining the SN curve and stress ratio correction factor; S84: Intelligent early warning of crack initiation, integrating acoustic emission event energy, impact rate and spectrum centroid migration characteristics, through Bayesian probability model Assessing the risk of crack initiation: When Early warning is triggered when is the comprehensive crack probability, is the characteristic probability of acoustic emission when crack occurs, is the crack probability based on damage only, is the prior probability of acoustic emission events; S85: Multi-domain evaluation of safety margin, construction of stress-damage joint boundary equation, and real-time calculation of safety margin value; S86: Hierarchical decision-making and twin verification, performing intelligent operation and maintenance according to the damage index D: D<0.5 normal operation, 0.5≤D<0.7 load reduction by 25% and shortened maintenance, D≥0.7 forced shutdown; and before critical operations, rehearsal and verification are carried out in the digital twin model to intercept potential failure risks.

9. The method for controlling the hoisting of a safe and portable hoisting cage according to claim 1, characterized in that: During the hoisting landing control process, the specific process is as follows: S91: Real-time monitoring of the cable tension value F, when it is detected When the hoisting cage begins to touch the ground, the soft landing control program is started, in which: is the rated working tension; S92: Control the drive source to restart briefly at 30-50% of the rated power, so that the pulling force increases to 1.1 within 1.5S. , and if the tension fails to reach 1.05 within 0.25S , then return to step S91; S93: Constant deceleration Control the tension to decrease and meet the following requirements: ； in, is the preset ramp-down time, and ; S94: When the tension is reduced to 0.3 After 1.5 seconds, it is determined to be a complete landing, the driving source power is cut off, and the braking system is activated at the same time.

Images