Remote intelligent control system and control method for crane

Abstract

The invention discloses a remote intelligent control system and control method for a crane, and relates to the technical field of crane control, the system comprises the following components: a physical entity layer, a digital twinborn layer, a 5G transmission layer, a VR interaction layer and a virtual-real calibration layer; according to the method, an equipment state predictive modeling technology is adopted through a digital twinborn layer, historical operation data of a crane are collected, an equipment state database is constructed, key features related to equipment faults are extracted, and an equipment state prediction model is constructed based on a long-short-term memory network; according to the method, the change trend of equipment state parameters in the future 72 hours can be predicted, when the predicted state parameters approach or exceed a fault threshold value, the system can automatically mark corresponding parts in the digital twin, generate maintenance suggestions and synchronize the maintenance suggestions to a VR interaction layer to prompt an operator, and the maintenance suggestions are displayed in the VR interaction layer. According to the predictive maintenance mechanism, equipment faults can be early warned and processed in advance, the non-planned downtime is shortened, and the operation and maintenance cost is reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of crane control, in particular to a crane remote intelligent control system and control method. BACKGROUND

[0002] With the continuous improvement of industrial automation and intelligence level, as the key equipment of material handling, cranes play an important role in many fields such as ports, logistics, construction, etc.

[0003] Although the traditional crane remote control technology realizes the separation of operation and site to some extent, it still has many shortcomings. First, due to network delay and unstable data transmission, the accuracy and real-time performance of remote control are difficult to guarantee, which can easily cause operation errors and safety accidents. Second, the traditional system lacks real-time monitoring and predictive maintenance capability for equipment status, and can only be repaired after the equipment fails, which increases the unplanned downtime and operation and maintenance cost. In addition, the traditional remote control interface is mostly two-dimensional plane display, which cannot provide immersive operation experience, and the operator cannot fully perceive the working environment, which increases the operation difficulty and risk. Finally, the traditional system lacks flexible response strategies and intelligent decision support when facing complex and variable working environment, which is difficult to meet the demand of modern industry for high efficiency, safety and intelligence.

[0004] In view of the problems of traditional crane remote control technology such as insufficient accuracy, poor real-time performance, lack of predictive maintenance, poor operation experience and limited ability to cope with complex environment, the present application proposes a crane remote intelligent control system and control method, which is particularly important. SUMMARY

[0005] The purpose of the present application is to make up for the shortcomings of the prior art, and to provide a crane remote intelligent control system and control method, which can realize real-time monitoring and accurate control of the running state of the crane by integrating the physical entity layer, digital twin layer, 5G transmission layer, VR interaction layer and virtual-real calibration layer, provide immersive operation experience and intelligent decision support, and at the same time, through the equipment state predictive modeling technology, realize the early warning and preventive maintenance of equipment failure, effectively reduce the operation and maintenance cost, and improve the service life of the equipment.

[0006] To solve the above technical problems, the present application provides the following technical solutions: on the one hand, a crane remote intelligent control system, which comprises the following components: physical entity layer, digital twin layer, 5G transmission layer, VR interaction layer and virtual-real calibration layer. The physical entity layer is deployed on the crane body and includes multiple sensor modules and actuators. The sensor modules collect crane operation data in real time, including motor speed, hydraulic pressure, and mechanism position. The digital twin layer is communicatively connected to the physical entity layer. It constructs a digital twin of the crane based on the operating data to achieve a 1:1 virtual restoration of the working environment and equipment status. The digital twin also supports virtual pre-simulation of the working path and simulation of load handling collision risks. The 5G transmission layer: uses a 5G network to establish a communication link between the physical entity layer and the digital twin layer, and transmits the operating data and control commands in real time; The VR interaction layer is connected to the digital twin layer, providing an immersive control perspective for remote operators and supporting operators to input remote control commands based on virtual scenes. The virtual-real calibration layer is connected to the physical entity layer and the digital twin layer respectively, and is equipped with a virtual-real linkage calibration mechanism to detect the state deviation between the physical device and the digital twin in real time. When the deviation is ≤3%, a parameter correction command is automatically triggered to synchronize the digital twin parameters with the physical device state.

[0007] Furthermore, the sensor module of the physical entity layer adopts a multi-source data synchronous acquisition and preprocessing mechanism. The specific implementation steps are as follows: First, a unified clock synchronization signal triggers all sensors to sample simultaneously, avoiding data timestamp deviations; second, the collected analog data such as motor speed and hydraulic pressure are subjected to sliding window mean filtering to remove instantaneous pulse interference; then, a density-based outlier detection method is used to identify and remove abnormal data to ensure data validity; finally, a weighted fusion strategy is used to fuse the same physical quantity data collected by different types of sensors. The weight allocation is dynamically adjusted based on the measurement accuracy and real-time stability of each sensor. Through this preprocessing mechanism, the signal-to-noise ratio of the sensor output data is improved by more than 40%, providing reliable data support for the accurate modeling of the digital twin layer and effectively solving the problems of single sensor data being susceptible to interference and insufficient accuracy.

[0008] Furthermore, the actuator of the physical entity layer adopts adaptive response control technology. The specific implementation steps are as follows: First, it receives the control command issued by the virtual-real calibration layer and parses the corresponding action type, target parameters, and execution speed. Second, it collects the current action execution status in real time through the feedback sensor built into the actuator, including lifting height, luffing angle, and slewing angle. Then, it compares the command target parameters with the actual execution status and calculates the deviation value. Then, according to the magnitude and trend of the deviation value, it dynamically adjusts the drive parameters of the actuator, including motor drive current and hydraulic valve opening. Finally, it establishes an actuator response delay compensation mechanism, predicts the action response delay time based on historical execution data, and compensates for the delay amount in advance when issuing control commands, ensuring that the actuator can respond to control commands accurately and in a timely manner, so that the action execution error is controlled within ±3mm, solving the problems of slow response speed and low action accuracy of traditional actuators.

[0009] Furthermore, the digital twin layer employs dynamic scene adaptive modeling technology. The specific implementation steps are as follows: First, an initial geometric model is constructed based on the crane's 3D design drawings, clarifying the structural dimensions and connection relationships of each component. Second, the physical attribute parameters of the digital twin, including motor output torque and hydraulic system damping coefficient, are dynamically updated using real-time received physical entity layer operation data. Then, obstacle and work surface scene elements are extracted using image segmentation algorithms from visual sensor-collected work environment image data and added to the virtual scene in real time. Finally, the model detail level is adaptively adjusted according to the complexity of the work scene, optimizing rendering efficiency while ensuring modeling accuracy. When the obstacle density in the work area exceeds a preset threshold, the model details of non-critical areas are automatically simplified, prioritizing the modeling accuracy of the core work area. This achieves real-time updates and efficient rendering of the virtual scene under complex working conditions, solving the deficiency of static modeling in adapting to dynamic environmental changes.

[0010] Furthermore, the collision risk simulation of the digital twin layer employs a collision risk quantification assessment algorithm, the formula of which is: ,in for The collision risk value at any given moment ranges from 0 to 10, with higher values ​​indicating a higher risk. The real-time distance between the load and the obstacle. The relative velocity between the load and the obstacle. This is the obstacle priority coefficient. , , The weighting coefficients are derived from training data from over 1000 actual crane operation cases and optimized using a support vector machine algorithm. This algorithm creatively combines distance, relative speed, and obstacle priority to achieve a quantitative assessment of collision risk. It automatically triggers warnings, which, compared to traditional distance threshold judgment methods, can more comprehensively and accurately identify potential collision risks, providing operators with a more scientific basis for avoidance decisions.

[0011] Furthermore, the digital twin layer employs equipment status predictive modeling technology. The specific implementation steps are as follows: First, historical crane operating data is collected, including sensor monitoring data, fault records, and maintenance logs, to construct an equipment status database. Second, feature extraction is performed on the historical data, extracting key features related to equipment faults such as motor current fluctuations, hydraulic pressure change rates, and component vibration frequencies. Then, an equipment status prediction model is constructed based on a long short-term memory network, inputting recent operating data to predict the trend of equipment status parameter changes over the next 72 hours. Finally, the prediction results are compared with preset fault thresholds. When the predicted status parameters approach or exceed the fault thresholds, the corresponding components are automatically marked in the digital twin, and maintenance suggestions are generated. Simultaneously, these suggestions are synchronized to the VR interaction layer to prompt the operator, enabling early warning and preventative maintenance of equipment faults, reducing equipment downtime, and lowering operation and maintenance costs.

[0012] Furthermore, the VR interaction layer employs multimodal fusion control command generation technology. The specific implementation steps are as follows: First, the VR headset collects the operator's visual focus data to determine the current work area and equipment component being focused on. Second, the operator's hand movement data, including grip strength, movement trajectory, and operation speed, is collected through the posture and pressure sensors of the control handle. Then, the operator's voice commands are received through the voice acquisition module and converted into digital commands via a voice recognition algorithm. Finally, a multimodal data fusion decision model is constructed to perform consistency verification and weight allocation on the visual, action, and voice command data. When conflicts exist between different modal commands, the action command is prioritized, the voice command is secondary, and the visual focus data is used as a reference to make the final control command. This improves the convenience of remote control and the accuracy of command recognition, solving the problems of single-modal control being easily interfered with and lacking operational flexibility.

[0013] Furthermore, the VR interaction layer employs immersive scene optimization and risk warning technology. The specific implementation steps are as follows: First, based on the operator's visual parameters and operating habits, the resolution, brightness, and contrast of the VR scene are adaptively adjusted to ensure visual comfort during prolonged use. Second, based on the work path planning results of the digital twin, the optimal work path is marked with dynamically highlighted lines in the VR scene to guide the operator. Third, based on the collision risk assessment results, different warning methods are used according to risk levels: a yellow warning box for low risk, a flashing orange warning box with slight vibration for medium risk, and a combination of a flashing red warning box, voice alarm, and strong vibration from the control handle for high risk. Finally, operators can customize warning parameters, including warning thresholds, vibration intensity, and voice volume, to meet the needs of different operators. Through multi-dimensional scene optimization and risk warning, the safety and convenience of remote control are improved, and the operator's workload is reduced.

[0014] Furthermore, the virtual-real linkage calibration mechanism of the virtual-real calibration layer adopts a dynamic deviation correction algorithm, the formula of which is: ,in For the corrected state parameters of the digital twin, The state parameters of the digital twin before correction. , These are the state parameters collected in real time for the physical entity layer. This is the proportional correction factor. The time decay coefficient, This is the integral correction factor. This is the cumulative time of the current calibration cycle. For environmental disturbance weights, parameters , , Derived from machine learning training on historical crane operating data, and optimized using a gradient descent algorithm, this ensures calibration accuracy under different load weights and operating speeds; environmental disturbance weights are also included. Based on temperature, humidity, and wind speed data collected by visual and environmental sensors, environmental factors are quantified into weight values ​​between 0 and 1 using a fuzzy logic algorithm. The greater the environmental interference, the lower the weight value. The closer the integral term is to 1, the stronger its effect on correcting the deviation. This algorithm creatively integrates proportional correction, time decay and environmental adaptive adjustment mechanisms. Compared with the traditional Kalman filter algorithm, the deviation convergence speed is improved by more than 30%, and the virtual-real deviation can be quickly controlled within 1%, ensuring real-time synchronization between the virtual model and the physical device.

[0015] On the other hand, a remote intelligent control method for a crane is characterized by the following specific steps: S1. Data Acquisition and Transmission: Each sensor module in the physical entity layer acquires crane operation data and working environment image data in real time. The data is uploaded to the digital twin layer through the 5G transmission layer. The UDP protocol is used to ensure real-time performance during transmission, and CRC check is used to prevent data loss or tampering. S2. Digital Twin Construction and Scene Recreation: After receiving data, the digital twin layer uses a geometric modeling module to construct an equipment model based on 3D drawings and adjusts the position of the virtual equipment's mechanism by combining displacement sensor data; a physical modeling module corrects the mechanical properties of the virtual model based on motor speed and hydraulic pressure data; a behavioral modeling module simulates the equipment's motion state based on kinematic equations; and simultaneously, it constructs an obstacle model in the virtual scene by combining image data collected by a visual sensor, thus completing a 1:1 recreation of the working environment. S3, VR Immersive Pre-show and Risk Assessment: Remote operators wear VR headsets and access the virtual scene of the digital twin layer. They adjust the movement of the virtual crane through the control handle to perform a pre-show of the work path. The collision detection module calculates the distance between the load and obstacles in real time. When the collision probability is ≥5%, the risk area is marked with a red warning box in the VR scene, and avoidance suggestions are output. S4. Control command input and deviation detection: After confirming the operation path based on the pre-rehearsal results, the operator inputs remote control commands through the control handle or voice. The commands are transmitted to the virtual-real calibration layer via the 5G transmission layer. The deviation detection module compares the actual state of the physical equipment with the simulated state of the digital twin in real time and calculates the deviation value. S5. Command Calibration and Execution: If the deviation value is ≤3%, the virtual-physical calibration layer directly sends the control command to the actuator of the physical entity layer to drive the crane to complete the lifting, luffing, and slewing actions; if the deviation value is >3%, the parameter correction module starts the calibration process: first, the data noise is filtered by the Kalman filter algorithm, and then the amplitude and phase of the digital twin parameters and control command are adjusted by the PID control algorithm until the deviation is ≤3%. Then, the calibrated command is sent to the actuator to ensure the operation accuracy.

[0016] Compared with existing technologies, this remote intelligent control system and method for cranes has the following advantages: I. This system employs predictive modeling technology for equipment status through a digital twin layer. By collecting historical operating data of the crane, it constructs an equipment status database and extracts key features related to equipment failures. Based on a long short-term memory network, it builds an equipment status prediction model that can predict the trend of equipment status parameters over the next 72 hours. When the predicted status parameters approach or exceed the failure threshold, the system automatically marks the corresponding component in the digital twin and generates maintenance suggestions, which are simultaneously synchronized to the VR interaction layer to prompt the operator. This predictive maintenance mechanism enables early warning and handling of equipment failures, reduces unplanned downtime, lowers maintenance costs, and extends the service life of the equipment.

[0017] Second, this system collects crane operation data in real time through the physical entity layer and uploads this data to the digital twin layer in real time using the 5G transmission layer. This creates a digital twin that is a 1:1 virtual replica of the actual working environment and equipment status. The digital twin not only supports virtual pre-simulation of the work path but also simulates the collision risk during load handling. Combined with the immersive control perspective provided by the VR interaction layer, operators can conduct pre-simulation and risk assessment in the virtual environment, thereby avoiding potential risks in advance. The virtual-real calibration layer detects the deviation between the physical equipment and the digital twin in real time and automatically triggers parameter correction commands when the deviation exceeds a threshold, ensuring the accuracy and safety of remote control. This series of technological innovations significantly improves the operating efficiency of cranes and personnel safety in complex working environments.

[0018] Other advantages, objectives and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination or study, or may be learned from the practice of the invention. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0020] Figure 1 A flowchart of a remote intelligent control system for a crane; Figure 2 A flowchart illustrating the virtual-real linkage calibration process of a remote intelligent control system for cranes; Figure 3 This is a flowchart of a remote intelligent control method for a crane. Detailed Implementation

[0021] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0022] Example 1 The physical entity layer deployed on the quay crane body starts operation. The sensor module operates according to a multi-source data synchronous acquisition and preprocessing mechanism. A unified clock synchronization signal triggers simultaneous sampling by motor speed sensors, hydraulic pressure sensors, and mechanism position sensors to avoid data timestamp deviations. The collected analog data such as motor speed and hydraulic pressure undergo sliding window mean filtering. A density-based outlier detection method is used to remove abnormal data caused by sea wind interference. A weighted fusion strategy is then used to fuse multi-sensor data for the same physical quantity. Simultaneously, visual sensors acquire real-time image data of the ship's cargo hold internal structure, container distribution in the yard, and the operating status of surrounding quay cranes and trucks. The above operational data and image data are transmitted in real-time via the 5G transmission layer using the UDP protocol to ensure real-time transmission, with CRC checks preventing data loss or tampering, and are continuously uploaded to the digital twin layer. Figure 3 As shown.

[0023] After receiving data, the digital twin layer constructs a digital twin of the quay crane using dynamic scene adaptive modeling technology. First, it builds an initial geometric model based on the crane's 3D design drawings, clarifying the structural dimensions and connection relationships of components such as the boom, hook, and trolley. Then, based on data uploaded from the physical entity layer, such as motor speed, hydraulic pressure, and mechanism position, it dynamically updates the physical attribute parameters of the digital twin. Image segmentation algorithms are used to process the working environment images, extracting obstacles and working surface elements such as ship bulkheads, stacked containers, and quay crane columns, which are added to the virtual scene in real time. Due to the dense stacking of containers in the yard, the obstacle density exceeds a preset threshold. The digital twin automatically simplifies the model details of non-critical areas at the yard edge, optimizing rendering efficiency while ensuring modeling accuracy around the working path. Ultimately, a 1:1 virtual restoration of the working environment and equipment status is achieved. Simultaneously, the digital twin layer initiates predictive modeling technology for equipment status, calling upon historical operating databases to extract key features such as motor current fluctuations, hydraulic pressure change rates, and component vibration frequencies. A long short-term memory network model predicts the equipment status for the next 72 hours. Comparison with preset fault thresholds shows no abnormalities, and no maintenance prompts are generated.

[0024] Remote operators at the terminal control center wear VR headsets and hold control handles to access the VR interaction layer. The VR interaction layer adaptively adjusts the scene resolution, brightness, and contrast based on the operator's vision and operating habits. The operator uses the VR headset to lock onto the target container in the cargo hold, adjusts the virtual crane's hook raising and lowering, and boom luffing movements using the control handles, and issues a voice command to raise the hook to a height of 3 meters. The decision model of the multimodal fusion control command generation technology performs consistency checks on the three types of data. If the commands are conflict-free, a pre-simulation command is generated. During the pre-simulation, the digital twin layer uses a collision risk quantification assessment algorithm to detect the collision risk between the container and the ship's bulkhead and surrounding containers in real time. The formula is: ,in for Collision risk value at any given moment The real-time distance between the load and the obstacle. The relative velocity between the load and the obstacle. This is the obstacle priority coefficient. , , As a weighting factor, when the pre-simulation reaches the point where the container is about to approach the edge of the cargo hold, if the probability of collision reaches 6% (≥5%), the VR scene immediately marks the risk area with a red warning box, outputs a suggestion to adjust the boom angle by 2° to avoid collision, and simultaneously activates the high-risk warning mode, with a flashing red prompt box, voice alarm, and strong vibration of the control handle. The operator corrects the action according to the prompts and re-simulates until the collision risk is eliminated.

[0025] After confirming the optimal operating path, the operator issues commands via the control handle to lift the container to a height of 6 meters, rotate the crane boom towards the yard, and uniformly adjust the luffing speed to the designated container lowering position. Voice commands are provided to maintain stable operation. Once the multimodal fusion decision model confirms consistency with the commands, the final control commands are sent via the 5G transmission layer to the virtual-real calibration layer. The deviation detection module in the virtual-real calibration layer collects real-time actual state parameters such as motor speed, hydraulic pressure, and mechanism position from the physical entity layer. These parameters are compared with the simulated state parameters of the digital twin, and a deviation value of 2.3% is calculated. Figure 2 As shown.

[0026] Because the deviation value is 2.3% ≤ 3%, the virtual-real calibration layer directly sends control commands to the actuators in the physical entity layer. The actuators employ adaptive response control technology, analyzing the action type, target parameters, and execution speed corresponding to the commands. Through built-in feedback sensors, the execution status is collected in real time, and the deviation value is calculated by comparing the target parameters with the actual status, dynamically adjusting the drive parameters. Simultaneously, a response delay compensation mechanism is activated, predicting and compensating for delays based on historical data to ensure precise action execution. During the operation, the virtual-real calibration layer continuously monitors the deviation, always maintaining a deviation value ≤ 3%. The digital twin synchronizes the physical equipment status in real time, and the VR interaction layer marks the actual operation path with dynamically highlighted lines until the container smoothly falls to the designated position in the yard, completing the loading and unloading operation. Figure 1 As shown.

[0027] Example 2 The physical entity layer deployed on the metallurgical crane body starts working. The sensor module operates according to a multi-source data synchronous acquisition and preprocessing mechanism. A unified clock synchronization signal triggers synchronous sampling of the motor speed sensor, hydraulic pressure sensor, and mechanism position sensor to avoid data timestamp deviations. The collected analog data such as motor speed and hydraulic pressure undergo sliding window mean filtering. Density-based isolated point detection removes abnormal data caused by sensor drift due to the high-temperature environment. A weighted fusion strategy then fuses multi-sensor data for the same physical quantity. The vision sensor, through a high-temperature resistant lens, acquires real-time images of the operating environment, capturing scene information such as the high-temperature area at the converter tapping spout, the outline of the molten steel ladle, plant columns, crane tracks, and high-temperature flues. All operational data and image data are transmitted through the 5G transmission layer using the UDP protocol to ensure real-time transmission. CRC checks are used to resist data anomalies caused by high-temperature electromagnetic interference, and the data is continuously uploaded to the digital twin layer. Figure 3 As shown.

[0028] After receiving the data, the digital twin layer constructs a digital twin of the metallurgical crane using dynamic scene adaptive modeling technology. First, an initial geometric model is constructed based on the crane's 3D design drawings, clarifying the structural dimensions and connection relationships of components such as the hoisting mechanism, slewing mechanism, and frame. Then, based on data such as motor speed, hydraulic pressure, and mechanism position uploaded from the physical entity layer, the physical attribute parameters of the digital twin are dynamically updated. Image segmentation algorithms are used to process the working environment images, extracting obstacles and working surface elements such as the converter body, plant columns, crane tracks, and high-temperature flues, integrating them into the virtual scene in real time. Due to the high density of obstacles in the workshop, the digital twin automatically simplifies the model details of non-critical support structures on the plant roof, improving rendering efficiency while ensuring the accuracy of the ladle hoisting path modeling, achieving a 1:1 virtual restoration of the working environment and equipment status. Simultaneously, the digital twin layer initiates predictive modeling technology for equipment status, extracting key features such as hydraulic pressure change rate, motor current fluctuations, and hoisting mechanism vibration frequency based on historical operating data. A long short-term memory network model predicts the equipment status for the next 72 hours. If the hydraulic system pressure change rate is detected to be close to a preset fault threshold, the hydraulic pump component is immediately marked in the digital twin, generating maintenance suggestions for checking the hydraulic pump seals after the operation, which are then synchronized to the VR interaction layer to prompt the operator.

[0029] Remote operators in the workshop control room wear anti-glare VR headsets and anti-slip control handles to access the VR interaction layer. The VR interaction layer adaptively adjusts scene resolution, brightness, and contrast based on the operator's vision and long-term work habits to avoid interference from strong light in high-temperature environments. The operator uses the VR headset to lock onto the converter tapping area, uses the control handles to adjust the lifting height and slewing angle of the virtual crane, and issues voice commands to approach the converter tapping area. A decision model based on multimodal fusion control command generation technology verifies the consistency of the three types of data; if the commands are conflict-free, a pre-simulation command is generated. During the pre-simulation, the digital twin layer uses a collision risk quantification assessment algorithm to monitor the collision risk between the molten steel ladle and the converter body, high-temperature flue, and crane rails in real time. When the molten steel ladle approaches the high-temperature flue, the collision probability reaches 7% (≥5%), and the VR scene marks the risk area with a red warning box, activating a high-risk warning mode with a flashing red warning box, a voice alarm "Please avoid the high-temperature flue," and strong vibration of the control handles. After the operator adjusts the slewing angle, the collision risk decreases to low risk, and the warning mode switches to a yellow warning box until the pre-simulation determines a safe hoisting path.

[0030] After confirming the optimal path, the operator precisely controls the ladle's lifting height, slewing speed, and luffing angle using a control handle, supplemented by voice commands to keep the ladle level. A multimodal fusion decision model weights the two types of data, and upon confirmation of consistency, generates the final control command, which is then transmitted via the 5G transmission layer to the virtual-real calibration layer. The deviation detection module in the virtual-real calibration layer collects real-time actual state parameters from the physical entity layer, such as motor speed, hydraulic pressure, and mechanism position, and compares them with the simulated state parameters of the digital twin. The calculated deviation value is 3.8%. Figure 2 As shown.

[0031] Command Calibration and Execution: Due to a deviation of 3.8% > 3%, the parameter correction module of the virtual-real calibration layer initiates the calibration process. First, it filters high-temperature electromagnetic interference noise from the data using a Kalman filter algorithm. Then, it adjusts the digital twin parameters and the amplitude and phase of the control commands using a PID control algorithm. After continuous optimization, the deviation is reduced to 2.5% (≤ 3%). The calibrated commands are then sent to the execution mechanism in the physical entity layer. The execution mechanism employs adaptive response control technology, analyzing the action type, target parameters, and execution speed corresponding to the command. It collects the action execution status in real time through built-in feedback sensors, compares the target parameters with the actual status to calculate the deviation, and dynamically adjusts the drive parameters. Simultaneously, a response delay compensation mechanism is activated, predicting the delay time based on historical data and compensating in advance to ensure the smooth hoisting of the molten steel ladle. During the operation, the virtual-real calibration layer continuously monitors the deviation through a dynamic deviation correction algorithm, always maintaining a deviation value ≤ 3%. The VR interaction layer marks the actual operation path with dynamically highlighted lines, allowing the operator to monitor the entire process until the molten steel ladle is smoothly hoisted to the continuous casting machine's casting position, completing the transfer operation. Figure 1 As shown.

[0032] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A remote intelligent control system for a crane, characterized in that The system comprises the following components: a physical entity layer, a digital twin layer, a 5G transmission layer, a VR interaction layer, and a virtual-real calibration layer: The physical entity layer is deployed on the crane body and comprises a plurality of sensor modules and actuators. The sensor modules collect real-time crane operation data, including motor speed, hydraulic pressure, and mechanism position. The digital twin layer is communicatively connected to the physical entity layer and constructs a crane digital twin based on the operation data, achieving 1:1 virtual restoration of the working environment and device state. The digital twin supports virtual pre-rehearsal of the operation path and simulation of load handling collision risks. The 5G transmission layer uses a 5G network to establish a communication link between the physical entity layer and the digital twin layer, and transmits the operation data and control instructions in real time. The VR interaction layer is connected to the digital twin layer and provides an immersive control perspective for remote operators, supporting remote control instruction input by operators based on virtual scenes. The virtual-real calibration layer is connected to the physical entity layer and the digital twin layer and is configured with a virtual-real linkage calibration mechanism. It detects the state deviation between the physical device and the digital twin in real time and automatically triggers parameter correction instructions when the deviation is ≤3%, synchronizing the digital twin parameters and the physical device state.

2. A remote intelligent control system for a crane as claimed in claim 1, characterized in that The sensor modules of the physical entity layer use a multi-source data synchronous collection and preprocessing mechanism. The specific implementation steps are as follows: first, trigger all sensors to sample simultaneously through a unified clock synchronization signal to avoid data timestamp deviation; second, perform sliding window mean filtering on collected analog data such as motor speed and hydraulic pressure; third, identify and remove abnormal data using a density-based outlier detection method; and finally, use a weighted fusion strategy to fuse the same physical quantity data collected by different types of sensors.

3. A remote intelligent control system for a crane as claimed in claim 1, wherein, The actuators of the physical entity layer use adaptive response control technology. The specific implementation steps are as follows: first, receive the control instructions issued by the virtual-real calibration layer, analyze the action type, target parameters, and execution speed corresponding to the instructions; second, collect the current action execution state in real time through the feedback sensors built-in the actuators; third, compare the instruction target parameters with the actual execution state and calculate the deviation value; fourth, dynamically adjust the drive parameters of the actuators according to the size and change trend of the deviation value; and finally, establish an execution response delay compensation mechanism, predict the action response delay time based on historical execution data, and compensate for the delay in advance when issuing control instructions.

4. The remote intelligent control system for a crane according to claim 1, wherein The digital twin layer adopts a dynamic scene self-adaptive modeling technology, and the specific implementation steps are as follows: firstly, an initial geometric model is constructed based on a three-dimensional design drawing of the crane, and the structure size and connection relationship of each component are determined; secondly, the physical attribute parameters of the digital twin are dynamically updated through the real-time received physical entity layer operation data; thirdly, the image data of the working environment collected by the visual sensor are used to extract obstacles and working surface scene elements by using an image segmentation algorithm, and the obstacles and working surface scene elements are added to the virtual scene in real time; and finally, the model detail level is adaptively adjusted according to the complexity of the working scene, the rendering efficiency is optimized under the premise of ensuring the modeling accuracy, and when the obstacle density of the working area exceeds a preset threshold, the model details of non-critical areas are automatically simplified.

5. A remote intelligent control system for a crane as claimed in claim 1, wherein, The collision risk simulation of the digital twin layer adopts a collision risk quantitative evaluation algorithm, and the formula is: Wherein is the collision risk value at the moment, is the real-time distance between the load and the obstacle, is the relative speed between the load and the obstacle, is the obstacle priority coefficient, , , , is the weight coefficient.

6. A remote intelligent control system for a crane as claimed in claim 1, wherein, The digital twin layer adopts a device state predictive modeling technology, and the specific implementation steps are as follows: firstly, historical operation data of the crane are collected to construct a device state database; secondly, feature extraction is performed on the historical data to extract key features related to device failure, such as motor current fluctuation, hydraulic pressure change rate, and component vibration frequency; thirdly, a device state prediction model is constructed based on a long short-term memory network, and recent operation data are input to predict the future 72-hour device state parameter change trend; and finally, the prediction result is compared with a preset fault threshold, when the predicted state parameter approaches or exceeds the fault threshold, the corresponding component in the digital twin is automatically marked, and a maintenance suggestion is generated, which is simultaneously synchronized to the VR interaction layer to prompt the operator.

7. A remote intelligent control system for a crane as claimed in claim 1, wherein, The VR interaction layer adopts a multi-modal fusion control instruction generation technology, and the specific implementation steps are as follows: firstly, the visual focus data of the operator are collected through the VR head-mounted device to determine the working area and device component currently focused by the operator; secondly, the hand action data of the operator are collected through the posture sensor and pressure sensor of the control handle; thirdly, the voice instructions of the operator are received through the voice collection module and converted into digital instructions through a voice recognition algorithm; and finally, a multi-modal data fusion decision model is constructed to perform consistency checking and weight distribution on the instruction data of the three modalities of vision, action and voice, and when there is a conflict between the different modal instructions, the action instruction is given priority, the voice instruction is given secondary priority, and the visual focus data is used as a reference to make a decision, and the final control instruction is generated.

8. A remote intelligent control system for a crane as claimed in claim 1, wherein, The VR interaction layer adopts an immersive scene optimization and risk warning technology, and the specific implementation steps are as follows: firstly, the resolution, brightness and contrast of the VR scene are adaptively adjusted according to the visual parameters and operation habits of the operator; secondly, the optimal working path is marked in the VR scene with a dynamic highlight line based on the working path planning result of the digital twin, to guide the operator; thirdly, different warning methods are divided according to the risk level based on the collision risk assessment result, a yellow prompt box is used for low risk, an orange flashing prompt box is used for medium risk with slight vibration, and a red flashing prompt box, voice alarm and strong vibration of the control handle are combined for high risk; and finally, the operator can customize warning parameters, including warning threshold, vibration intensity and voice volume, to meet the needs of different operators.

9. A remote intelligent control system for a crane as claimed in claim 1, wherein, The virtual-real linkage calibration mechanism of the virtual-real calibration layer adopts a deviation dynamic correction algorithm, and the formula is: wherein is a corrected digital twin state parameter, is a state parameter before correction, , is a state parameter collected in real time by the physical entity layer, is a proportional correction coefficient, is a time attenuation coefficient, is an integral correction coefficient, is the cumulative time of the current calibration period, is an environmental interference weight.

10. A remote intelligent control method for a crane, suitable for a remote intelligent control system for a crane according to any one of claims 1-9, characterized in that, The specific steps of the method are as follows: S1, data acquisition and transmission: the sensor modules of the physical entity layer collect crane operation data and job environment image data in real time, upload the data to the digital twin layer through the 5G transmission layer, use UDP protocol to ensure real-time during transmission, and use CRC check to avoid data loss or tampering; S2, digital twin construction and scene restoration: after receiving the data, the digital twin layer uses the geometric modeling module to build the device model based on the three-dimensional drawing, adjusts the mechanism position of the virtual device based on the displacement sensor data, uses the physical modeling module to correct the mechanical properties of the virtual model based on the motor speed and hydraulic pressure data, uses the behavior modeling module to simulate the device motion state based on the kinematics equation, and at the same time, constructs the obstacle model in the virtual scene based on the image data collected by the visual sensor, and completes the 1:1 job environment restoration; S3, VR immersive rehearsal and risk assessment: the remote operator wears a VR headset device and accesses the virtual scene of the digital twin layer, adjusts the action of the virtual crane through the control handle, rehearses the operation path, and the collision detection module calculates the distance between the load and the obstacle in real time. When the collision probability is greater than or equal to 5%, the risk area is marked with a red warning box in the VR scene, and the avoidance suggestion is output; S4, control instruction input and deviation detection: after the operator confirms the operation path according to the rehearsal result, the control instruction is input through the control handle or voice, the instruction is transmitted to the virtual-real calibration layer through the 5G transmission layer, and the deviation detection module compares the actual state of the physical device with the simulation state of the digital twin in real time to calculate the deviation value; S5, instruction calibration and execution: if the deviation value is less than or equal to 3%, the virtual-real calibration layer directly issues the control instruction to the actuator of the physical entity layer to drive the crane to complete the lifting, amplitude, and rotation actions; if the deviation value is greater than 3%, the parameter correction module starts the calibration process: first, filter the data noise through Kalman filtering algorithm, then adjust the amplitude and phase of the digital twin parameters and control instruction through PID control algorithm, until the deviation is less than or equal to 3%, then issue the calibrated instruction to the actuator.

Images