A monitoring and early warning method and system for a container gantry crane

Abstract

The present invention discloses a monitoring and early warning method and system for a container gantry crane, which relates to the technical field of crane monitoring and early warning. The monitoring and early warning method for the container gantry crane continuously obtains the monitoring time-series data of the gantry crane, and respectively performs risk extraction processing on the monitoring time-series data of the gantry crane to obtain a risk assessment index set for each time period of the gantry crane, including a structural risk aggregation index, a lifting behavior risk index, and an operating environment interference aggregation index; comprehensively analyzes the risk assessment index set for each time period of the gantry crane to obtain a risk coupling determination index for each time period of the gantry crane. The present invention takes preset early warning measures for the gantry crane based on the risk coupling determination index for each time period, thereby significantly improving the prediction and early warning capabilities for potential safety hazards, further effectively guaranteeing the safety during the crane operation process, and then avoiding misjudgment and missed judgment.

Description

Technical Field

[0001] The present invention relates to the technical field of crane monitoring and early warning, and in particular to a monitoring and early warning method and system for a container gantry crane. Background Art

[0002] As one of the important equipment in the modern port and logistics industries, container gantry cranes are widely used in the loading, unloading, transportation, and stacking operations of containers. Such cranes have strong load-bearing capacity and high operating accuracy, and are usually used in large-scale container storage and transportation sites such as docks and warehouses. With the development of the container logistics industry, the application of gantry cranes has been increasing continuously. Their working efficiency and safety are crucial for modern logistics systems. However, under high-intensity working environments and during long-term operation, there are some potential safety hazards in gantry cranes, which can affect the working efficiency of the cranes, and may also lead to equipment damage, personnel casualties, or serious economic losses. Moreover, traditional gantry crane monitoring systems mainly rely on manual inspections and regular equipment detections, and their monitoring means are relatively single, making it difficult to comprehensively and real-time grasp the operating status of the equipment, and thus difficult to accurately reflect the overall operating risks of the equipment.

[0003] The prior art, such as the container gantry crane yard operation status real-time monitoring system and monitoring method disclosed in the patent application with the publication number of CN114132842B, includes a single-machine control system and a remote operation console in the central control room. The single-machine control system is communicatively connected to the remote operation console and includes a single-machine PLC, several network cameras, and a distributed processing platform. The single-machine PLC controls the operation of the crane and sends the current operation status to the distributed processing platform. The several network cameras are arranged at different positions of the crane and can capture the operation dynamics or working areas of the crane from different perspectives in real time. The distributed processing platform includes a video stream access module, a core algorithm module, a PLC interaction module, an alarm information output module, and a WEB server. The present invention makes full use of existing remote control cameras, distributes and pre-sets the calculation and processing units, integrates the operation status of the crane and multi-perspective cameras, and automatically switches functions according to the operation status, with a high degree of intelligence, effectively improving the operation efficiency and safety protection level.

[0004] Based on the above solutions, it is found that the limitations of the existing technology at least include the following problems. It is difficult for the existing technology to extract and analyze multi-dimensional data involved in the operation of gantry cranes, such as structural operation data, lifting behavior data, and environmental interference data. There is a lack of comprehensive analysis and efficient processing of different types of data, resulting in insufficient accuracy of risk judgment. For example, the stacking of containers under the spreader is not standardized, and at the same time, the load is eccentric and shaking during the lifting process, which is likely to cause the spreader to become unstable or the structural stress to be too large. However, the existing technology often analyzes these factors independently, ignoring their mutual influence, and it is difficult to comprehensively judge the combined impact of structural risk and operating environment risk on safety, and thus it is difficult to accurately judge the potential safety hazards existing in the crane during operation. Summary of the Invention

[0005] In view of the deficiencies of the existing technology, the present invention provides a monitoring and early warning method and system for container gantry cranes, which solves the problem that it is difficult for the existing technology to conduct comprehensive multi-dimensional analysis and ignores their interaction, resulting in inaccurate risk judgment.

[0006] To achieve the above objectives, the present invention is realized through the following technical solutions: A monitoring and early warning method for a container gantry crane includes the following steps: continuously obtaining the monitoring time-series data of the gantry crane, where the monitoring time-series data includes structural operation time-series data, lifting behavior time-series data, and operation video stream time-series data; respectively performing risk extraction processing on the monitoring time-series data of the gantry crane to obtain a risk assessment index set for each period of the gantry crane, including a structural risk aggregation index, a lifting behavior risk index, and an operating environment interference aggregation index; comprehensively analyzing the risk assessment index set for each period of the gantry crane to obtain a risk coupling determination index for each period of the gantry crane; and taking a preset early warning measure for the gantry crane based on the risk coupling determination index for each period.

[0007] Further, the specific formula for calculating the risk coupling determination index for a certain period of the gantry crane is as follows: ; where FpD is the risk coupling determination index for a certain period of the gantry crane, GxH is the structural risk aggregation index for a certain period of the gantry crane, α1 is the structural risk adjustment coefficient stored in the database, DxF is the lifting behavior risk index for a certain period of the gantry crane, ZgR is the operating environment interference aggregation index for a certain period of the gantry crane, α2 is the behavior environment aggregation adjustment coefficient stored in the database, and α3 is the structural behavior aggregation adjustment coefficient stored in the database.

[0008] Further, the specific steps for obtaining the structural risk aggregation index for each time period of the gantry crane by including the main girder impact response index, outrigger force distribution value, main girder strain variation index, and vibration coordination index in the structural operation time series data are as follows: comprehensively analyze the structural operation time series data of the gantry crane respectively to obtain the structural operation risk assessment index set for each time period of the gantry crane, including the structural abnormal response intensity index and the structural coordination stability index; and comprehensively analyze the structural operation risk assessment index set for each time period of the gantry crane to obtain the structural risk aggregation index for each time period of the gantry crane.

[0009] Further, the specific steps for obtaining the structural operation risk assessment index set for each time period of the gantry crane are as follows: read the main girder impact response index and the main girder strain variation index for each time period of the gantry crane, and conduct a comprehensive analysis to obtain the structural abnormal response intensity index for each time period of the gantry crane; read the outrigger force distribution value and the vibration coordination index for each time period of the gantry crane, and conduct a comprehensive analysis to obtain the structural coordination stability index for each time period of the gantry crane.

[0010] Further, the specific formula for calculating the structural risk aggregation index for a certain time period of the gantry crane is as follows: ; where GxH is the structural risk aggregation index for a certain time period of the gantry crane, YcX is the structural abnormal response intensity index for a certain time period of the gantry crane, β1 is the abnormal response adjustment coefficient stored in the database, XtW is the structural coordination stability index for a certain time period of the gantry crane, β2 is the coordination stability adjustment coefficient stored in the database, β3 is the interaction coupling adjustment coefficient stored in the database, and β4 is the coupling sensitivity adjustment coefficient stored in the database.

[0011] Further, the hoisting behavior time series data includes the jerk frequency value of the spreader, the load eccentricity value, the load sway coordination index, the inertial energy accumulation index of the spreader, the operation smoothness index, and the braking inertia drag distance value for each time period. The specific steps for obtaining the hoisting behavior risk index for each time period of the gantry crane are as follows: comprehensively analyze the hoisting behavior time series data of the gantry crane respectively to obtain the hoisting risk assessment index set for each time period of the gantry crane, including the operation consistency risk index and the load coupling disturbance risk index; and comprehensively analyze the hoisting risk assessment index set for each time period of the gantry crane to obtain the hoisting behavior risk index for each time period of the gantry crane.

[0012] Further, the operation video stream timing data includes operation video stream data for each period. The operation video stream data includes a number of frames of operation image data. The operation image data specifically includes the pixel values and two-dimensional coordinates of each pixel point in the operation image. The specific steps to obtain the operation environment interference aggregation index for each period of the gantry crane are as follows: Input the pixel values and two-dimensional coordinates of each pixel point in each frame of the operation image of the gantry crane for each period into a pre-trained operation risk identification model for comprehensive analysis to obtain an operation risk assessment index set for each period of the gantry crane, including an operation obstacle interference index, a non-operation target intrusion index, and a container stacking interference index; conduct comprehensive analysis on the operation risk assessment index set for each period of the gantry crane to obtain the operation environment interference aggregation index for each period of the gantry crane.

[0013] Further, the operation risk identification model is specifically a vision transformer network. The vision transformer network includes an input layer, an image encoding layer, a vision transformer encoder layer, a risk attention feature extraction layer, and a task output layer. The specific steps to obtain the operation risk assessment index set for each period of the gantry crane are as follows: In the input layer of the vision transformer network, receive the operation video stream timing data of the gantry crane and perform preprocessing; in the image encoding layer of the vision transformer network, perform image block encoding processing on the preprocessed operation video stream timing data of the gantry crane to obtain an image block embedding vector sequence for each frame of the operation image of the gantry crane for each period; in the vision transformer encoder layer of the vision transformer network, perform multi-layer attention transformation encoding processing on the image block embedding vector sequence for each frame of the operation image of the gantry crane for each period to obtain an image block global feature vector sequence for each frame of the operation image of the gantry crane for each period; in the risk attention feature extraction layer of the vision transformer network, perform feature extraction on the image block global feature vector sequence for each frame of the operation image of the gantry crane for each period to obtain a risk region feature vector set for each frame of the operation image of the gantry crane for each period; in the task output layer of the vision transformer network, perform branch prediction processing on the risk region feature vector set for each frame of the operation image of the gantry crane for each period to obtain the operation obstacle interference index, the non-operation target intrusion index, and the container stacking interference index for each period of the gantry crane, that is, the operation risk assessment index set.

[0014] Further, the specific steps for taking preset warning measures for the gantry crane based on the risk coupling determination index of each time period are as follows: Read the risk coupling determination index of each time period of the gantry crane and conduct comprehensive analysis to obtain the risk coupling determination index of the next time period of the gantry crane; Compare and analyze the risk coupling determination index of the next time period of the gantry crane with the preset risk coupling determination index threshold interval respectively; If the risk coupling determination index of the next time period of the gantry crane is lower than the lower limit of the preset risk coupling determination index threshold interval, it is marked as low - level risk and the first warning measure is taken;

[0015] If the risk coupling determination index of the next time period of the gantry crane is within the preset risk coupling determination index threshold interval, it is marked as medium - level risk and the second warning measure is taken; If the risk coupling determination index of the next time period of the gantry crane is higher than the upper limit of the preset risk coupling determination index threshold interval, it is marked as high - level risk and the third warning measure is taken.

[0016] A monitoring and warning system for a container gantry crane includes: a monitoring data acquisition module for continuously acquiring the monitoring time - series data of the gantry crane, where the monitoring time - series data includes structure operation time - series data, lifting operation time - series data, and operation video stream time - series data; a risk extraction and processing module for respectively performing risk extraction and processing on the monitoring time - series data of the gantry crane to obtain a risk assessment index set for each time period of the gantry crane, including a structure risk aggregation index, a lifting operation risk index, and an operation environment interference aggregation index; a comprehensive risk assessment module for comprehensively analyzing the risk assessment index set for each time period of the gantry crane to obtain the risk coupling determination index for each time period of the gantry crane; a risk intelligent warning module for taking preset warning measures for the gantry crane based on the risk coupling determination index of each time period.

[0017] The present invention has the following beneficial effects:

[0018] (1) The monitoring and warning method for the container gantry crane realizes multi - dimensional data fusion and risk assessment by comprehensively analyzing the structure operation data, lifting operation data, and operation video stream data of the gantry crane, and conducts coupling analysis on multiple risk assessment index sets for each time period, thereby obtaining a more accurate and comprehensive risk judgment, ensuring the safety of the gantry crane during operation. For example, by combining the structure risk and the lifting operation risk, potential instability risks caused by factors such as load eccentricity and spreader sway can be identified, significantly improving the prediction and warning ability for potential safety hazards, effectively guaranteeing the safety during the crane operation process, and avoiding misjudgment and missed judgment.

[0019] (2) The monitoring and early warning method of this container gantry crane analyzes the operation video stream data of the gantry crane through deep learning by introducing a vision transformer network, so as to accurately identify potential interference factors in the operation environment. By processing the complex information in each frame of the video, through image block coding, global semantic modeling, and local attention mechanism, problems such as operation obstacle interference, entry of non-operation targets, and container stacking interference are deeply explored. Furthermore, it can identify and evaluate the dynamic risks in the operation environment with higher accuracy, thereby improving the real-time monitoring ability of the operation environment, providing more accurate data support for early warning, and ensuring the timeliness and accuracy of early warning.

[0020] (3) The monitoring and early warning method of this container gantry crane automatically predicts the risk status of the next time period by performing trend analysis and risk change rate adjustment on the risk coupling determination index of each time period of the gantry crane, and triggers different levels of early warning measures according to the prediction results to ensure operation safety. Thereby, it improves the timeliness and accuracy of safety response, and can adjust the early warning strategy in real time according to the actual operation situation, thus improving the timeliness and accuracy of safety response, and further avoiding the lag and false alarm rate of early warning. Subsequently, it ensures that the gantry crane can achieve accurate early warning in a complex and changeable operation environment, thereby improving the safety and operation efficiency during the operation process.

[0021] (4) The monitoring and early warning system of this container gantry crane realizes real-time monitoring and intelligent early warning of multi-dimensional risks during the operation process of the container gantry crane through modular design, thereby improving the automation and intelligence level of the system. For example, the risk extraction and processing module conducts detailed risk extraction and analysis on various types of data to obtain the structural risk aggregation index, lifting behavior risk index, and operation environment interference aggregation index of each time period. Through the collaborative analysis of multi-source data, it comprehensively evaluates various potential risks that may exist during the operation process, so as to reduce the risk of human operation errors, and further improve the system's autonomous decision-making and real-time response capabilities. Subsequently, it provides more accurate, real-time, and intelligent guarantee for the safe operation of the container gantry crane.

[0022] Of course, it is not necessary for any product implementing the present invention to achieve all the above-mentioned advantages simultaneously. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 It is a flowchart of the monitoring and early warning method of a container gantry crane according to the present invention.

[0024] Figure 2 It is a specific step flowchart for obtaining the operation environment interference aggregation index of each time period of the gantry crane in the monitoring and early warning method of a container gantry crane according to the present invention.

[0025] Figure 3 It is the time sequence diagram of the operation risk assessment index set of the gantry crane in the monitoring and early warning method of a container gantry crane of the present invention.

[0026] Figure 4 It is the block diagram of the monitoring and early warning system of a container gantry crane of the present invention. Specific embodiments

[0027] Please refer to Figure 1 , the embodiment of the present invention provides a technical solution: a monitoring and early warning method for a container gantry crane, including the following steps: continuously obtaining the monitoring time sequence data of the (container) gantry crane, and the monitoring time sequence data includes the structure operation time sequence data, the hoisting behavior time sequence data, and the operation video stream time sequence data; respectively performing risk extraction processing on the monitoring time sequence data of the gantry crane to obtain the risk assessment index set of each period of the gantry crane, including the structure risk aggregation index, the hoisting behavior risk index, and the operation environment interference aggregation index; comprehensively analyzing the risk assessment index set of each period of the gantry crane to obtain the risk coupling determination index of each period of the gantry crane; and taking a preset early warning measure for the gantry crane based on the risk coupling determination index of each period.

[0028] The specific formula for calculating the risk coupling determination index of a certain period of the gantry crane is as follows: ; where FpD is the risk coupling determination index of a certain period of the gantry crane, GxH is the structure risk aggregation index of a certain period of the gantry crane, α1 is the structure risk adjustment coefficient stored in the database, DxF is the hoisting behavior risk index of a certain period of the gantry crane, ZgR is the operation environment interference aggregation index of a certain period of the gantry crane, α2 is the behavior environment aggregation adjustment coefficient stored in the database, and α3 is the structure behavior aggregation adjustment coefficient stored in the database.

[0029] It should be explained that in the formula This term is used to adjust the mutual influence between the structure risk aggregation index and the hoisting behavior risk index, and control the amplification or suppression of the interaction between the structure risk aggregation index and the hoisting behavior risk index on the risk coupling determination index.

[0030] α1, α2, and α3 can be obtained through the following steps: Using historical data, combined with the structural risk aggregation index, the hoisting behavior risk index, and the operation environment interference aggregation index, perform statistical regression analysis to quantify the specific impact of each factor on the risk coupling determination index, thereby fitting the initial weight values. Secondly, use the sensitivity analysis method to adjust the value range of each coefficient and observe its impact on the risk coupling determination result to ensure the stability and rationality of the model. Based on the actual situation, correct and optimize the preliminarily fitted coefficients, and finally determine the coefficient values.

[0031] Specifically, the structural risk aggregation index for each time period of the structural operation time series data includes the main girder impact response index, the leg force distribution value, the main girder strain variation index, and the vibration coordination index. The specific steps to obtain the structural risk aggregation index for each time period of the gantry crane are as follows: Conduct a comprehensive analysis of the structural operation time series data of the gantry crane respectively to obtain the structural operation risk assessment index set for each time period of the gantry crane, including the structural abnormal response intensity index and the structural coordination and stability index; and conduct a comprehensive analysis of the structural operation risk assessment index set for each time period of the gantry crane to obtain the structural risk aggregation index for each time period of the gantry crane.

[0032] Among them, the main girder impact response index is the structural response caused by the impact or load change of the main girder. It can obtain the vibration acceleration values at each time point within this time period through an acceleration sensor, and perform statistical analysis and mean processing respectively to obtain the maximum vibration acceleration value and the vibration acceleration mean value, and perform a ratio processing, that is, the maximum vibration acceleration value / the vibration acceleration mean value. The obtained result is the main girder impact response index.

[0033] The leg force distribution value is the degree of imbalance of the pressure borne by each leg of the container gantry crane. That is, it can obtain the pressure borne by each leg at each time point within this time period through a pressure sensor, perform standard deviation processing, and perform mean processing based on the standard deviation processing result. The obtained result is the leg force distribution value.

[0034] The main girder strain variation index is used to measure the strain value dispersion degree of multiple strain measurement points on the main girder of the container gantry crane within the same monitoring time period during this time period. It can be obtained by arranging multiple resistance strain gauges on the surface of the main girder, and obtaining the strain values of each strain measurement point at each time point within this time period, and performing mean and standard deviation processing respectively to obtain the strain mean value and the strain standard deviation at each time point, and performing a ratio processing, that is, the strain standard deviation / the strain mean value, and performing mean processing based on the ratio processing result. The obtained result is the main girder strain variation index.

[0035] The vibration coordination index measures the similarity and coordination degree among the vibration amplitudes of key parts of multiple structures (such as the mid-span area of the main girder, the connection between the main girder and the leg, the connection point between the bottom of the leg and the track, etc.). It can obtain the vibration amplitude of each key part of each structure at each time point through vibration sensors, calculate the correlation value of the vibration amplitudes of adjacent key parts of each structure at each time point based on the Pearson correlation coefficient, and perform mean processing. The resulting value is the vibration coordination index.

[0036] The specific steps to obtain the structural operation risk assessment index set for each period of the gantry crane are as follows: Read the main girder impact response index and the main girder strain variation index for each period of the gantry crane, and conduct comprehensive analysis (first perform normalization processing, then perform weighted processing based on the standardized processing results, and map the weighted processing results to between 0 and 1 based on the sigmoid function) to obtain the structural abnormal response intensity index for each period of the gantry crane; Read the leg force distribution value and the vibration coordination index for each period of the gantry crane, and conduct comprehensive analysis (first perform normalization processing, then perform weighted processing based on the standardized processing results, and map the weighted processing results to between 0 and 1 based on the sigmoid function) to obtain the structural coordination and stability index for each period of the gantry crane.

[0037] The specific formula for calculating the structural risk aggregation index for a certain period of the gantry crane is as follows: ; where, GxH is the structural risk aggregation index for a certain period of the gantry crane, YcX is the structural abnormal response intensity index for a certain period of the gantry crane, β1 is the abnormal response adjustment coefficient stored in the database, XtW is the structural coordination and stability index for a certain period of the gantry crane, β2 is the coordination and stability adjustment coefficient stored in the database, β3 is the interaction coupling adjustment coefficient stored in the database, and β4 is the coupling sensitivity adjustment coefficient stored in the database.

[0038] It should be explained that in the formula This term is used to adjust the risk linkage amplification effect that is prone to occur under the critical instability condition of the gantry crane structure when the structural abnormal response intensity index is high and the structural coordination and stability index is low, and to avoid the structural risk aggregation index being too high or too low.

[0039] β1, β2, β3, and β4 can be obtained through the following steps: Based on historical data, determine the initial influence weights of each variable (structural abnormal response intensity index, structural coordination and stability index) on the structural risk aggregation index through statistical regression analysis. Then, use the sensitivity analysis method to adjust the value range of the coefficients to evaluate the stability and applicability of these parameters to the formula output. Next, further fit the weights through model optimization (such as machine learning algorithms or multi-objective optimization) to ensure that the formula can accurately reflect the risk state of the actual crane structure.

[0040] In this implementation scheme, through the fusion analysis of multiple structural data, the limitations of a single index are avoided, and the key factors in the structural operation are accurately considered, so that complex structural risks can be effectively identified. Secondly, statistical regression analysis is carried out using historical data to quantify the influence of different structural risk factors, and the adjustment coefficient is adjusted through sensitivity analysis to ensure the real-time update and adaptability of the structural risk assessment formula, so as to be able to dynamically respond to the changes of gantry cranes in different working environments, ensure the real-time and accuracy of risk assessment, and by introducing an interactive coupling adjustment coefficient, the structural risk aggregation index is finely adjusted, thus effectively preventing the imbalance of the overall risk assessment caused by the abnormality of a single risk index, providing a more accurate risk judgment. Finally, through a more refined structural risk aggregation analysis, the risks of potential structural instability or excessive stress during the operation of the crane are identified, providing a more accurate decision-making basis for the warning system, and then effectively improving the overall safety guarantee ability of the gantry crane, and then preventing accidents caused by inaccurate risk judgment.

[0041] Specifically, the time-series data of the lifting behavior includes the jerk frequency value of the spreader, the load eccentricity value, the load sway coordination index, the spreader inertial energy accumulation index, the control smoothness index, and the braking inertia drag distance value in each time period. The specific steps to obtain the lifting behavior risk index of the gantry crane in each time period are as follows: comprehensively analyze the time-series data of the lifting behavior of the gantry crane respectively to obtain the lifting risk assessment index set of the gantry crane in each time period, including the operation consistency risk index and the load coupling disturbance risk index; comprehensively analyze (i.e., perform weighted processing) the lifting risk assessment index set of the gantry crane in each time period to obtain the lifting behavior risk index of the gantry crane in each time period.

[0042] Among them, the jerk frequency value of the spreader is the number of acceleration mutations of the spreader in this time period, which is used to identify the operation smoothness during the lifting process. It can obtain the acceleration value of the spreader at each time point in this time period through an acceleration sensor, and judge and analyze it with a set acceleration threshold (which can be obtained by obtaining the historical acceleration values of the spreader at several historical time points and performing average processing, and setting the result as the acceleration threshold). Mark the time points higher than the acceleration threshold as an acceleration mutation once, and perform statistical analysis. The obtained result is the jerk frequency value of the spreader.

[0043] The load eccentricity value is the degree of offset between the actual load-bearing mass center and the geometric center, reflecting whether there is asymmetric loading of the suspended load. It can obtain the four-point force values of the four hoisting point wire ropes and the corresponding three-dimensional position coordinates through a tension sensor, calculate the three-dimensional position coordinates of the load center based on the static moment balance, obtain the three-dimensional position coordinates of the geometric center of the load, and analyze and calculate the three-dimensional position coordinates of the load center and the three-dimensional position coordinates of the geometric center based on the Euclidean distance formula. The obtained result is the load eccentricity value.

[0044] The load sway coordination index is the synchronization degree of the forward and backward swing angles and the left and right swing angles during the swinging process of the spreader. It can obtain the pitch angle and roll angle at each time point during this period through an inertial measurement unit, calculate the time synchronization degree between the two based on Pearson correlation, and perform mean processing. The obtained result is the load sway coordination index.

[0045] The spreader inertial energy accumulation index is the total inertial kinetic energy accumulated by the spreader due to continuous movement during this period. It can collect the horizontal acceleration signal in real time through the spreader inertial measurement unit, combine the current load mass (which can obtain the total vertical tension borne by the spreader through a tension sensor and analyze the load mass based on Newton's second law), calculate the square value of the acceleration per unit time during this period and integrate to obtain its inertial kinetic energy accumulation, which is used to identify whether there is a continuous accumulation state of energy during the operation of the spreader, and further evaluate the potential structural excitation and impact risks during the hoisting process.

[0046] The operation smoothness index is the degree of deviation between the running speeds of the trolley and the spreader and the trolley speed reference value (by obtaining the trolley speeds at several historical time points and performing mean processing) and the spreader speed reference value (by obtaining the spreader speeds at several historical time points and performing mean processing) during this period. It can obtain the running speeds of the trolley and the spreader at each time point during this period through a speed sensor, perform mean processing to obtain the average running speed of the trolley and the average running speed of the spreader, and perform ratio processing with the trolley speed reference value and the spreader speed reference value respectively. Based on the ratio processing results, perform weighted processing. The obtained result is the operation smoothness index.

[0047] The braking inertia drag distance value is the additional distance dragged by inertia after the stop commands for the trolley and the spreader are issued during this period. It can obtain the displacement change amount from the command issuance to the complete stop of the trolley or the spreader through a displacement sensor, and perform weighted processing on the displacement change amounts of the trolley and the spreader. The obtained result is the braking inertia drag distance value.

[0048] The specific steps to obtain the hoisting risk assessment index set for each period of the gantry crane are as follows: Read the jerk frequency value of the spreader, the control smoothness index, and the braking inertia drag distance value for each period of the gantry crane, and conduct a comprehensive analysis (first perform standardization processing, and then perform weighted processing based on the results of the standardization processing) to obtain the operation consistency risk index (operation execution stability and dynamic response consistency) for each period of the gantry crane; Read the load eccentricity value, the load sway coordination index, and the spreader inertial energy accumulation index for each period of the gantry crane, and conduct a comprehensive analysis (first perform standardization processing, and then perform weighted processing based on the results of the standardization processing) to obtain the load coupling disturbance risk index (the degree of risk of the load state disturbing the sway behavior of the spreader) for each period of the gantry crane.

[0049] In this implementation plan, through the real-time monitoring of multiple key parameters such as the spreader jerk frequency, load eccentricity, load sway coordination, and inertial energy accumulation, it is possible to comprehensively evaluate the stability and safety of various behaviors during the hoisting process, thus ensuring the accurate assessment of the crane hoisting behavior. Secondly, through the analysis of indicators such as the spreader jerk frequency and the control smoothness index, it helps to judge whether the crane operation is stable and whether there are abnormal situations such as sudden acceleration changes, and avoids potential safety hazards such as structural instability or spreader collision caused by inconsistent operation or unstable load during the operation process. Finally, the comprehensive analysis of the hoisting behavior data can help the system make more detailed risk predictions during the real-time monitoring process. For example, when the spreader jerks, the load is unstable or sways, it can quickly identify and evaluate potential risks, trigger the warning mechanism, and avoid the occurrence of accidents. This refined risk identification and warning ability can effectively improve the safety and response speed during the entire crane operation process, and through the weighted processing and dynamic evaluation of different risk indicators, it is possible to adjust the risk index according to the actual operation status and optimize the warning strategy.

[0050] Specifically, such as Figure 2As shown in the figure, the time-series data of the operation video stream includes the operation video stream data for each period. The operation video stream data includes several frames of operation image data. The operation image data is specifically the pixel value and two-dimensional coordinates of each pixel point in the operation image. The specific steps to obtain the operation environment interference aggregation index for each period of the gantry crane are as follows: Input the pixel value and two-dimensional coordinates of each pixel point in each frame of the operation image for each period of the gantry crane into the pre-trained operation risk recognition model for comprehensive analysis to obtain the operation risk assessment index set for each period of the gantry crane, including the operation obstacle interference index (identifying the interference situation of static objects in the hoist lowering path), the non-operation target intrusion index (identifying dynamic intrusion behaviors such as unplanned entry of personnel and vehicles in the lifting operation area), and the container stacking interference index (identifying abnormal container stacking states at the target landing point of the hoist); conduct comprehensive analysis on the operation risk assessment index set for each period of the gantry crane to obtain the operation environment interference aggregation index for each period of the gantry crane.

[0051] Among them, the specific formula for calculating the operation environment interference aggregation index for a certain period of the gantry crane is as follows: ; where ZgR is the operation environment interference aggregation index for a certain period of the gantry crane, ZaW is the operation obstacle interference index for a certain period of the gantry crane, FmB is the non-operation target intrusion index for a certain period of the gantry crane, μ1 is the superposition adjustment coefficient stored in the database, DgR is the container stacking interference index for a certain period of the gantry crane, and μ2 is the stacking interference adjustment coefficient stored in the database.

[0052] It should be explained that μ1 and μ2 can be obtained through the following steps: Based on historical data, determine the initial influence weights of each variable (such as the coupling interaction between the operation obstacle interference index and the non-operation target intrusion index, and the container stacking interference index) on the operation environment interference aggregation index through statistical regression analysis. Then, use the sensitivity analysis method to adjust the value range of the coefficients to evaluate the stability and applicability of these parameters to the formula output. Next, further fit the weights through model optimization (such as multi-objective optimization) to ensure that the formula can accurately reflect the interference state of the actual operation environment.

[0053] The specific implementation example of calculating the operation environment interference aggregation index for a certain period of the gantry crane is as follows. The following data is available: including the operation obstacle interference index, non-operation target intrusion index, and container stacking interference index for 5 periods of the gantry crane, as shown in Table 1 and Figure 3 :

[0054] Table 1 Example of time-series set data of operation risk assessment index of gantry crane

[0055]

[0056] The interaction adjustment coefficient μ1 stored in the database is approximately: 0.317;

[0057] The stacking interference adjustment coefficient μ2 stored in the database is approximately: 0.573;

[0058] Substitute the data in Table 1 and the above adjustment coefficients into the specific formula for calculating the operating environment interference aggregation index of the gantry crane for a certain period, and we get:

[0059] The operating environment interference aggregation index of the gantry crane for the first period = ln(1 + 0.317×(0.103 + 1)×(0.042 + 1))×√(exp(0.537×(1 + 0.121))) ≈ 0.427;

[0060] The operating environment interference aggregation index of the gantry crane for the second period = ln(1 + 0.317×(0.254 + 1)×(0.126 + 1))×√(exp(0.537×(1 + 0.194))) ≈ 0.521;

[0061] The operating environment interference aggregation index of the gantry crane for the third period = ln(1 + 0.317×(0.346 + 1)×(0.413 + 1))×√(exp(0.537×(1 + 0.253))) ≈ 0.676;

[0062] The operating environment interference aggregation index of the gantry crane for the fourth period = ln(1 + 0.317×(0.268 + 1)×(0.234 + 1))×√(exp(0.537×(1 + 0.401))) ≈ 0.601;

[0063] The operating environment interference aggregation index of the gantry crane for the fifth period = ln(1 + 0.317×(0.153 + 1)×(0.092 + 1))×√(exp(0.537×(1 + 0.687))) ≈ 0.475.

[0064] The operation risk identification model is specifically a Vision Transformer network (Vision Transformer). The Vision Transformer network includes an input layer, an image encoding layer, a Vision Transformer encoder layer, a risk attention feature extraction layer, and a task output layer. The specific steps to obtain the operation risk assessment index set for each time period of the gantry crane are as follows: In the input layer of the Vision Transformer network, receive the operation video stream time series data of the gantry crane (the pixel values and two-dimensional coordinates of each pixel point in each frame of the operation image for each time period), and perform preprocessing; in the image encoding layer of the Vision Transformer network, perform image block encoding processing on the preprocessed operation video stream time series data of the gantry crane (divide each frame of the operation image into several image blocks according to a fixed size, then perform linear mapping processing on the pixel values of each image block, convert the two-dimensional image block into a vector representation of a fixed dimension, that is, an image block embedding vector. At the same time, embed the two-dimensional coordinate position of each image block in the original image into the corresponding vector, and use the position encoding method to embed the position information into the corresponding vector), to obtain the image block embedding vector sequence of each frame of the operation image for each time period of the gantry crane (several image block embedding vector sequences); in the Vision Transformer encoder layer of the Vision Transformer network, perform multi-layer attention transformation encoding processing on the image block embedding vector sequence of each frame of the operation image for each time period of the gantry crane (each layer of the Transformer encoder includes a multi-head self-attention mechanism sub-module, a feed-forward neural network sub-module, and a residual connection and layer normalization module. In each encoder layer, first establish the dependence relationship between the current image block and all other image blocks through the multi-head self-attention mechanism, and enable each image block vector to fuse the context information within the entire image range through the global attention calculation mechanism. Then, through the feed-forward neural network, that is, two fully connected neural networks, perform non-linear transformation on the sub-attention output result to enhance the feature expression ability. At the same time, introduce residual connection and layer normalization operations after the output of each sub-module to maintain feature stability and improve the model training efficiency, so that the image block vectors gradually obtain deeper and richer global semantic information. Each vector not only retains the position information of its original image block but also fuses the context features from other regions in the entire image), to obtain the image block global feature vector sequence of each frame of the operation image for each time period of the gantry crane;In the risk attention feature extraction layer of the vision transformer network, feature extraction is performed on the sequence of global feature vectors of image patches in each frame of the operation image for each time period of the gantry crane (call the target detection network stored in the database to automatically identify and locate the spreader in the image, so as to obtain the rectangular area range of the spreader in the current frame image. According to the position of the frame in the image space, several subordinate sub-regions related to the spreader space are automatically generated, including: the descending path region composed of several rows of image patches extending downward with the spreader frame as the center, the spreader landing point region composed of image patches aligned directly below the spreader frame, and the operation buffer region formed by image patches within a certain range around the spreader frame. The above regions can be accurately calibrated in the image patch sequence through simple pixel coordinate expansion and image patch index mapping, and the set of image patch vectors corresponding to the calibrated risk sub-regions is extracted from the global feature sequence of image patches. And a local self-attention mechanism is constructed for each sub-region. For example, three groups of vector representations of query, key, and value are generated for each sub-region image patch set, and the semantic correlation weights between image patches are calculated based on the softmax attention function. Then, the original feature vectors within the region are weighted and fused to obtain an enhanced image patch sequence. Then, further pooling processing is performed on this enhanced sequence, including methods such as weighted average or maximum response selection, and a region feature vector with the same dimension as the original image patch is output as a high-dimensional semantic expression representing the potential risk structure, abnormal target, or dynamic interference state of the sub-region. For example, the region feature vector of the operation obstacle region contains feature dimensions such as regional surface continuity, texture mutation density, and gray gradient variance that characterize the existence of obstacles. The region feature vector of the non-operation target intrusion region contains structural activation features such as contour symmetry, target movement trajectory pattern, and height-width ratio deviation that reflect the intrusion behavior of dynamic targets. And the region feature vector of the container stacking interference region contains region discriminant features such as edge line alignment, height gradient anomaly, and irregular corner distribution that characterize the structural stability of container stacking), and the risk region feature vector set of each frame of the operation image for each time period of the gantry crane is obtained;In the task output layer of the vision transformer network, branch prediction processing is performed on the set of risk area feature vectors of each frame of the operation image in each period of the gantry crane (an independent task branch network structure is set for each type of operation risk, corresponding to the operation obstacle interference risk index, the non-operation target intrusion risk index, and the container stacking interference risk index respectively. Each task branch includes a set of fully connected neural network modules for risk discrimination. The input is the feature vector of the corresponding area, and then through multi-layer linear transformation, non-linear activation functions such as ReLU or GELU, and regularization processing such as Dropout, deep semantic mapping and pattern classification compression are performed on the input features. Finally, a scalar value is output, and the scalar value is normalized to between 0 and 1 through the Sigmoid activation function, which is used to represent the occurrence probability or confidence intensity of the risk corresponding to the current area in this frame of image. For example, after the feature vector of the operation obstacle area is input into the obstacle risk discrimination network, if the output value is close to 1, it means that the system highly determines that there is a significant object interference risk in the area directly below the spreader. If the output of the non-operation target intrusion branch is close to 1, it means that there is a clear humanoid target or dynamic object in the spreader operation area in the current image; when the output of the container stacking interference branch is relatively high, it indicates that the container structure in the current spreader landing area has abnormal stacking features such as tilting and misalignment. The three risk indexes are calculated for each frame of image respectively, and finally, in a monitoring period, aggregation analysis is performed on the three index values of all frames, including strategies such as average value, maximum value, and sliding weighting, to obtain the operation obstacle interference index, the non-operation target intrusion index, and the container stacking interference index of this period, which constitute the operation risk assessment index set of the current operation cycle of the gantry crane). The operation obstacle interference index, the non-operation target intrusion index, and the container stacking interference index of each period of the gantry crane are obtained, that is, the operation risk assessment index set.

[0065] Among them, the input layer is used to receive the operation video stream image of the gantry crane and perform image standardization and pixel information preprocessing.

[0066] The image encoding layer is used to receive the operation video stream image of the gantry crane and perform image standardization and pixel information preprocessing.

[0067] The vision transformer encoder layer is used to model the global semantic relationship between image patches through a multi-layer attention mechanism and generate a sequence of global feature vectors of image patches with context awareness ability.

[0068] The risk attention feature extraction layer is used to delimit the risk area based on the spreader position, perform local attention enhancement on the image patches in each sub-area, and extract the risk area feature vectors.

[0069] The task output layer is used to extract the feature vectors of three regions and input them into their respective risk discrimination networks, and output the interference index of operation obstacles, the intrusion index of non-operation targets, and the interference index of container stacking.

[0070] And the pre-training process of the vision transformer network is as follows:

[0071] Obtain the labeled image dataset, including a large number of labeled image data with targets such as spreaders, containers, and operation areas, preprocess the images and target location information, and divide the labeled image dataset into a sample training set and a sample validation set.

[0072] Initialize the vision transformer network, including the weight and bias parameters of the image patch embedding encoding layer, the transformer encoder layer, and the task output layer, and use random initialization or load existing model parameters.

[0073] Train based on the sample training set, set the number of loops (such as 100 times), and in each training loop, perform the forward propagation process (input the sample data into each layer of the vision transformer network, and then output the predicted values of the three risk indices in the current batch), calculate the loss function (calculate the loss function based on the difference between the predicted output value and the true label, for example, use the binary cross-entropy loss function to calculate the prediction error of the operation obstacle interference index, or use the mean square error function to measure the error size), backpropagation, and parameter update (automatically calculate the gradient values of each neural network parameter, and update the weight and bias parameters in the network based on the set optimization algorithm, such as Adam or SGD).

[0074] After each training loop, perform evaluation and analysis based on the sample validation set. That is, for each frame of validation image, the model will sequentially generate image patch embedding vectors, global feature sequences, and risk region feature representations, and output three risk prediction indices: the interference index of operation obstacles, the intrusion index of non-operation targets, and the interference index of container stacking. Next, compare and analyze the predicted three risk indices with the true labels annotated in the validation set samples, calculate the validation loss values output for each type of risk, such as cross-entropy loss or mean square error, and at the same time, count multiple performance evaluation metrics, such as average prediction accuracy (Accuracy), AUC (area under the curve), Precision, Recall, F1 score, etc., to comprehensively measure the performance of the model in each risk identification task. In addition, the distribution trend, abnormal prediction frequency, and gradient fluctuation degree of the output values of each risk index in the validation set will also be monitored to determine whether the model has a tendency of overfitting or under-training, and an early stopping mechanism (Early Stopping) is set to automatically abort subsequent training in the case of no obvious improvement in consecutive rounds of validation metrics, retain the optimal model parameters, and end the training process to obtain the trained network model.

[0075] In this implementation scheme, by combining the operation video stream data with the vision transformer network for multi-dimensional risk identification, it can not only handle the static obstacles in the lifting operation during each period, but also dynamically identify the intrusion of non-operation targets and the abnormal situation of container stacking, so as to comprehensively monitor the crane operation environment, and then improve the comprehensiveness of operation safety monitoring. Secondly, through the multi-layer attention mechanism of the vision transformer network, it can deeply extract features at the image patch level, accurately identify the potential risks of the spreader descent path, the landing area and the surrounding buffer area. For example, when identifying the interference of operation obstacles, the system can analyze features such as texture changes, surface continuity, and gray-scale gradient in the image to detect possible object interference, thus effectively reducing the risks of false alarms and missed alarms. In addition, through the pre-training and adaptive optimization of the vision transformer network, the accuracy of the model can be continuously improved in different operation environments, and the misjudgment caused by overfitting or insufficient training can be effectively reduced, improving the long-term stability and accuracy. Finally, by adopting the global feature and regional feature extraction method based on image patches, the risk assessment result not only has high accuracy, but also has strong interpretability, so that the operator can intuitively understand the safety risk level of the current operation area, further improving the operation management efficiency and safety.

[0076] Specifically, the specific steps for taking preset warning measures for a gantry crane based on the risk coupling determination index for each time period are as follows: Read the risk coupling determination index for each time period of the gantry crane and conduct comprehensive analysis to obtain the risk coupling determination index for the next time period of the gantry crane; Compare and analyze the risk coupling determination index for the next time period of the gantry crane with the preset risk coupling determination index threshold range respectively; If the risk coupling determination index for the next time period of the gantry crane is lower than the lower limit of the preset risk coupling determination index threshold range, then (mark the next time period) as low risk and take the first warning measure, specifically record the current state as stable operation and do not trigger any active control intervention; If the risk coupling determination index for the next time period of the gantry crane is within the preset risk coupling determination index threshold range, then (mark the next time period) as medium risk and take the second warning measure, specifically trigger a prompt warning signal, such as a voice prompt, to remind the operator to pay attention to the operation state, and automatically turn on the frequency enhancement monitoring mode, accelerate the data refresh frequency, track the risk fluctuation in real time, and restrict some risk actions, such as prohibiting rapid hoisting and restricting the left and right extreme lateral movement of the spreader; If the risk coupling determination index for the next time period of the gantry crane is higher than the upper limit of the preset risk coupling determination index threshold range, then (mark the next time period) as high risk and take the third warning measure, specifically trigger a red alarm signal and an audible and visual warning, and simultaneously inform the operator of the gantry crane of the alarm information, and start the automatic intervention mechanism, including: decelerating operation, pausing some actions, locking the operation permission for high-risk areas, and record this time period as a high-risk historical sample for risk model adjustment and working condition reconstruction.

[0077] Among them, the specific steps for obtaining the risk coupling determination index for the next time period of the gantry crane are as follows: Conduct trend analysis on the risk coupling determination index for each time period of the gantry crane (such as the change rate between adjacent two time periods, that is, the difference divided by the value of the previous time period, and then conduct weighted processing) to obtain several groups of predicted time period (such as a group of 3 consecutive time periods) risk coupling determination index change rates for the gantry crane; And conduct mean value processing on the risk coupling determination index for each time period of the gantry crane (that is, conduct mean value processing on the risk coupling determination index for 3 consecutive time periods) to obtain the operation risk determination benchmark value for each group of predicted time periods of the gantry crane, and conduct risk trend adjustment processing in combination with the risk coupling determination index change rate, that is, according to the risk coupling determination index change rate (rising or falling), adjust the benchmark value of this group up and down to obtain the predicted risk coupling determination index for each group of predicted time periods of the gantry crane, and use the corresponding group of predicted time periods as the risk coupling determination index for the next time period.

[0078] In this implementation scheme, through trend analysis and risk change rate adjustment based on the risk coupling determination index for each time period, it is possible to accurately predict the risk state of the gantry crane in the next time period, and take corresponding early warning measures according to the prediction results. Furthermore, it can respond in real time to changes in the operating environment, avoid the lag and false alarm rate of early warnings, and ensure a more timely safety response. Secondly, according to the comparison between the risk coupling determination index and the preset threshold, the system can divide the risk state into low, medium, and high risks, and trigger corresponding early warning measures, so as to refine the risk control strategy, help operators and administrators take actions in advance, reduce the occurrence of emergencies, and thus enhance predictability and foresight. Finally, through the dynamic adjustment of the risk change rate, it is possible to intelligently optimize the risk management strategy, track risk fluctuations in real time, and make rapid responses according to the actual situation, further improving the safety and stability of gantry crane operations.

[0079] Please refer to Figure 4 , an embodiment of the present invention provides a technical solution: a monitoring and early warning system for a container gantry crane, including: a monitoring data acquisition module for continuously acquiring the monitoring time-series data of the gantry crane, where the monitoring time-series data includes structural operation time-series data, hoisting behavior time-series data, and operation video stream time-series data; a risk extraction and processing module for respectively performing risk extraction and processing on the monitoring time-series data of the gantry crane to obtain a set of risk assessment indexes for each time period of the gantry crane, including a structural risk aggregation index, a hoisting behavior risk index, and an operation environment interference aggregation index; a comprehensive risk assessment module for comprehensively analyzing the set of risk assessment indexes for each time period of the gantry crane to obtain a risk coupling determination index for each time period of the gantry crane; and a risk intelligent early warning module for taking preset early warning measures for the gantry crane based on the risk coupling determination index for each time period.

[0080] Although the preferred embodiments of the present invention have been described, those skilled in the art can make additional changes and modifications to these embodiments once they know the basic creative concept. Therefore, the appended claims are intended to be construed as including the preferred embodiments as well as all changes and modifications falling within the scope of the present invention.

[0081] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention also intends to include these modifications and variations.

Claims

1. A monitoring and early warning method for a container gantry crane, characterized in that, The method includes the following steps: Continuously obtain the monitored time-series data of the gantry crane. The monitored time-series data includes structural operation time-series data, lifting behavior time-series data, and operation video stream time-series data. The lifting behavior time-series data includes the jerk frequency value of the spreader, the load eccentricity value, the load sway coordination index, the spreader inertial energy accumulation index, the control smoothness index, and the braking inertia drag distance value for each period. The operation video stream time-series data includes the operation video stream data for each period, which includes a number of frame operation image data, specifically the pixel value and two-dimensional coordinates of each pixel point in the operation image; Respectively perform risk extraction processing on the monitored time-series data of the gantry crane to obtain the risk assessment index set for the corresponding period, including the structural risk aggregation index, the lifting behavior risk index, and the operation environment interference aggregation index. The specific steps are as follows: Respectively perform comprehensive analysis on the structural operation time-series data of the gantry crane to obtain the structural operation risk assessment index set for the corresponding period, including the structural abnormal response intensity index and the structural coordination stability index; And perform comprehensive analysis on the structural operation risk assessment index set for each period of the gantry crane to obtain the structural risk aggregation index for the corresponding period. The specific formula for calculating the structural risk aggregation index of the gantry crane for a certain period is as follows: ； Wherein, GxH, YcX, and XtW are the structural risk aggregation index, the structural abnormal response intensity index, and the structural coordination stability index of the gantry crane for a certain period in sequence, and β1, β2, β3, and β4 are the abnormal response adjustment coefficient, the coordination stability adjustment coefficient, the interaction coupling adjustment coefficient, and the coupling sensitivity adjustment coefficient stored in the database in sequence; Respectively perform comprehensive analysis on the lifting behavior time-series data of the gantry crane to obtain the lifting risk assessment index set for the corresponding period, including the operation consistency risk index and the load coupling disturbance risk index; Perform comprehensive analysis on the lifting risk assessment index set for each period of the gantry crane to obtain the lifting behavior risk index for the corresponding period; Input the pixel value and two-dimensional coordinates of each pixel point in each frame of the operation image of the gantry crane for each period into the pre-trained operation risk recognition model for comprehensive analysis to obtain the operation risk assessment index set for the corresponding period, including the operation obstacle interference index, the non-operation target intrusion index, and the container stacking interference index; Perform comprehensive analysis on the operation risk assessment index set for each period of the gantry crane to obtain the operation environment interference aggregation index for the corresponding period; The specific formula for calculating the operation environment interference aggregation index of a gantry crane during a certain period is as follows: ; where, ZgR is the operation environment interference aggregation index of the gantry crane during a certain period, ZaW is the operation obstacle interference index of the gantry crane during a certain period, FmB is the non-operation target intrusion index of the gantry crane during a certain period, μ1 is the superposition adjustment coefficient stored in the database, DgR is the container stacking interference index of the gantry crane during a certain period, and μ2 is the stacking interference adjustment coefficient stored in the database; Perform comprehensive analysis on the risk assessment index set for each period of the gantry crane to obtain the risk coupling determination index for the corresponding period. The specific formula for calculating the risk coupling determination index of the gantry crane for a certain period is as follows: ； Wherein, FpD, GxH, DxF, and ZgR are the risk coupling determination index, the structural risk aggregation index, the lifting behavior risk index, and the operation environment interference aggregation index of the gantry crane for a certain period in sequence, and α1, α2, and α3 are the structural risk adjustment coefficient, the behavior environment aggregation adjustment coefficient, and the structural behavior aggregation adjustment coefficient stored in the database in sequence; Based on the risk coupling determination index for each time period, preset warning measures are taken for the gantry crane, and the specific steps are as follows: Read the risk coupling determination index for each time period of the gantry crane, and conduct comprehensive analysis to obtain the risk coupling determination index for the next time period of the gantry crane; Compare and analyze the risk coupling determination index for the next time period of the gantry crane with the preset risk coupling determination index threshold range respectively; If the risk coupling determination index for the next time period of the gantry crane is lower than the lower limit of the preset risk coupling determination index threshold range, it is marked as low - risk, and the first warning measure is taken; If the risk coupling determination index for the next time period of the gantry crane is within the preset risk coupling determination index threshold range, it is marked as medium - risk, and the second warning measure is taken; If the risk coupling determination index for the next time period of the gantry crane is higher than the upper limit of the preset risk coupling determination index threshold range, it is marked as high - risk, and the third warning measure is taken.

2. The monitoring and early warning method for a container gantry crane according to claim 1, characterized in that The structural operation time - series data includes the main girder impact response index, leg force distribution value, main girder strain variation index, and vibration coordination index for each time period. The specific steps to obtain the structural operation risk assessment index set for each time period of the gantry crane are as follows: Read the main girder impact response index and main girder strain variation index for each time period of the gantry crane, and conduct comprehensive analysis to obtain the structural abnormal response intensity index for each time period of the gantry crane; Read the leg force distribution value and vibration coordination index for each time period of the gantry crane, and conduct comprehensive analysis to obtain the structural coordination and stability index for each time period of the gantry crane.

3. The monitoring and early warning method of the container gantry crane according to claim 1, characterized in that, The operation risk identification model is specifically a vision transformer network. The vision transformer network includes an input layer, an image encoding layer, a vision transformer encoder layer, a risk attention feature extraction layer, and a task output layer. The specific steps to obtain the operation risk assessment index set for each time period of the gantry crane are as follows: In the input layer of the vision transformer network, receive the operation video stream time - series data of the gantry crane and conduct pre - processing; In the image encoding layer of the vision transformer network, conduct image block encoding processing on the pre - processed operation video stream time - series data of the gantry crane to obtain the image block embedding vector sequence of each frame of the operation image for each time period of the gantry crane; In the vision transformer encoder layer of the vision transformer network, conduct multi - layer attention transformation encoding processing on the image block embedding vector sequence of each frame of the operation image for each time period of the gantry crane to obtain the image block global feature vector sequence of each frame of the operation image for each time period of the gantry crane; In the risk attention feature extraction layer of the vision transformer network, conduct feature extraction on the image block global feature vector sequence of each frame of the operation image for each time period of the gantry crane to obtain the risk area feature vector set of each frame of the operation image for each time period of the gantry crane; In the task output layer of the vision transformer network, branch prediction processing is performed on the risk area feature vector set of each frame of operation image in each period of the gantry crane, and the operation obstacle interference index, non-operation target misentry index, and container stacking interference index in each period of the gantry crane are obtained, that is, the operation risk assessment index set.

4. A monitoring and early warning system for a container gantry crane, which applies the monitoring and early warning method for a container gantry crane described in any one of claims 1-3, characterized in that, Including: A monitoring data acquisition module, which is used to continuously acquire the monitoring time series data of the gantry crane, and the monitoring time series data includes structural operation time series data, lifting behavior time series data, and operation video stream time series data; A risk extraction and processing module, which is used to perform risk extraction and processing on the monitoring time series data of the gantry crane respectively, and obtain the risk assessment index set in each period of the gantry crane, including the structural risk aggregation index, the lifting behavior risk index, and the operation environment interference aggregation index; A comprehensive risk assessment module, which is used to comprehensively analyze the risk assessment index set in each period of the gantry crane, and obtain the risk coupling determination index in each period of the gantry crane; A risk intelligent early warning module, which is used to take preset early warning measures for the gantry crane based on the risk coupling determination index in each period.

Images