Intelligent crane operating environment construction method based on digital twinning technology

Abstract

The present application relates to the field of hoisting equipment safety monitoring, and particularly relates to an intelligent crane operating environment construction method based on digital twin technology. The intelligent crane operating environment construction method based on digital twin technology comprises: constructing a digital twin model; collecting meteorological data, working condition data and response data; formulating a monitoring strategy; performing crane stability analysis to obtain overturning risk and execute overturning risk grading early warning; and real-time mapping and synchronous display of the obtained overturning risk grade and monitoring strategy in the digital twin model. The present application constructs a digital twin model that is real-time synchronized with a physical crane, and dynamically adjusts the monitoring strategy in combination with a historical fault cause-and-effect diagram, so that the sensor sampling frequency and redundant channels can be automatically distributed according to actual risk; the method can greatly reduce redundant collection of low-risk parts while ensuring high-precision monitoring of key parts, reduce data transmission and storage costs, and improve monitoring resource utilization efficiency.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of hoisting machinery safety monitoring, in particular to an intelligent crane operating environment construction method based on digital twin technology. BACKGROUND

[0002] With the development of building industrialization and the increasing demand for large equipment construction, various types of cranes are increasingly developing towards long-arm, high-load, and high-frequency collaborative operation. Traditional safety management methods mainly rely on periodic manual inspection and independent sensor alarm devices installed at key positions. Although these methods can to some extent detect overt risks such as overload and excessive wind speed, they are difficult to capture hidden faults caused by early structural fatigue, hydraulic instability, or uneven leg stress. At the same time, the current monitoring system generally uses a fixed sampling strategy, which cannot automatically adjust the monitoring focus according to the operation scene and environmental conditions, resulting in missing of key data or waste of resources. The existing stability evaluation is mainly based on static specification formulas, and only one calculation is performed before operation, which cannot be continuously corrected in combination with real-time wind load changes, actual leg pressure distribution, and dynamic factors such as hoisting load swing, thereby reducing the timeliness and accuracy of the overturning risk warning.

[0003] In recent years, digital twin technology has attracted widespread attention in manufacturing and equipment operation and maintenance. It establishes a virtual model that evolves synchronously with the physical entity, realizes state visualization and fault prediction. However, there are still two major difficulties in directly applying digital twin to the crane field: first, the crane working conditions are greatly affected by the environment and operation instructions, with multiple data dimensions and rapid changes. There is still a lack of unified solution on how to build a stable and efficient multi-source data fusion mechanism. Second, although the digital twin model can present the overall state, it lacks risk priority evaluation and adaptive monitoring strategy for key positions, making it difficult to capture fault precursors and control sampling resources in a timely manner under large data conditions. Therefore, the present application proposes an intelligent crane operating environment construction method based on digital twin technology, which synchronizes the model and entity in real time, integrates historical fault causal relationships, real-time working condition parameters, and leg stress information, realizes dynamic generation of monitoring strategy and graded warning of overturning risk, and improves equipment safety and monitoring efficiency. SUMMARY

[0004] The present application constructs a digital twin model that synchronizes with the physical crane in real time, and dynamically adjusts the monitoring strategy in combination with the historical fault causal graph, so that the sensor sampling frequency and redundant channels can be automatically allocated according to the actual risk. Compared with the fixed sampling or manual setting scheme, this method can greatly reduce the redundant collection of low-risk parts while ensuring high-precision monitoring of key parts, reduce data transmission and storage costs, and improve the utilization efficiency of monitoring resources.

[0005] The intelligent crane operating environment construction method based on digital twin technology comprises:

[0006] Based on the three-dimensional structure information and configuration parameters of the crane, a digital twin model is constructed, and the posture, load and running state of the model are continuously synchronized with the physical crane by receiving sensor data in real time;

[0007] During the operation of the crane, meteorological data, working condition data and response data are collected, wherein: the meteorological data includes wind speed, ambient temperature and relative humidity; the working condition data includes hoisting load, working radius, boom elevation angle, rotation angular velocity and working mode; the response data includes main arm strain, hook-steel wire rope tension, slewing bearing torque, hydraulic system pressure and outrigger pressure;

[0008] Using the historical fault records of the crane, the association of key part failures with historical meteorological data and working condition data is analyzed, the key parts including the main arm, the hook assembly, the slewing bearing, the hydraulic system and the outrigger; then, according to the current meteorological data and working condition data, the key parts that need to be focused on are determined, and the corresponding monitoring strategy is formulated;

[0009] Based on the current meteorological data, working condition data and outrigger pressure, the stability of the crane is analyzed, the overturning risk is obtained and the overturning risk grading early warning is performed;

[0010] The obtained overturning risk level and monitoring strategy are mapped and synchronously displayed in the digital twin model in real time.

[0011] Preferably, the association of key part failures with historical meteorological data and working condition data is analyzed, and the specific operation is as follows:

[0012] Align the historical meteorological data, working condition data and fault event records according to a unified time reference, delete data entries with missing time stamps or missing fields, and reconstruct short missing sections using linear interpolation method, thereby obtaining a time series data set containing complete fields of wind speed, ambient temperature, relative humidity, hoisting load, working radius, boom elevation angle, rotation angular velocity and working mode;

[0013] Five training samples are established for main arm failure, hook assembly failure, slewing bearing failure, hydraulic system failure and outrigger failure, the meteorological data and working condition data are encoded into input vectors, and the occurrence of failure is encoded into output identifier, forming five groups of causal candidate pairs;

[0014] Subsequently, a conditional independence-based search algorithm is applied to each set of causal candidate pairs in turn, first a stepwise breadth-limited relationship screening method is used to exclude input fields that have no significant statistical association with the target fault variable, then a pair-wise conditional independence test is performed on the remaining input fields to obtain an initial causal direction set that satisfies the acyclic constraint; then, taking the initial causal direction set as the search space, a score-based incremental greedy optimization strategy is used to minimize the data reconstruction error while maintaining the stability of the causal graph structure, and finally a multiple-input single-output directed acyclic causal graph is generated for each target fault variable.

[0015] For each input edge in the directed acyclic causal graph, the conditional probability of the input edge is calculated using the joint frequency statistics of the historical samples, and the conditional probability is taken as the influence weight. After standardizing all influence weights, the influence strength of each input variable on the target fault is obtained.

[0016] The input variables are sorted in descending order of influence strength, and a group of input variables that rank high and are conditionally independent of each other are selected to form a fault root cause path. The fault root cause path and the corresponding influence strength are written into the monitoring data structure of the digital twin model for subsequent monitoring strategy module calling.

[0017] Preferably, according to the current meteorological data and the working condition data, the key parts that need to be focused on are determined, and the specific operation is as follows:

[0018] A monitoring period is set, and current meteorological data and current working condition data are received in each monitoring period. The meteorological data and the working condition data are taken as influence variables, and each influence variable is compared with its preset historical normal range to convert into a dimensionless real-time index to form an instant index set.

[0019] Based on the influence weight of each influence variable on the five target variables of main arm fault, hook assembly fault, slewing bearing fault, hydraulic system fault and leg fault; for each key part, the real-time index of the influence variable is multiplied by the influence weight to obtain a contribution value, and all contribution values are added to generate an original risk score of the key part. The original risk score is compared with the historical statistical upper and lower bounds to linearly map to a standardized risk score between zero and one. According to the preset risk threshold, the standardized risk score is divided into a high-risk monitoring level, a medium-risk monitoring level or a low-risk monitoring level.

[0020] Preferably, a corresponding monitoring strategy is formulated, which specifically includes:

[0021] For key positions in high-risk monitoring levels, the highest sampling frequency is set, dual-channel redundant acquisition is enabled, and instant alarm pushing is activated. For key positions in medium-risk monitoring levels, standard sampling frequency is maintained and periodic channel verification is performed. For key positions in low-risk monitoring levels, sparse sampling is adopted and estimation models are called to fill in non-key missing windows. Subsequently, the monitoring level and corresponding sampling parameters of each key position are written into the monitoring strategy list, and the current computing power and network bandwidth occupancy are compared. If the high-risk monitoring level persists for multiple monitoring periods or resources are scarce, the sampling frequency of the low-risk monitoring level is reduced in priority order to ensure that the real-time monitoring requirements of the high-risk monitoring level are met.

[0022] Preferably, based on current meteorological data, working condition data and outrigger pressure, crane stability analysis is performed to obtain overturning risk, and the specific operation is as follows:

[0023] In each monitoring period, the current hoisting load, working radius, boom angle and wind speed are read, and these input data are mapped to the unified coordinate system used by the digital twin model. Based on the input data, the stability moment is calculated, the weight, hoisting load moment and wind load moment are vector synthesized at the center of rotation of the crane to obtain the real-time stability moment. The outrigger pressure is obtained synchronously, and the overturning moment is calculated by combining it with the geometric distribution of the outrigger. The additional overturning moment caused by the imbalance of the outrigger pressure is superimposed into the load overturning moment to obtain the real-time overturning moment.

[0024] Based on the difference between the stability moment and the overturning moment, the safety margin is defined, and the sliding window algorithm is used to extrapolate the safety margin in the short term to generate a safety margin trend curve for future monitoring periods. Then, the current value and predicted value of the safety margin are compared with the preset threshold to divide them into three risk levels: normal zone, attention zone and danger zone.

[0025] Preferably, the synchronization process of the digital twin model includes:

[0026] The real-time collected attitude, load and running state data are extracted, the coordinate system is unified and the time tag is aligned. Then, recursive filtering is used to weaken random noise and occasional burrs. In each sampling period, the corrected data is written into the digital twin model to update the geometric node position and dynamics parameters, so that the digital twin model is always synchronized with the physical crane.

[0027] Preferably, the process of real-time mapping and synchronous display of the overturning risk level and monitoring strategy in the digital twin model includes:

[0028] The risk heat layer of color gradient is superimposed in the three-dimensional scene, and different color scales correspond to three levels of normal, attention and danger; then, the monitoring levels and sampling frequencies of the main arm, hook assembly, rotary support, hydraulic system and support leg are listed in the form of a list in real time, and the sampling parameters are allowed to be manually adjusted by the operator; finally, the risk level change, monitoring strategy adjustment and operator feedback are written into the monitoring data structure synchronously, which is used for subsequent auditing and model correction.

[0029] The present application has the following advantages:

[0030] 1、The present application constructs a digital twin model synchronized with the physical crane in real time, and dynamically adjusts the monitoring strategy combined with the historical fault causal diagram, so that the sensor sampling frequency and redundant channels can be automatically allocated according to the actual risk; compared with the fixed sampling or manual setting scheme, this method can greatly reduce the redundant collection of low-risk parts while ensuring high-precision monitoring of key parts, reduce data transmission and storage costs, and improve the utilization efficiency of monitoring resources.

[0031] 2、The present application introduces real-time support leg pressure, unbalanced additional torque and short-term wind load change in stability analysis, continuously calculates safety margin and performs multi-level overturning risk warning; compared with the traditional method which only relies on pre-operation static checking, this method can identify dynamic dangerous factors such as foundation settlement and gust impact in advance, issue timely and classified safety instructions to the control system and the operator, and significantly improve the safety and accident prevention ability of the crane operation. BRIEF DESCRIPTION OF DRAWINGS

[0032] Figure 1 The flowchart of the intelligent crane operation environment construction method based on digital twin technology adopted by the embodiment of the present application is shown. DETAILED DESCRIPTION

[0033] In order to make the personnel in the technical field better understand the technical solutions in the present application, the technical solutions in the embodiments of the present application will be described clearly and completely in combination with the drawings in the embodiments of the present application.

[0034] The embodiment of the intelligent crane operation environment construction method based on digital twin technology, as shown in Figure 1 , includes:

[0035] Based on the three-dimensional structure information and configuration parameters of the crane, a digital twin model is constructed, and through real-time reception of sensor data, the attitude, load and operating state of the model are continuously synchronized with the physical crane; The "sensor data" referred to herein refers to the original signals collected and returned by various measuring devices in real time during the operation of the crane, which are continuously driving the digital twin model to update its three-dimensional attitude, stress state and working environment after unified time stamping, coordinate conversion and filtering processing, thereby realizing high-precision synchronization of the virtual model and the physical crane.

[0036] During the operation of the crane, meteorological data, working condition data and response data are collected, wherein: the meteorological data includes wind speed, ambient temperature and relative humidity; the working condition data includes hoisting load, working radius, boom elevation angle, rotation angular velocity and working mode (hoisting, rotating, walking); the response data includes main arm strain, hook-steel wire rope tension, rotary support torque, hydraulic system pressure and leg pressure; the above three types of data collectively constitute the real-time input of the crane digital twin, the meteorological data reflects the direct influence of the external environment on the safety of the operation, the wind speed determines the size of the wind load, and the ambient temperature and relative humidity can indicate the trend of change in material performance or hydraulic efficiency; the working condition data describes the current operation mode and load state, the hoisting load, working radius, boom elevation angle and rotation angular velocity determine the stress distribution of the structure, and the working mode distinguishes the hoisting, rotating or walking working condition characteristics; the response data directly presents the health and stress of the key parts, the main arm strain, hook-steel wire rope tension, rotary support torque, hydraulic system pressure and leg pressure can be used to judge fatigue, overload, wear or uneven support; after the three types of data are comprehensively input into the digital twin model, the attitude and stress state of the crane can be reproduced in real time, providing complete basis for fault cause analysis, monitoring strategy dynamic adjustment and stable moment-overturning moment calculation, thereby improving the accuracy of safety warning and the utilization efficiency of monitoring resources;

[0037] The historical failure records of the crane are used to analyze the association between key part failures (main arm, hook assembly, slewing bearing, hydraulic system and outrigger) and historical meteorological and working condition data. Then, according to the current meteorological and working condition data, the key parts that need to be focused on are determined, and the corresponding monitoring strategy is formulated. The principle here is to convert "post-failure experience" into "pre-risk identification". By looking back at the history, the system quantifies the statistical association between environmental-working condition factors such as wind speed, working radius and boom elevation and key part failures such as main arm, hook assembly, slewing bearing, hydraulic system and outrigger, thereby establishing an influence weight diagram that can be used for real-time inference. When new meteorological and working condition data arrive, the system can instantly calculate the risk scores of each key part, automatically determine the part that is most likely to fail, and allocate monitoring resources (sampling frequency, redundant channels, alarm threshold) to these high-risk parts first, avoiding data redundancy and waste of computing power caused by "full monitoring", while improving the detection rate of potential failures and the early warning lead time.

[0038] Based on the current meteorological data, working condition data and outrigger pressure, the crane stability is analyzed to obtain the overturning risk and perform the overturning risk classification warning. This step uses real-time meteorological and working condition information combined with outrigger pressure to calculate the instantaneous stability moment and overturning moment of the crane, thereby obtaining the margin size and determining whether the device is in a safe, caution or dangerous state. By continuously updating the safety margin curve and setting multiple threshold values, the system can identify the overturning trend in advance in the case of sudden increase in wind speed, deviation of hoisting load or uneven force on outriggers, automatically issue speed limit, load reduction or emergency stop instructions to the control system, and intuitively present the risk level to the operator in a color classification manner, ensuring that the operating personnel have sufficient reaction time before an accident occurs, significantly reducing the probability of overturning accidents.

[0039] The obtained overturning risk level and monitoring strategy are mapped and synchronously displayed in the digital twin model in real time, enabling the operator to intuitively view the current risk state and monitoring focus of the crane.

[0040] The association between key part failures and historical meteorological and working condition data is analyzed as follows:

[0041] First, align the historical meteorological and working condition data with the failure event records according to a unified time reference, delete data entries with missing timestamps or missing fields, and use linear interpolation to reconstruct short missing sections, thereby obtaining a time series data set containing complete fields such as wind speed, environmental temperature, relative humidity, hoisting load, working radius, boom elevation, slewing angular velocity and working mode.

[0042] Five training samples are respectively established with five result variables of main arm failure, hook assembly failure, slewing bearing failure, hydraulic system failure and outrigger failure, weather data and working condition data are coded as input vectors, and whether failure occurs or not is coded as output identifier to form five groups of causal candidate pairs;

[0043] Subsequently, a search algorithm based on conditional independence is applied to each group of causal candidate pairs in turn. Firstly, a stepwise width-limited relationship screening method is used to exclude input fields that have no significant statistical association with the target failure variable. Then, a pair-by-pair conditional independence test is performed on the remaining input fields to obtain an initial causal direction set that satisfies the acyclic constraint. Then, taking the initial causal direction set as the search space, an incremental greedy optimization strategy based on scoring is used to minimize the data reconstruction error while maintaining the stability of the causal graph structure, and finally a multiple-input single-output directed acyclic causal graph for each target failure variable is generated.

[0044] For each input edge in the directed acyclic causal graph, the conditional probability of the input edge is calculated using the joint frequency statistics of the historical samples, and the conditional probability is taken as the influence weight. After standardizing all influence weights, the influence strength of each input variable on the target failure is obtained.

[0045] The input variables are sorted in descending order of influence strength, and a group of input variables that rank in the front and are conditionally independent of each other are selected to form a fault root cause path. The fault root cause path and the corresponding influence strength are written into the monitoring data structure of the digital twin model for subsequent calling by the monitoring strategy module.

[0046] According to the current weather data and working condition data, the key parts that need to be focused on are determined, and the specific operation is as follows:

[0047] Firstly, the current weather data and the current working condition data are received in each monitoring period. The weather data and the working condition data are taken as influence variables, and each influence variable is compared with its preset historical normal range to convert into a dimensionless real-time index to form an instant index set.

[0048] Based on the influence weight of each influence variable on the five target variables of main arm failure, hook assembly failure, slewing bearing failure, hydraulic system failure and outrigger failure. For each key part, the real-time index of the influence variable is multiplied by the influence weight to obtain a contribution value, and all contribution values are added to generate an original risk score of the key part. The original risk score is compared with the historical statistical upper and lower bounds to linearly map into a standardized risk score between zero and one. According to the preset risk threshold, the standardized risk score is divided into a high-risk monitoring level, a medium-risk monitoring level or a low-risk monitoring level.

[0049] A corresponding monitoring strategy is formulated, specifically including:

[0050] For key positions in the high-risk monitoring level, the highest sampling frequency is set, dual-channel redundant acquisition is enabled, and instant alarm pushing is activated. For key positions in the medium-risk monitoring level, the standard sampling frequency is maintained and periodic channel verification is performed. For key positions in the low-risk monitoring level, sparse sampling is adopted and an estimation model is called to fill in non-key missing windows. Subsequently, the monitoring level and corresponding sampling parameters of each key position are written into a monitoring strategy list, and the current computing power and network bandwidth occupancy are compared. If the high-risk monitoring level persists for multiple monitoring periods or resources are scarce, the sampling frequency of the low-risk monitoring level is reduced in priority order to ensure that the real-time monitoring requirements of the high-risk monitoring level are met. Finally, the monitoring strategy list is synchronized to the monitoring data structure of the digital twin model, and the monitoring level state of the key position is marked in real time in the visualization interface in the form of color or icon.

[0051] Based on the current meteorological data, working condition data, and leg pressure, the stability of the crane is analyzed to obtain the overturning risk and perform the overturning risk grading warning. The specific operations are as follows:

[0052] In each monitoring period, the current hoisting load, working radius, boom angle, and wind speed are read as input data, and these input data are mapped to the unified coordinate system used by the digital twin model. Based on the input data, the stability moment is calculated, and the weight, hoisting load moment, and wind load moment are vector synthesized at the center of rotation of the crane to obtain the real-time stability moment. The leg pressure is obtained synchronously, and the overturning moment is calculated by combining the leg pressure with the geometric distribution of the legs. The additional overturning moment caused by the imbalance of the leg pressure is superimposed into the load overturning moment to obtain the real-time overturning moment.

[0053] Based on the difference between the stability moment and the overturning moment, the safety margin is defined, and the sliding window algorithm is used to extrapolate the safety margin for a short time to generate a safety margin trend curve for future monitoring periods. Subsequently, the current value and predicted value of the safety margin are compared with the preset threshold to divide them into three risk levels: normal zone, attention zone, and danger zone.

[0054] The overturning risk grading warning specifically includes:

[0055] When the safety margin is in the normal zone and the trend is stable, only the existing monitoring strategy is maintained; when the safety margin enters the attention zone or shows a continuous downward trend, the swing angle speed is automatically reduced and the key part associated with the hazard source is monitored to the high level; when the safety margin falls into the danger zone or is predicted to fall into the danger zone within the preset time window, the load limiting, arm retracting or emergency stopping instructions are immediately sent to the crane control system, and the risk source position and safety suggestions are highlighted in the digital twin model scene; finally, the risk level change, automatic control instructions and operator confirmation information are written into the monitoring data structure for subsequent audit, model correction and trend analysis, and the above process is repeated in the next monitoring cycle to realize the continuous assessment and graded early warning of the overturning risk.

[0056] The synchronization process of the digital twin model includes:

[0057] The real-time data such as attitude angle, load weight, lifting height, swing angle speed and leg pressure are supplemented with time stamps and converted to the coordinate system consistent with the three-dimensional structure model;

[0058] Recursive low-pass filters are used to remove high-frequency noise, and three-point sliding median method is used to remove occasional burrs, and linear interpolation repair is performed on the short-time packet sampling segment;

[0059] In each sampling period, the filtered and corrected data are written into the digital twin data buffer, the three-dimensional coordinates of the main arm node, the swing center and the lifting point are updated, and the load, wind load and inertia term in the dynamics parameter table are reset;

[0060] If there are missing data or value overruns for consecutive multiple periods, the corresponding node update is suspended and a data integrity warning is triggered;

[0061] After synchronization, the time stamp, data integrity identifier and update result are output to ensure the real-time consistency of the digital twin model and the physical crane in position, attitude and stress state.

[0062] The process of real-time mapping and synchronization of overturning risk level and monitoring strategy in the digital twin model includes:

[0063] In the three-dimensional scene of the digital twin model, a risk heat layer is superimposed on the geometric bodies such as the lifting arm, lifting hook and leg, and green, yellow and red gradient color scales are used to represent normal, attention and danger risk levels respectively, and the heat layer transparency increases with the risk level;

[0064] A monitoring strategy panel is generated in the scene sidebar, which lists the monitoring level, current sampling frequency and redundant channel state of the main arm, lifting hook assembly, swing bearing, hydraulic system and leg in table form in real time, and flashes or highlights when the monitoring level changes;

[0065] Providing the interactive control allows the operator to manually increase or decrease the sampling frequency for any critical site, the increase operation takes effect immediately, and the decrease operation needs to be verified by the system that the current risk level is not higher than the attention level before it can be confirmed;

[0066] The risk level change, monitoring strategy adjustment, manual intervention record and operation confirmation information are written in the monitoring data structure in chronological order to form a complete audit chain;

[0067] When the system detects that the risk level falls back and remains stable, the default sampling frequency is automatically restored and a "restored" mark is displayed on the monitoring strategy panel, so that the operator can quickly identify the current state.

[0068] It should be understood that, for those skilled in the art, improvements or changes can be made according to the above description, and all these improvements and changes shall belong to the protection scope of the appended claims of the present application. The parts not described in detail in the specification belong to the prior art known to those skilled in the art.

Claims

1. A method for constructing an intelligent crane operating environment based on digital twinning technology, characterized in that, The method comprises the following steps: Based on the three-dimensional structure information and configuration parameters of the crane, a digital twin model is constructed, and the model is kept synchronized with the physical crane in terms of posture, load and operating state by receiving sensor data in real time; During the operation of the crane, meteorological data, working condition data and response data are collected, wherein the meteorological data includes wind speed, ambient temperature and relative humidity; the working condition data includes hoisting load, working radius, boom elevation angle, rotation angular velocity and working mode; the response data includes main arm strain, hook-steel wire rope tension, slewing bearing torque, hydraulic system pressure and leg pressure; Using the historical fault records of the crane, the correlation between key part failures and historical meteorological data and working condition data is analyzed, the key parts including the main arm, hook assembly, slewing bearing, hydraulic system and legs; then, according to the current meteorological data and working condition data, the key parts that need to be focused on are determined, and the corresponding monitoring strategy is formulated; Based on the current meteorological data, working condition data and leg pressure, the stability of the crane is analyzed to obtain the overturning risk and perform the overturning risk classification warning; The obtained overturning risk level and monitoring strategy are mapped and synchronously displayed in the digital twin model in real time; Based on the current meteorological data, working condition data and leg pressure, the stability of the crane is analyzed to obtain the overturning risk, which is specifically operated as follows: In each monitoring period, the current hoisting load, working radius, boom elevation angle and wind speed are read as input data, and these input data are mapped to a unified coordinate system used by the digital twin model; based on the input data, the stable moment is calculated, the weight, hoisting load moment and wind load moment are vector synthesized at the rotation center of the crane to obtain the real-time stable moment; the leg pressure is synchronously obtained, the overturning moment is calculated by combining the leg pressure with the geometric distribution of the legs, and the additional overturning moment caused by the imbalance of the leg pressure is superimposed into the load overturning moment to obtain the real-time overturning moment; Based on the difference between the stable moment and the overturning moment, the safety margin is defined, and the sliding window algorithm is used to extrapolate the safety margin in the short term to generate the safety margin trend curve for future monitoring periods; then, the current value and the predicted value of the safety margin are compared with the preset threshold to divide the safety margin into three risk levels: normal zone, attention zone and danger zone. 2.The smart crane operating environment construction method based on digital twin technology according to claim 1, characterized in that, The correlation between key part failures and historical meteorological data and working condition data is analyzed, which is specifically operated as follows: The historical meteorological data, working condition data and fault event records are aligned according to a unified time reference, data entries with missing time stamps or missing fields are deleted, and linear interpolation is used to reconstruct the short missing section, thereby obtaining a time series data set containing complete fields of wind speed, ambient temperature, relative humidity, hoisting load, working radius, boom elevation angle, rotation angular velocity and working mode; Five result variables are established, including main arm failure, hook assembly failure, slewing bearing failure, hydraulic system failure and leg failure, five groups of causal candidate pairs are constructed by encoding the meteorological data and working condition data into input vectors and encoding the occurrence of failure into output identifiers. Subsequently, a conditional independence-based search algorithm is applied to each group of causal candidate pairs, first a stepwise breadth-limited relationship screening method is used to exclude input fields that have no significant statistical association with the target fault variable, then a pair-by-pair conditional independence test is performed on the remaining input fields to obtain an initial causal direction set that satisfies the acyclic constraint; then, taking the initial causal direction set as the search space, a score-based incremental greedy optimization strategy is used to minimize the data reconstruction error while maintaining the stability of the causal graph structure, and finally a multiple-input single-output directed acyclic causal graph is generated for each target fault variable; For each input edge in the directed acyclic causal graph, the conditional probability of the input edge is calculated using the joint frequency statistics of the historical samples, and the conditional probability is used as the influence weight. After standardizing all influence weights, the influence strength of each input variable on the target fault is obtained; The input variables are sorted in descending order of influence strength, and a group of input variables with high ranking and conditional independence with each other are selected to form a fault root cause path. The fault root cause path and the corresponding influence strength are written into the monitoring data structure of the digital twin model for subsequent monitoring strategy module calling. 3.The smart crane operating environment construction method based on digital twin technology according to claim 2, characterized in that, According to the current meteorological data and working condition data, the key parts that need to be focused on are determined, and the specific operations are as follows: Set the monitoring period, receive the current meteorological data and the current working condition data in each monitoring period; take the meteorological data and the working condition data as influence variables, compare each influence variable with its pre-set historical normal range, convert it into a dimensionless real-time index, and form an instant index set; Based on the influence weight of each influence variable on the five target variables of main arm fault, hook assembly fault, slewing bearing fault, hydraulic system fault and leg fault; for each key part, traverse its corresponding influence variable, multiply the real-time index of the influence variable by the influence weight to obtain the contribution value, and then add all the contribution values to generate the original risk score of the key part; compare the original risk score with the historical statistical upper and lower bounds, and linearly map it to a standardized risk score between zero and one; According to the pre-set risk threshold, the standardized risk score is divided into a high-risk monitoring level, a medium-risk monitoring level or a low-risk monitoring level.

4. The intelligent crane operating environment construction method based on digital twin technology according to claim 3, characterized in that, Formulate the corresponding monitoring strategy, which specifically includes: For the key parts in the high-risk monitoring level, set the highest sampling frequency, enable dual-channel redundant acquisition and instant alarm push, maintain the standard sampling frequency for the key parts in the medium-risk monitoring level and perform periodic channel verification, and use sparse sampling for the key parts in the low-risk monitoring level and call the estimation model to fill in the non-key missing window; then, write the monitoring level and the corresponding sampling parameters of each key part into the monitoring strategy list, and compare the current computing power with the network bandwidth occupancy rate; if the high-risk monitoring level lasts for multiple monitoring periods or resources are scarce, reduce the sampling frequency of the low-risk monitoring level in priority order to ensure that the real-time monitoring requirements of the high-risk monitoring level are met.

5. The intelligent crane operating environment construction method based on digital twin technology according to claim 4, characterized in that, The synchronization process of the digital twin model includes: The real-time collected attitude, load and running state data are extracted, the coordinate system is unified and the time tag is aligned. Then the recursive filtering is used to weaken the random noise and accidental burr. In each sampling period, the corrected data is written into the digital twin model, the geometric node position and the dynamics parameters are updated, so that the digital twin model is always synchronized with the physical crane.

6. The intelligent crane operating environment construction method based on digital twin technology according to claim 5, characterized in that, The process of real-time mapping and synchronous display of the overturning risk level and monitoring strategy in the digital twin model includes: The risk heat layer with color gradient is superimposed in the three-dimensional scene, and different color scales correspond to three levels of normal, attention and danger. Then the monitoring level and sampling frequency of the main arm, hook assembly, slewing bearing, hydraulic system and outrigger are listed in the form of list in real time, and the operation personnel are allowed to manually adjust the sampling parameters. Finally, the risk level change, monitoring strategy adjustment and operator feedback are written into the monitoring data structure synchronously for subsequent audit and model correction.

Images