Digital twinborn model construction method, life prediction method, model construction device and equipment for crane

Abstract

The invention discloses a digital twinborn model construction method, a life prediction method, a model construction device and equipment of a crane, and relates to the technical field of industrial Internet of Things, the method comprises the following steps: constructing a multi-dimensional digital twinborn model, the multi-dimensional digital twinborn model comprises a physical entity layer, a connection layer, a virtual model layer, a twinborn data layer and a service function layer, the virtual model layer comprises a target twin model, and the service function layer corrects key components of the crane; transmitting the physical parameters of the crane to a connection layer for preprocessing to obtain a physical data set; the physical data set is transmitted to a virtual model layer, and an initial twin model is obtained through kinetic analysis and knowledge graph processing; and transmitting the virtual data set to a twin data layer, calculating target correction parameters according to the physical data set and the virtual data set, and correcting model parameters of the initial twin model to obtain a target twin model. According to the method, the problem of insufficient physical mapping precision of the existing virtual model is solved.

Description

Technical Field

[0001] This application relates to the field of industrial Internet of Things (IoT) technology, and in particular to a method for constructing a digital twin model of a crane, a life prediction method, a model construction device, and equipment. Background Technology

[0002] With the deepening of industrial modernization, bridge cranes are widely used as key production equipment in workshops, warehouses, and ports. However, the aging of their main drive systems, slings, and other core components, coupled with frequent use, has led to an increasing number of failures. Traditional monitoring methods (relying on periodic manual inspections and static assessments based on historical data) are insufficient for real-time perception of structural stress states under dynamic operating conditions, resulting in frequent safety accidents.

[0003] Currently, although simulation technology can be used to build virtual crane models as a solution, the physical mapping accuracy of existing models is insufficient, making it difficult to meet the needs of high-precision monitoring.

[0004] Therefore, improving the mapping accuracy of virtual models of bridge cranes to physical entities is an urgent problem to be solved. Summary of the Invention

[0005] The main objective of this application is to provide a method for constructing a digital twin model of a crane, a life prediction method, a model construction device and equipment, which aims to solve the technical problem of insufficient physical mapping accuracy of existing models.

[0006] To achieve the above objectives, this application proposes a method for modeling a crane, the method comprising:

[0007] A multidimensional digital twin model is constructed, comprising at least a physical entity layer, a connection layer, a virtual model layer, a twin data layer, and a service function layer. The physical entity layer is used to collect the current operating parameters of the crane. The virtual model layer includes a target twin model, which is used to map the physical entity of the crane into a virtual space. The twin data layer is used to manage the physical dataset and the virtual dataset. The service function layer is used to correct key components of the crane.

[0008] The current operating parameters of the crane are transmitted to the connection layer for preprocessing to obtain the physical dataset. The preprocessing includes protocol conversion, interface standardization, and encryption.

[0009] The physical dataset is transmitted to the virtual model layer, and an initial twin model is obtained through dynamic analysis and knowledge graph processing. The initial twin model is used to output the virtual dataset.

[0010] The virtual dataset is transmitted to the twin data layer, and the target correction parameters are calculated based on the physical dataset and the virtual dataset.

[0011] Based on the target correction parameters, the model parameters of the initial twin model are corrected to obtain the target twin model.

[0012] In one embodiment, the virtual simulation layer includes a multi-level geometric model, a multi-dimensional physical model, and a behavioral rule model. The physical dataset includes physical geometric data, material data, load data, environmental data, and preset rules. The step of transmitting the physical dataset to the virtual model layer and obtaining an initial twin model through dynamic analysis and knowledge graph processing, wherein the initial twin model is used to output the virtual dataset, includes:

[0013] The physical geometry data and material data are transmitted to the multi-level geometric model to obtain design information, which includes part-level design information and component-level design information.

[0014] The load data and environmental data are transmitted to the multidimensional physical model, and a quantitative mapping relationship is obtained through dynamic analysis. The quantitative mapping relationship is the quantitative mapping relationship between load, internal force and stress. The multidimensional physical model includes a structural mechanics model and a temperature field model.

[0015] The preset rules are transmitted to the behavior rule model, and the preset rules are integrated through a knowledge graph to obtain a special rule knowledge graph;

[0016] Based on the design information, the quantization mapping relationship, and the special rule knowledge graph, an initial twin model is obtained, which is used to output the virtual dataset.

[0017] In one embodiment, the step of transmitting the virtual dataset to the twin data layer and calculating the target correction parameters based on the physical dataset and the virtual dataset includes:

[0018] Identify the parameter deviations between the physical dataset and the virtual dataset to obtain the deviation dataset;

[0019] The similarity between real-time sensor data and historical interference data is calculated using principal component analysis to obtain similarity results.

[0020] Based on the similarity results, the target correction parameters are obtained by filtering from the deviation dataset. In one embodiment, the step of calculating the similarity between real-time sensor data and historical interference data using principal component analysis to obtain the similarity results includes:

[0021] The sampling data matrix and the interference data matrix are determined based on real-time sensor data and historical interference data.

[0022] The sampled data matrix and the interference data matrix are standardized to obtain a standardized real-time matrix and a standardized historical matrix.

[0023] Principal component analysis is performed on the standardized real-time matrix and the standardized historical matrix to extract the first K principal component vectors.

[0024] Based on the first K principal component vectors, the time series data of the first principal component vector is obtained;

[0025] Based on the time-series data of the first principal component vector, the minimum cumulative distance between the standardized real-time matrix and each of the standardized historical matrices is calculated using a dynamic time-normalization algorithm to determine the similarity result.

[0026] In one embodiment, the sampling data matrix is ​​represented as:

[0027]

[0028] Where S(t) represents the sensor from x1 to x K At time [t] s ,t w The collected data matrix within ], x1(t s ) indicates that the first sensor is at time t s The sampled data, x K (t s ) represents the Kth sensor at time t s The sampled data, x1(t) w ) indicates that the first sensor is at time t w The sampled data, x K (t w ) represents the Kth sensor at time t w The sampling data;

[0029] The interference data matrix is ​​represented as follows:

[0030]

[0031] Among them, S j (t) is the j-th type of interference data matrix, x1(t) jc The first sensor operates for the time t during which the interference factor j takes effect. jc The collected interference data, x1(t) jd The first sensor operates for the time t during which the interference factor j takes effect. jd The collected interference data, x K (t jcThe time t for the Kth sensor to be affected by interference factor j jc The collected interference data, x K (t jd The time t for the Kth sensor to be affected by interference factor j jd The collected interference data.

[0032] In one embodiment, the step of correcting the model parameters of the initial Siamese model based on the target correction parameters to obtain the target Siamese model includes:

[0033] Construct a correlation model between the target correction parameters and the target variable, and obtain the output value of the correlation model;

[0034] The output value of the associated model is input into the initial twin model for dynamic mapping and correction of model parameters to obtain the target twin model.

[0035] Furthermore, to achieve the above objectives, this application also proposes a method for predicting the lifespan of a crane, the method comprising:

[0036] Based on the model output value of the target twin model, the twin data is determined. The target twin model is constructed using the crane model construction method described above.

[0037] The fatigue life state parameters of key components are extracted by fusing the physical dataset and the twin data through the first neural network.

[0038] The fatigue life state parameters are input into a second neural network for time series degradation analysis to obtain load spectrum statistics.

[0039] Based on the fatigue damage accumulation theory and the statistical results of the load spectrum, the predicted remaining life of the key components is output.

[0040] In one embodiment, the step of extracting fatigue life state parameters of key components by fusing the physical dataset and twin data through a first neural network includes:

[0041] The location of fatigue-prone cracks was determined based on finite element analysis.

[0042] Convert the fatigue life curve of the material into a fatigue life curve under actual working conditions.

[0043] Based on the location of the fatigue-prone crack and the fatigue life curve under the actual working conditions, real-time load data and stress distribution data are determined.

[0044] The real-time load data and stress distribution data are input into a convolutional neural network for feature fusion, and the fatigue life state parameters of key components are obtained through convolutional kernel operations.

[0045] Furthermore, to achieve the above objectives, this application also proposes a crane model building apparatus, which includes:

[0046] The model building module is used to construct a multidimensional digital twin model. The multidimensional digital twin model includes at least a physical entity layer, a connection layer, a virtual model layer, a twin data layer, and a service function layer. The physical entity layer is used to collect the current operating parameters of the crane. The virtual model layer includes a target twin model, which is used to map the physical entity of the crane into a virtual space. The twin data layer is used to manage the physical dataset and the virtual dataset. The service function layer is used to correct key components of the crane.

[0047] The data preprocessing module is used to transmit the current operating parameters of the crane to the connection layer for preprocessing to obtain the physical dataset. The preprocessing includes protocol conversion, interface standardization, and encryption.

[0048] An initial twin module is used to transmit the physical dataset to the virtual model layer, obtain an initial twin model through dynamic analysis and knowledge graph processing, and the initial twin model is used to output the virtual dataset;

[0049] The parameter acquisition module allows the user to transmit the virtual dataset to the twin data layer and calculate the target correction parameters based on the physical dataset and the virtual dataset.

[0050] The model correction module is used to correct the model parameters of the initial twin model based on the target correction parameters to obtain the target twin model.

[0051] In addition, to achieve the above objectives, this application also proposes a crane model building device, the device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the crane model building method or the crane life prediction method as described above.

[0052] One or more technical solutions proposed in this application have at least the following technical effects:

[0053] A multi-dimensional digital twin model is constructed, comprising at least a physical entity layer, a connection layer, a virtual model layer, a twin data layer, and a service function layer. The physical entity layer collects the crane's current operating parameters, providing a real-time data source for the model. The virtual model layer includes a target twin model, which maps the crane's physical entity to the virtual space, establishing a mapping relationship between the physical and virtual models. The twin data layer manages the physical and virtual datasets, while the service function layer corrects key crane components, avoiding data silos and functional gaps. This layered structure (physical entity layer, connection layer, virtual model layer, twin data layer, and service function layer) achieves multi-dimensional integration of the crane model, ensuring clear responsibilities for data acquisition, transmission, modeling, and data management, and providing systematic support for high-precision mapping. The crane's current operating parameters are transmitted to the connection layer for preprocessing to obtain the physical dataset. Preprocessing includes protocol conversion, interface standardization, and encryption, providing highly reliable and high-concurrency data support for the service function layer and the twin data layer. The physical dataset is transmitted to the virtual model layer. An initial twin model is obtained through dynamic analysis and knowledge graph processing. Dynamic analysis ensures the virtual model conforms to physical laws, while the knowledge graph integrates domain knowledge, giving the initial twin model higher physical mapping accuracy. Furthermore, the initial twin model also outputs a more accurate virtual dataset. The virtual dataset is then transmitted to the twin data layer. Target correction parameters are calculated based on the physical and virtual datasets. Quantifying the differences between the physical and virtual model data provides a data basis for high-precision correction. Based on the target correction parameters, the model parameters of the initial twin model are corrected to obtain the target twin model. By dynamically adjusting the model parameters, the virtual model continuously approximates real physical behavior, forming a high-fidelity mapping. A closed-loop correction mechanism ensures continuous model optimization, ultimately improving mapping accuracy. Attached Figure Description

[0054] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0055] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0056] Figure 1 This is a flowchart illustrating the first embodiment of the crane model construction method of this application;

[0057] Figure 2 This is a schematic diagram of the overall architecture of the multidimensional digital twin model in the embodiments of this application;

[0058] Figure 3 This is a flowchart illustrating the second embodiment of the crane model construction method of this application;

[0059] Figure 4 This is a simplified mechanical model diagram of the main beam in the vertical direction according to an embodiment of this application;

[0060] Figure 5 This is a flowchart illustrating the third embodiment of the crane model construction method of this application;

[0061] Figure 6 This is a schematic diagram of the model modification process in an embodiment of this application;

[0062] Figure 7 This is a flowchart illustrating the fourth embodiment of the crane model construction method of this application;

[0063] Figure 8 This is a schematic diagram of the modular structure of the crane model building device according to an embodiment of this application;

[0064] Figure 9 This is a schematic diagram of the equipment structure of the hardware operating environment involved in the crane model construction method in this application embodiment.

[0065] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0066] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0067] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0068] Due to their large size, complex structure, and frequent start-stop operations, bridge cranes face challenges in monitoring, including limited inspection coverage, low model accuracy, and high risk of fatigue failure. Traditional monitoring methods (relying on periodic manual inspections and static assessments based on historical data) cannot achieve real-time perception of structural stress states under dynamic operating conditions, making it difficult to provide timely warnings of potential safety hazards. Currently, while simulation technology can be used to construct virtual crane models as a solution, the physical mapping accuracy of existing models is insufficient to meet the requirements of high-precision monitoring. Therefore, improving the mapping accuracy of virtual bridge crane models to physical entities is an urgent problem to be solved.

[0069] Based on this, embodiments of this application provide a method for constructing a digital twin model of a crane, referring to... Figure 1 , Figure 1This is a flowchart illustrating the first embodiment of the crane model construction method of this application. Due to the large size, complex structure, and frequent start-stop characteristics of bridge cranes, this application will be described in detail using a bridge crane as an example.

[0070] In this embodiment, the crane model construction method includes steps S10 to S50:

[0071] Step S10: Construct a multidimensional digital twin model. The multidimensional digital twin model includes at least a physical entity layer, a connection layer, a virtual model layer, a twin data layer, and a service function layer. The physical entity layer is used to collect the current operating parameters of the crane. The virtual model layer includes a target twin model, which is used to map the physical entity of the crane into the virtual space. The twin data layer is used to manage the physical dataset and the virtual dataset. The service function layer is used to correct the key components of the crane.

[0072] It should be noted that the overall architecture of the multidimensional digital twin model can be referenced. Figure 2 Specifically, the physical entity layer can be understood as the real-world foundation of the five-dimensional digital twin model of the bridge crane. For example, the physical entity layer can be divided into the bridge crane body, the sensor network, and the data acquisition and transmission system. The bridge crane body includes the main beam, end beams, trolley traveling mechanism, gantry traveling mechanism, and hoisting mechanism. The main beam is the core load-bearing structure supporting the lateral movement of the large vehicle traveling mechanism. The end beams connect the main beam to the rails; the trolley traveling mechanism drives the crane forward along the rails, and the gantry traveling mechanism controls the movement along the direction of the main beam. The sensor network consists of multiple sensors; for example, the sensor network may include, but is not limited to, non-contact current transformers (measuring real-time motor current), temperature sensors (motor temperature), displacement sensors (traveling speed of the trolley and gantry), strain sensors (strain of key welds on the main beam and end beams), and acceleration sensors (traveling acceleration of the trolley and gantry). In physical space, industrial sensor technology is used to collect and transmit real-time dynamic parameters of the crane across all work positions and elements (such as motor current and temperature of the hoisting mechanism, deformation of crane structural components, and running speed and acceleration of the trolley and crane traveling mechanisms). The data acquisition and transmission system can acquire heterogeneous sensor signals through embedded multi-channel acquisition modules and process the acquired data (such as vibration waveform noise reduction and stress feature extraction). A hybrid communication architecture is adopted to achieve real-time synchronous acquisition and low-latency transmission of multi-source data.

[0073] It should be noted that the virtual model layer, through multi-dimensional modeling and high-fidelity simulation technology, achieves accurate mapping and dynamic simulation of physical entities in virtual space, supporting the full lifecycle management needs of bridge cranes, including service status monitoring, dynamic fault prediction, and service life prediction. The service function layer can be understood as the application outlet of the bridge crane's multi-dimensional digital twin model, realizing intelligent services throughout the equipment's lifecycle. This layer mainly includes two parts: functional services (for developers) and business services. Business services include equipment status monitoring, lifespan prediction, and maintenance decision-making.

[0074] Step S20: Transmit the current operating parameters of the crane to the connection layer for preprocessing to obtain a physical dataset. The preprocessing includes protocol conversion, interface standardization, and encryption.

[0075] It should be noted that the physical dataset can be understood as a structured set of runtime parameters preprocessed by the connection layer. The connection layer is responsible for efficient communication and collaborative interaction between physical devices and virtual models, and between lower-level data and upper-level services. The core functions of the connection layer include multi-source data acquisition, protocol conversion, interface standardization, and encryption processing, ensuring communication protocol compatibility and network security.

[0076] Step S30: The physical dataset is transmitted to the virtual model layer, and an initial twin model is obtained through dynamic analysis and knowledge graph processing. The initial twin model is used to output the virtual dataset.

[0077] It should be noted that dynamic analysis is the computational process of constructing the kinematics or dynamics equations of a crane based on Newtonian mechanics and multibody system theory. Knowledge graph processing can be understood as a technique that uses graph structures to integrate knowledge in the field of crane machinery to assist in model construction. The initial twin model can be understood as a primary virtual mapping generated based on the physical dataset and prior knowledge. The virtual dataset corresponds to the initial twin model and can be understood as the equivalent physical parameters generated by simulation. The virtual dataset can maintain the same dimension and sampling rate as the physical dataset.

[0078] Step S40: Transfer the virtual dataset to the twin data layer and calculate the target correction parameters based on the physical dataset and the virtual dataset.

[0079] It should be noted that the twin data layer is used to manage physical datasets, virtual datasets, and service-related data, providing data support for other layers of the multidimensional digital twin model. The physical dataset primarily reflects the static attributes and dynamic process data of the bridge crane's physical entity. For example, the physical dataset may include, but is not limited to, geometric dimensions, material properties, load distribution, track length, wheel specifications, and real-time monitoring data such as operating status (e.g., lifting, translation, braking), vibration frequency, temperature changes, and lifting load. The virtual dataset mainly reflects the simulation data of the virtual model, such as the three-dimensional geometric dimensions, assembly relationships, and positional information of components like the main beam, trolley, and hook; structural mechanics and thermodynamics simulation data such as material properties, stress distribution, and temperature field distribution; and dynamic behavior data and constraint conditions, logical rule data such as correlations and fault propagation paths, specifically driving factors, environmental disturbances, and operating mechanisms. Service-related data includes service status monitoring data, fault diagnosis algorithms, dynamic life prediction models, and other business-oriented data, as well as function-oriented data such as model inconsistency correction during the application of the bridge crane digital twin model. The target correction parameters can characterize the set of compensation amounts for the initial twin model error.

[0080] Step S50: Based on the target correction parameters, correct the model parameters of the initial twin model to obtain the target twin model.

[0081] It should be noted that the internal variables of the initial twin model are dynamically adjusted according to the target correction parameters. After parameter correction, the target twin model is obtained, which significantly improves the mapping accuracy of the physical entity. Understandably, since the crane's current operating parameters are constantly updated, the physical dataset is also constantly updated. Through continuous correction, the final target twin model infinitely approximates the state of the physical entity.

[0082] In this embodiment, a multi-dimensional digital twin model is constructed, which includes at least a physical entity layer, a connection layer, a virtual model layer, a twin data layer, and a service function layer. The physical entity layer collects the crane's current operating parameters, providing a real-time data source for the model. The virtual model layer includes a target twin model, which maps the crane's physical entity to the virtual space, establishing a mapping relationship between the physical and virtual models. The twin data layer manages the physical and virtual datasets, and the service function layer corrects key crane components, avoiding data silos and functional gaps. This layered structure (physical entity layer, connection layer, virtual model layer, twin data layer, and service function layer) achieves multi-dimensional integration of the crane model, ensuring clear responsibilities for data collection, transmission, modeling, and data management, and providing systematic support for high-precision mapping. The crane's current operating parameters are transmitted to the connection layer for preprocessing to obtain the physical dataset. Preprocessing includes protocol conversion, interface standardization, and encryption, providing highly reliable and high-concurrency data support for the service function layer and the twin data layer. The physical dataset is transmitted to the virtual model layer. An initial twin model is obtained through dynamic analysis and knowledge graph processing. Dynamic analysis ensures the virtual model conforms to physical laws, while the knowledge graph integrates domain knowledge, giving the initial twin model higher physical mapping accuracy. Furthermore, the initial twin model also outputs a more accurate virtual dataset. The virtual dataset is then transmitted to the twin data layer. Target correction parameters are calculated based on the physical and virtual datasets. Quantifying the differences between the physical and virtual model data provides a data basis for high-precision correction. Based on the target correction parameters, the model parameters of the initial twin model are corrected to obtain the target twin model. By dynamically adjusting the model parameters, the virtual model continuously approximates real physical behavior, forming a high-fidelity mapping. A closed-loop correction mechanism ensures continuous model optimization, ultimately improving mapping accuracy.

[0083] Reference Figure 3 , Figure 3 This is a flowchart illustrating the second embodiment of the crane model construction method of this application, based on the above. Figure 1 The first embodiment shown presents a second embodiment of the crane model construction method of this application.

[0084] It should be noted that the virtual simulation layer includes a multi-level geometric model, a multi-dimensional physical model, and a behavioral rule model. The physical dataset includes physical geometric data, material data, load data, environmental data, and preset rules. Based on this, in the second embodiment, step S30 includes:

[0085] Step S301: Transfer the physical geometry data and material data to the multi-level geometric model to obtain design information, including part-level design information and component-level design information.

[0086] For example, a three-dimensional multi-level geometric model is constructed for a bridge crane, including part-level and component-level models. The part-level model includes the geometric features and material properties of individual components, supporting rapid iteration and variant design during the design phase, and ensuring the accurate representation of each component in virtual space. The component-level model focuses on the constraint relationships and kinematic properties of subassemblies, such as defining the contact relationship between the track and wheels, and key relationships like the kinematic coupling of wire ropes and pulleys to ensure consistency between the virtual model and the physical entity in mechanical behavior.

[0087] Step S302: Transfer the load data and environmental data to the multidimensional physical model, and obtain the quantitative mapping relationship through dynamic analysis. The quantitative mapping relationship is the quantitative mapping relationship between load, internal force and stress. The multidimensional physical model includes a structural mechanics model and a temperature field model.

[0088] It should be noted that transmitting load and environmental data to a multi-dimensional physical model and obtaining the quantitative mapping relationship between load, internal force, and stress through dynamic analysis can be understood as comprehensively simulating the physical behavior of a bridge crane under different working conditions through structural mechanics and temperature field simulations. Specifically, it can simulate the stress distribution at different load locations using a structural mechanics model and an integrated thermodynamic (motor temperature rise) simulation model simulating multiple environmental factors. Since the structural stress index of a bridge crane is directly affected by both structural characteristic parameters (such as geometric dimensions and material properties) and service internal forces, and the internal forces essentially originate from the dynamic action of external loads, the prerequisite for constructing a high-precision stress spectrum is the systematic analysis of the structural service load spectrum information and the establishment of a quantitative mapping relationship between load, internal force, and stress.

[0089] For example, Figure 4 A simplified mechanical model of the main beam in the vertical direction is used, where the uniformly distributed self-weight load q of the main beam, the span S of the main beam, and the wheel track b of the trolley are inherent design parameters, and the wheel pressure of the trolley is directly determined by P1 and the self-weight P2 of the trolley. Based on Newton's third law, the support reaction force F... l and F r The bending moment M in the vertical direction at the mid-span section of the trolley can be determined by formulas (1) to (3). mid-span See formula (4).

[0090] P1 + P2 + qS = F l +F r (1)

[0091]

[0092] In formulas (1) to (4), P1 represents the lifting load, P2 represents the trolley's self-weight, q represents the uniformly distributed load of the main beam's self-weight, S represents the main beam's span, and F... l and F rThe values ​​represent the support reactions at the left and right ends of the main beam, respectively; b represents the wheel track of the trolley; x represents the distance from the left wheel of the trolley to the left end of the main beam; and M represents the distance from the left wheel of the trolley to the left end of the main beam. mid-span This represents the bending moment at the mid-span section. Furthermore, a parametric main beam model can be created in simulation software (such as ANSYS), and boundary conditions based on measured load spectra can be applied, combined with measured sensor data from the bridge crane, to verify simulation errors. The temperature field model integrates thermodynamic simulation to model the impact of various environmental factors, such as motor temperature rise, on crane performance, providing data support for equipment thermal management.

[0093] Step S303: The preset rules are transmitted to the behavior rule model, and the preset rules are integrated through the knowledge graph to obtain the special rule knowledge graph.

[0094] It should be noted that preset rules can be understood as a set of domain constraints and logical conditions that must be followed during crane operation. For example, preset rules may include operational logic (mandatory procedures based on physical laws and equipment limitations), fault evolution mechanisms (causal chains in the process of component failure), safety standards (quantified execution conditions of industry regulations), etc. These rules are integrated through a knowledge graph and transformed into a machine-parseable and reasonable knowledge framework (i.e., a special rule knowledge graph), laying the foundation for subsequent real-time safety monitoring and intelligent decision-making.

[0095] Step S304: Based on the design information, quantized mapping relationship and special rule knowledge graph, an initial twin model is obtained, which is used to output a virtual dataset.

[0096] For example, part-level design information and component-level design information drive geometric reconstruction, define the spatial topology of the computational domain, and inject rules into the initial twin model in real time by establishing a physical field mapping database and behavioral constraint logic (such as forcibly adding inequality constraints in the solution function).

[0097] In this embodiment, design information provides geometric entities and material distribution, quantifies mapping relationships to define mechanical response calculation rules, and a special rule knowledge graph delineates rule boundary constraints. Through multi-dimensional modeling and high-fidelity simulation technology, accurate mapping and dynamic deduction of physical entities in virtual space are achieved. This ensures that the initial twin model possesses both physical realism and meets engineering safety constraints, establishing a high-precision benchmark for subsequent real-time corrections.

[0098] Reference Figure 5 , Figure 5 This is a flowchart illustrating the third embodiment of the method for constructing a digital twin model of a crane according to this application. Based on the above... Figure 3 The second embodiment shown presents a third embodiment of the method for constructing a digital twin model of the crane of this application.

[0099] In the third embodiment, step S40 includes:

[0100] Step S401: Identify the parameter deviation between the physical dataset and the virtual dataset to obtain the deviation dataset.

[0101] It should be noted that due to data transmission and processing delays, as well as changes in the external environment and actual working conditions, inconsistencies may exist between the virtual model (i.e., the target twin model) and the physical model. Furthermore, since the number of sensors cannot fully reflect all aspects of the crane's operation, the virtual model and the physical model may exhibit different operating conditions. This can cause subsequent simulation analyses to deviate from reality, reducing the reliability and confidence of the results. Therefore, parameter deviation can be considered as the numerical difference between the physical dataset and the virtual dataset under the same spatiotemporal reference. Understandably, due to different interference causes, multiple different deviation data points may arise, which can be aggregated into a deviation dataset.

[0102] Step S402: Calculate the similarity between real-time sensor data and historical interference data using principal component analysis to obtain the similarity results.

[0103] It should be noted that when errors are not caused by human intervention, it is assumed that there is an unknown interference in the model building process. It is necessary to determine the cause of the error between the initial twin model and the physical model, and then analyze and resolve the reasons for the inconsistency. See step S402 for details.

[0104] Step S402 includes: determining the sampling data matrix and interference data matrix based on real-time sensor data and historical interference data; standardizing the sampling data matrix and interference data matrix respectively to obtain a standardized real-time matrix and a standardized historical matrix; performing principal component analysis on the standardized real-time matrix and the standardized historical matrix to extract the first K principal component vectors; obtaining the time series data of the first principal component vector based on the first K principal component vectors; and calculating the minimum cumulative distance between the standardized real-time matrix and each standardized historical matrix using a dynamic time normalization algorithm based on the time series data of the first principal component vector to determine the similarity result.

[0105] Specifically, the sampled data matrix is ​​represented by formula (5):

[0106]

[0107] In formula (5), S(t) represents the value from sensor x1 to x2. K At time [t] s ,t w The collected data matrix within ], x1(t s ) indicates that the first sensor is at time t s The sampled data, x K (ts ) represents the sampling data of the K-th sensor at time t s , and x1(t w ) represents the sampling data of the 1st sensor at time t w , and x K (t w ) is the sampling data of the K-th sensor at time t w . Assume that the unknown interference is divided into j categories (such as environmental interference, material interference, etc.). The historical time period [t jc , t jd for collecting interference data matrix is represented as the interference sampling data matrix, as shown in formula (6).

[0108]

[0109] In formula (6), S j (t) is the interference data matrix of the j-th category. x1(t jc ) is the interference data collected by the 1st sensor at time t when the interference factor j takes effect jc , and x1(t jd ) is the interference data collected by the 1st sensor at time t when the interference factor j takes effect jd , and x K (t jc ) is the interference data collected by the K-th sensor at time t when the interference factor j takes effect jc , and x K (t jd ) is the interference data collected by the K-th sensor at time t when the interference factor k takes effect jd . Use the principal component analysis method to calculate the similarity between S(t) and S j (t), and perform dynamic time warping. Specifically, normalize the real-time sensor matrix S(t) and the historical matrix S j (t) of the j-th category of interference respectively (eliminating the influence of dimensions) to form an m×K-dimensional data matrix (m is the number of time points, and K is the number of sensors). Perform PCA on the two groups of matrices respectively, extract the orthogonal eigenvectors (principal component directions), and retain the first p principal components (p << K) with a cumulative contribution rate > 95%. Take the time series data of the first principal component as the representative feature (because it has the largest contribution), generate two time series, find the minimum cumulative distance path through dynamic programming, and output the minimum distance as the similarity calculation result.

[0110] Step S403: Screen from the deviation dataset based on the similarity result to obtain the target correction parameter.

[0111] Understandably, based on the similarity calculation results, the output minimum distance indicates the highest similarity. The highest similarity is considered the current cause of interference, and parameters are then adjusted or corrected based on this cause. See the detailed process below. Figure 6 .

[0112] In this embodiment, a dynamic error matrix is ​​constructed by quantifying the deviation between virtual and real data. Principal component analysis is then used to perform similarity matching between real-time sensor data and historical interference patterns. Based on this, target correction parameters that characterize the defects of the system model are intelligently selected from the deviation dataset. This significantly suppresses the pollution of the model correction process by environmental noise and instantaneous interference, ensuring that the correction parameters accurately point to the essential modeling error of the crane twin model. This improves the effectiveness and anti-interference ability of model parameter correction and provides a pure error driving signal for the construction of the target twin model.

[0113] Reference Figure 7 , Figure 7 This is a flowchart illustrating the fourth embodiment of the method for constructing a digital twin model of a crane according to this application. Based on the above... Figure 5 The third embodiment shown presents a fourth embodiment of the crane model construction method of this application.

[0114] In the fourth embodiment, step S50 includes:

[0115] Step S501: Construct a correlation model between the target correction parameters and the target variable, and obtain the output value of the correlation model.

[0116] It should be noted that the target correction parameter can be understood as the variable to be adjusted that has a significant impact on the model output, selected through sensitivity analysis. The target variable can be understood as a latent variable that affects the model results but is not directly determined (such as component wear, environmental interference factors, etc.). Specifically, the target correction parameter can be input into the correlation model, and based on prior experimental data (the target correction parameter as the independent variable and the sensor data as the dependent variable), a regression model (i.e., the correlation model) between the target correction parameter and the target variable can be established; then, the objective function can be constructed, that is, the correlation model can be transformed into a calculation function of the target variable; finally, the target correction parameter can be input into the correlation model to obtain the estimated value of the target variable.

[0117] Step S502: Input the output value of the associated model into the initial twin model to perform dynamic mapping correction of the model parameters, and obtain the target twin model.

[0118] It should be noted that the association model can be understood as a mathematical mapping relationship (such as a regression model) between the target adjustment parameters and the target variable, used to quantify the influence mechanism between the two. The output value of the association model is the estimated value of the target variable calculated after inputting the target adjustment parameters into the association model. Specifically, the output value of the association model is used as a new input for the target variable, replacing the undetermined latent variables in the original twin model; then, through a parameter mapping mechanism, the estimated value of the target variable is written into the corresponding parameter positions of the initial twin model in real time; finally, a target twin model with dynamically updated parameters is completed.

[0119] In this embodiment, by constructing a correlation model between target correction parameters and target variables, target variables that are difficult to measure directly (such as wear and environmental interference) are transformed into computable explicit functions. This breaks through the error bottleneck caused by neglecting latent variables in traditional models. The corrected target twin model can adapt to more complex operating scenarios (such as variable load and multiple disturbances), and ultimately solves the defects of insufficient physical mapping accuracy and poor real-time monitoring in the technical problems.

[0120] In one embodiment, based on the above embodiments, this application proposes a method for predicting the lifespan of a crane, comprising: determining twin data based on the model output value of a target twin model, wherein the target twin model is constructed using the method described in steps S10 to S50; fusing the physical dataset and twin data through a first neural network to extract fatigue life state parameters of key components; inputting the fatigue life state parameters into a second neural network for time series degradation analysis to obtain load spectrum statistical results; and outputting the predicted remaining lifespan value of the key components based on fatigue damage accumulation theory and load spectrum statistical results.

[0121] It should be noted that twin data can be virtual operating parameters generated for the target twin model (which can be compared to virtual datasets, such as simulated stress distribution data; in other words, twin data is more precise virtual data). Fatigue life state parameters can be quantitative indicators (such as stress amplitude, cycle count, and damage factor) characterizing the degree of fatigue damage in key components (such as main beams and welds). Load spectrum statistical results can be used to quantify the contribution of actual working conditions to fatigue damage based on load distribution characteristics extracted from time-series data (such as peak load frequency and average stress level). The remaining life prediction value is the expected service life of key components before failure.

[0122] Specifically, the steps for extracting fatigue life state parameters of key components by fusing physical datasets and twin data through a first neural network include: determining the location of fatigue-prone cracks based on finite element analysis; converting the fatigue life curve of the material into a fatigue life curve under actual working conditions; determining real-time load data and stress distribution data based on the location of fatigue-prone cracks and the fatigue life curve under actual working conditions; inputting the real-time load data and stress distribution data into a convolutional neural network for feature fusion, and obtaining the fatigue life state parameters of key components through convolutional kernel operations.

[0123] For example, the first neural network can be a one-dimensional convolutional neural network (CNN). Understandably, before using the CNN for mapping, it needs to be trained using historical data. The input values ​​to the CNN can be measured data and twin data at various historical time points, and the output is fatigue life state parameters. By performing finite element analysis on the crane structure, and based on the load conditions at local locations prone to fatigue cracking, fatigue prediction software is used to process and transform the material's fatigue life curve (SN curve) to obtain the SN curve under actual working conditions, thereby achieving the prediction and analysis of fatigue life under the current condition. The output state parameters are referenced to crane safety requirements, and the measured values ​​are compared with standard requirements to determine its reliability.

[0124] CNNs utilize convolutional layers on top of neural networks, extracting features from the input data through convolution operations and adjusting parameters during training using backpropagation. During computation, the convolutional kernel can learn multiple features while reducing the number of parameters in the entire model, thus lowering the risk of overfitting. The calculation process of the convolutional kernel is shown in formula (7).

[0125]

[0126] In formula (7), This represents the input and bias of the k-th neuron in the l-th convolutional layer; Let i be the weights of neurons i and k in layer l-1; N is the output of the i-th neuron in layer l-1; l-1 This represents the number of neurons in the (l-1)th convolutional layer.

[0127] CNN models use fully connected layers instead of pooling layers to avoid feature loss during pooling. Fully connected layers enable the combination of data features. After the features have undergone convolution operations, they are processed by fully connected layers to output the final result. The calculation process of fully connected layers is shown in formula (8).

[0128]

[0129] In formula (8), This represents the output of the l-th fully connected layer; This represents the input of the l-th fully connected layer; Indicates the weights of the l-th layer; This represents the bias parameters of the l-th layer. The CNN-based crane service process operation state analysis combines all parameters from the physical model and the twin model, achieving higher complexity with fewer parameters and more accurate state analysis.

[0130] It should be noted that the second neural network can be a gated recurrent unit (GRU). Combining fatigue damage accumulation theory and load statistics, the GRU is used to predict the remaining life of the evaluation state data and obtain the reference failure time. The GRU can filter the time series data information through a recurrent neural network composed of an update gate and a reset gate composed of two sigmoid functions. The sigmoid function is used to transform the previous memory state information and the current input data, thereby realizing the function of data information memory and loss compared with traditional time series prediction models. The GRU has fewer parameters and is less prone to overfitting. At the same time, due to the reduction of parameters, the amount of computation is greatly reduced, and the efficiency and practicality are greatly improved. In order to realize the degradation prediction of the crane's service performance, the state parameters of the crane during operation are first recorded in real time to form a historical database. The gated neural network GRU is trained through historical data. The input of a single GRU neural unit is the value of the previous moment, and the output is the predicted value of the next moment. The predicted value is expressed as formula (9).

[0131] h t =(1-Z) t )×h t-1 +Z×h t (9)

[0132] In formula (9), Z t Indicates updating the gating, Z t The closer to 1, the more data information is "memorized"; h t h represents an intermediate variable containing the current input information. t-1 This represents the output of the hidden layer from the previous time step. Based on the load conditions at local locations of the bridge crane and the SN curve of the material, a gated neural network is used to continuously predict the state parameters, thereby enabling the prediction and analysis of fatigue life.

[0133] In this embodiment, CNN utilizes convolutional kernels to extract spatial correlation features between physical and twin data, generating highly reliable fatigue state parameters; GRU's gating mechanism filters key historical states, accurately statistically analyzes load spectrum distribution, and avoids information loss caused by pooling layers. Combining fatigue damage theory with GRU temporal extrapolation, the remaining life is dynamically calculated, providing real-time decision support for safety assessment, and realizing the prediction and demonstration application of the remaining life of bridge cranes.

[0134] It should be noted that the above examples are only for understanding this application and do not constitute a limitation on the model construction method of the crane in this application. Any simple modifications based on this technical concept are within the protection scope of this application.

[0135] This application also provides a crane model building device, please refer to... Figure 8 The crane model building apparatus includes:

[0136] The model building module 10 is used to build a multidimensional digital twin model. The multidimensional digital twin model includes at least a physical entity layer, a connection layer, a virtual model layer, a twin data layer, and a service function layer. The physical entity layer is used to collect the current operating parameters of the crane. The virtual model layer includes a target twin model, which is used to map the physical entity of the crane into a virtual space. The twin data layer is used to manage the physical dataset and the virtual dataset. The service function layer is used to correct key components of the crane.

[0137] Data preprocessing module 20 is used to transmit the current operating parameters of the crane to the connection layer for preprocessing to obtain the physical dataset. The preprocessing includes protocol conversion, interface standardization and encryption.

[0138] An initial twin module 30 is used to transmit the physical dataset to the virtual model layer, obtain an initial twin model through dynamic analysis and knowledge graph processing, and the initial twin model is used to output the virtual dataset;

[0139] The parameter acquisition module 40 allows the user to transmit the virtual dataset to the twin data layer and calculate the target correction parameters based on the physical dataset and the virtual dataset.

[0140] The model correction module 50 is used to correct the model parameters of the initial twin model based on the target correction parameters to obtain the target twin model.

[0141] The crane model building apparatus provided in this application, employing the crane model building method described in the above embodiments, can solve the technical problem of insufficient physical mapping accuracy in existing models. Compared with the prior art, the beneficial effects of the crane model building apparatus provided in this application are the same as those of the crane model building method provided in the above embodiments, and other technical features in the crane model building apparatus are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.

[0142] This application provides a crane model building device. The crane model building device 2000 includes: at least one processor 2001; and a memory 2003 communicatively connected to the at least one processor 2001. The memory 2003 stores instructions that can be executed by the at least one processor 2001. The instructions are executed by the at least one processor 2001 to enable the at least one processor 2001 to execute the crane model building method or life prediction method in the above embodiments.

[0143] The following is for reference. Figure 9 This diagram illustrates a structural schematic of a crane model-building device 2000 suitable for implementing embodiments of this application. The crane model-building device 2000 in this application embodiment may include a processor 2001 and a memory 2003. The processor 2001 and memory 2003 are connected, for example, via a bus 2002. Optionally, the crane model-building device 2000 may also include a transceiver 2004. It should be noted that in practical applications, the transceiver 2004 is not limited to one, and the structure of this crane model-building device 2000 does not constitute a limitation on the embodiments of this application.

[0144] Processor 2001 may be a CPU, a general-purpose processor, a DSP, an ASIC, an FPGA, or other programmable logic device, transistor logic device, hardware component, or any combination thereof. It may implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. Processor 2001 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0145] Bus 2002 may include a pathway for transmitting information between the aforementioned components. Bus 2002 may be a PCI bus or an EISA bus, etc. Bus 2002 can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 9 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0146] The memory 2003 may be ROM or other type of static storage device capable of storing static information and instructions, RAM or other type of dynamic storage device capable of storing information and instructions, or EEPROM, CD-ROM or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto.

[0147] The memory 2003 is used to store the application code that executes the scheme of this application, and its execution is controlled by the processor 2001. The processor 2001 is used to execute the application code stored in the memory 2003.

[0148] It should be understood that although the steps in the flowcharts of the accompanying figures are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the accompanying figures may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0149] The above description is only a partial embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for constructing a digital twin model of a crane, characterized in that, The method includes: A multidimensional digital twin model is constructed, comprising at least a physical entity layer, a connection layer, a virtual model layer, a twin data layer, and a service function layer. The physical entity layer is used to collect the current operating parameters of the crane. The virtual model layer includes a target twin model, which is used to map the physical entity of the crane into a virtual space. The twin data layer is used to manage the physical dataset and the virtual dataset. The service function layer is used to correct key components of the crane. The current operating parameters of the crane are transmitted to the connection layer for preprocessing to obtain the physical dataset. The preprocessing includes protocol conversion, interface standardization, and encryption. The physical dataset is transmitted to the virtual model layer, and an initial twin model is obtained through dynamic analysis and knowledge graph processing. The initial twin model is used to output the virtual dataset. The virtual dataset is transmitted to the twin data layer, and the target correction parameters are calculated based on the physical dataset and the virtual dataset. Based on the target correction parameters, the model parameters of the initial twin model are corrected to obtain the target twin model.

2. The method as described in claim 1, characterized in that, The virtual simulation layer includes a multi-level geometric model, a multi-dimensional physical model, and a behavioral rule model. The physical dataset includes physical geometric data, material data, load data, environmental data, and preset rules. The step of transmitting the physical dataset to the virtual model layer and obtaining an initial twin model through dynamic analysis and knowledge graph processing, wherein the initial twin model is used to output the virtual dataset, includes: The physical geometry data and material data are transmitted to the multi-level geometric model to obtain design information, which includes part-level design information and component-level design information. The load data and environmental data are transmitted to the multidimensional physical model, and a quantitative mapping relationship is obtained through dynamic analysis. The quantitative mapping relationship is the quantitative mapping relationship between load, internal force and stress. The multidimensional physical model includes a structural mechanics model and a temperature field model. The preset rules are transmitted to the behavior rule model, and the preset rules are integrated through a knowledge graph to obtain a special rule knowledge graph; Based on the design information, the quantization mapping relationship, and the special rule knowledge graph, an initial twin model is obtained, which is used to output the virtual dataset.

3. The method as described in claim 1, characterized in that, The step of transmitting the virtual dataset to the twin data layer and calculating the target correction parameters based on the physical dataset and the virtual dataset includes: Identify the parameter deviations between the physical dataset and the virtual dataset to obtain the deviation dataset; The similarity between real-time sensor data and historical interference data is calculated using principal component analysis to obtain similarity results. Based on the similarity results, the target correction parameters are obtained by filtering from the deviation dataset.

4. The method as described in claim 3, characterized in that, The step of calculating the similarity between real-time sensor data and historical interference data using principal component analysis to obtain the similarity result includes: The sampling data matrix and the interference data matrix are determined based on real-time sensor data and historical interference data. The sampled data matrix and the interference data matrix are standardized to obtain a standardized real-time matrix and a standardized historical matrix. Principal component analysis is performed on the standardized real-time matrix and the standardized historical matrix to extract the first K principal component vectors. Based on the first K principal component vectors, the time series data of the first principal component vector is obtained; Based on the time-series data of the first principal component vector, the minimum cumulative distance between the standardized real-time matrix and each of the standardized historical matrices is calculated using a dynamic time-normalization algorithm to determine the similarity result.

5. The method as described in claim 4, characterized in that, The sampled data matrix is ​​represented as follows: Where S(t) represents the sensor from x1 to x K At time [t] s ,t w The collected data matrix within ], x1(t s ) indicates that the first sensor is at time t s The sampled data, x K (t s ) represents the Kth sensor at time t s The sampled data, x1(t) w ) indicates that the first sensor is at time t w The sampled data, x K (t w ) represents the Kth sensor at time t w The sampling data; The interference data matrix is ​​represented as follows: Among them, S j (t) is the j-th type of interference data matrix, x1(t) jc The first sensor operates for the time t during which the interference factor j takes effect. jc The collected interference data, x1(t) jd The first sensor operates for the time t during which the interference factor j takes effect. jd The collected interference data, x K (t jc The time t for the Kth sensor to be affected by interference factor j jc The collected interference data, x K (t jd The time t for the Kth sensor to be affected by interference factor j jd The collected interference data.

6. The method as described in claim 1, characterized in that, The step of correcting the model parameters of the initial twin model based on the target correction parameters to obtain the target twin model includes: Construct a correlation model between the target correction parameters and the target variable, and obtain the output value of the correlation model; The output value of the associated model is input into the initial twin model for dynamic mapping and correction of model parameters to obtain the target twin model.

7. A method for predicting the lifespan of a crane, characterized in that, The crane life prediction method includes: Based on the model output value of the target twin model, twin data is determined, wherein the target twin model is constructed using the method described in any one of claims 1 to 6; The fatigue life state parameters of key components are extracted by fusing the physical dataset and the twin data through the first neural network. The fatigue life state parameters are input into a second neural network for time series degradation analysis to obtain load spectrum statistics. Based on the fatigue damage accumulation theory and the statistical results of the load spectrum, the predicted remaining life of the key components is output.

8. The method as described in claim 7, characterized in that, The step of extracting fatigue life state parameters of key components by fusing the physical dataset and twin data through a first neural network includes: The location of fatigue-prone cracks was determined based on finite element analysis. Convert the fatigue life curve of the material into a fatigue life curve under actual working conditions. Based on the location of the fatigue-prone crack and the fatigue life curve under the actual working conditions, real-time load data and stress distribution data are determined. The real-time load data and stress distribution data are input into a convolutional neural network for feature fusion, and the fatigue life state parameters of key components are obtained through convolutional kernel operations.

9. A model-building device for a crane, characterized in that, The crane model building device includes: The model building module is used to construct a multidimensional digital twin model. The multidimensional digital twin model includes at least a physical entity layer, a connection layer, a virtual model layer, a twin data layer, and a service function layer. The physical entity layer is used to collect the current operating parameters of the crane. The virtual model layer includes a target twin model, which is used to map the physical entity of the crane into a virtual space. The twin data layer is used to manage the physical dataset and the virtual dataset. The service function layer is used to correct key components of the crane. The data preprocessing module is used to transmit the current operating parameters of the crane to the connection layer for preprocessing to obtain the physical dataset. The preprocessing includes protocol conversion, interface standardization, and encryption. An initial twin module is used to transmit the physical dataset to the virtual model layer, obtain an initial twin model through dynamic analysis and knowledge graph processing, and the initial twin model is used to output the virtual dataset; The parameter acquisition module allows the user to transmit the virtual dataset to the twin data layer and calculate the target correction parameters based on the physical dataset and the virtual dataset. The model correction module is used to correct the model parameters of the initial twin model based on the target correction parameters to obtain the target twin model.

10. A crane model building device, characterized in that, The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the crane modeling method of any one of claims 1 to 6 or the crane life prediction method of any one of claims 7 to 8.

Images