Crane with vibration fatigue damage deduction and control functions

Abstract

The invention provides a crane with vibration fatigue damage deduction and control functions, which comprises a first sample generation module used for generating a first sample data set taking a generated lifting point and load combination as input and taking the fatigue life cycle index as output; the second sample generation module is used for generating a second sample data set taking the control parameter as input and the lifting dynamic load coefficient as output; and the optimization module is used for carrying out load combination according to an output result of the second prediction model and then inputting the load combination result into the first prediction model, carrying out fatigue residual life deduction of the crane, optimizing control parameters and realizing control on fatigue damage of the crane. According to the method, the influence degree of the dynamic load effect on the fatigue life can be deduced, a fatigue damage control strategy is given, the control parameters matched with the deduction result are fed back to the physical entity, regulation, control, improvement and upgrading of the physical entity are guided, fatigue damage delay and control of the crane are achieved, and the method has important engineering value.

Description

Technical Field

[0001] The present invention relates to the technical field of cranes, and in particular to a crane with vibration fatigue damage deduction and control functions. Background Art

[0002] Cranes are widely used in industrial manufacturing, warehousing, logistics and other logistics. As the skeleton of the crane, the fatigue life of the metal structure determines the safety, performance and economic benefits of the crane. Therefore, in-depth research on its fatigue life is of great significance for quickly and accurately judging the safety of the crane when in use.

[0003] The vibration of a gantry crane during cargo lifting will reduce the smoothness of the lifting process, shorten the fatigue remaining life of the metal structure, accelerate the process of fatigue damage, and affect the intrinsic safety of the equipment. However, the problem of structural vibration stress feedback caused by the transient load impact of the crane in the complex service process has not been systematically solved, resulting in the accuracy of the prediction of the vibration fatigue remaining life of the crane cannot be guaranteed. Solving the above problems faces the following three challenges: the prediction accuracy of the crane load spectrum under service is also inconsistent with the complexity of the original structure; the mechanism of transient load generation that induces the structural vibration response of the crane is unclear; and it is difficult to couple the structural vibration with the stress indicators of the crane's service hoisting cycle.

[0004] Therefore, how to provide an accurate fatigue life prediction research plan for vibration fatigue damage of cranes has become a technical problem that needs to be solved urgently by technical personnel in this field. Summary of the invention

[0005] In order to solve the above technical problems, the present invention provides a crane with vibration fatigue damage deduction and control functions.

[0006] The technical solution adopted by the present invention is as follows:

[0007] An embodiment of the present invention provides a crane with the function of vibration fatigue failure deduction and control, including: a modeling module for constructing a three-dimensional virtual model according to the physical model of the crane; a first sample generation module for obtaining the fatigue remaining life of the crane under different lifting points and load combinations based on the three-dimensional virtual model, generating a first sample data set with the lifting point and load combination as the input and the fatigue remaining life as the output, where the load combination includes the lifting load and the lifting dynamic load coefficient during lifting operation; a second sample generation module for establishing the lifting dynamic model of the crane, solving the lifting dynamic model to obtain the lifting dynamic load coefficient under different control parameters, and generating a second sample data set with the control parameter as the input and the lifting dynamic load coefficient as the output, where the control parameters include the cargo mass, the initial lifting speed, and the wire rope stiffness; a training module for training the GWO-XGBoost model using the first sample data set to generate a first prediction model, and training the GWO-XGBoost model using the second sample data set to generate a second prediction model; an optimization module for using the trained GWO-XGBoost model, inputting different control parameters into the second prediction model, performing load combination according to the output result of the second prediction model and then inputting it into the first prediction model, obtaining the fatigue remaining life of the crane under different control parameters, deducing the fatigue remaining life of the crane, taking the difference between the target life and the fatigue remaining life as the objective function, aiming at minimizing the difference, optimizing the control parameters, and realizing the control of the fatigue failure of the gantry crane according to the found optimal control parameters.

[0008] The crane with the function of vibration fatigue failure deduction and control proposed above in the present invention further has the following additional technical features:

[0009] According to an embodiment of the present invention, the optimization module is used to optimize the control parameters by using the particle swarm optimization algorithm.

[0010] According to an embodiment of the present invention, it further includes: a virtual platform building module, and the virtual platform building module is used for: importing a three-dimensional virtual model in the.STL (a file format of a stereolithography computer-aided design software) format into the Unity 3D (a creation engine and development tool) software to perform material and color configuration of the crane, and at the same time building a virtual scene in Unity 3D; exporting the built virtual scene as an HTML (a hypertext markup language) file through the WebGL (Web Graphics Library, a 3D drawing protocol) platform of Unity 3D, using the IIS Web (website server) to build a website with the virtual model, and configuring the HTML file of the virtual scene into the root directory of the built website to form a virtual space so that users can open the virtual space by entering the website address in the web page; nesting the web page of the virtual space into the twin system platform framework built by Visual Studio Code (a cross-platform source code editor); building an architecture of function buttons, data input boxes, and prediction result display areas on the twin system platform; monitoring user operations and sending requests through the twin system platform, calling the GWO-XGBoost model according to the requests to complete the prediction of the fatigue remaining life of the crane, and returning and displaying the results on the interface of the twin system platform.

[0011] According to an embodiment of the present invention, the first sample generation module is specifically configured to: discretize the crane main girder, trolley track size, and actual operating conditions into different lifting points along the length direction of the main girder, combine the rated lifting load, consider the influence of the lifting dynamic load coefficient, set the interval value after the actual load combination, take a load gradient every Δm increments, and set k input sample data corresponding to n lifting points, where Δm, n, and k are preset positive integers; import the three-dimensional virtual model into Abaqus software to establish a global rough model of the crane gantry structure; set the classical working conditions, material properties, and constraint conditions, analyze the stress nephogram and displacement nephogram information of the gantry structure through finite element static simulation, and determine the key parts prone to fatigue failure in combination with the engineering practice and according to the most unfavorable principle; cut out the fatigue failure parts from the global rough model and transplant the boundary conditions to establish a local fine model, and perform simulation analysis on the local fine model to obtain the dangerous welds on the key parts; compile a generalized load spectrum according to the lifting points and lifting loads, combined with the lifting dynamic load coefficient, apply it to the local fine model for finite element analysis, import the simulation results and the local fine model into the FE-SAFE software, and use the VERITY module for simulation analysis to obtain the equivalent structural stress at the dangerous welds; set the material properties of the dangerous welds, analyze the fatigue life cycle times at the dangerous welds according to the equivalent structural stress at the dangerous welds, and obtain the total damage at the dangerous welds according to the fatigue life cycle times in combination with the Miner linear cumulative damage theory. After conversion, obtain the fatigue remaining life in years, and construct a first sample data set with the lifting points and load combinations as inputs and the fatigue remaining life as the output.

[0012] According to an embodiment of the present invention, the second sample generation module is specifically configured to: equivalent the crane to a lifting dynamic model composed of mass / rotational inertia, stiffness, and damping according to the structural characteristics, load characteristics, and service characteristics of the crane; establish the vibration differential equation of the crane according to the lifting dynamic model; calculate and solve the parameters of the vibration differential equation according to the structural dimensions, cross-sectional dimensions, connection methods, and material properties of the crane physical entity, where the parameters include the mass of each component, equivalent stiffness, equivalent damping, and rotational inertia, and use MATLAB software to establish a lifting dynamic simulation model of the gantry crane; adjust the control parameters by adjusting the cargo mass, initial lifting speed, wire rope stiffness, and response time, and perform simulation solution on the lifting dynamic simulation model; use an oscilloscope to obtain the change of the wire rope tension F within the set time, and combine the wire rope tension F and the lifting dynamic load coefficient relationship formula to obtain the lifting dynamic load coefficient under different control parameters n is the pulley block ratio of the crane, and m 2 is the cargo mass.

[0013] According to an embodiment of the present invention, the vibration differential equation of the crane is obtained according to the following formula:

[0014]

[0015] where M is the output torque of the speed reducer, F is the wire rope tension, R is the drum radius, i is the transmission ratio of the commutator, n is the pulley block ratio, J 1 is the moment of inertia of the reducer motor and the coupling, J 2 is the moment of inertia of the commutator, the drum and the coupling, θ 1 is the angular displacement corresponding to J 1 is the angular displacement corresponding to θ 2 is the angular displacement corresponding to J 2 c 0 , k 0 are respectively the damping and stiffness coefficients of the coupling, c 1 , k 1 , m 1 are respectively the damping, stiffness coefficient and mass of the main beam, c 2 , k 2 , m 2 are respectively the damping, stiffness coefficient and mass of the pulley block wire rope, c 3 , k 3 , m 3 are the damping, stiffness coefficient and mass of the left outrigger, c 4 , k 4 , m 4 are respectively the damping, stiffness coefficient and mass of the right outrigger, x 1 , x 2 , x 3 , x 4 are respectively the displacements of the main beam, the goods, the left outrigger and the right outrigger, represents the first-order differential of θ 1 with respect to time, represents the second-order differential of θ 1 with respect to time, represents the first-order differential of θ 2 with respect to time, represents the second-order differential of θ 2 with respect to time, represents the first-order differential of x 1 with respect to time, represents the second-order differential of x 1 with respect to time, represents the first-order differential of x 2 with respect to time, represents the second-order differential of x 2 with respect to time, represents the first-order differential of x 3The first-order derivative with respect to time, represents x 3 The second-order derivative with respect to time, represents x 4 The first-order derivative with respect to time, represents x 4 The second-order derivative with respect to time.

[0016] According to an embodiment of the present invention, the IGWO-XGBoost model specifically includes: using the XGBoost regression algorithm as a prediction model, and using the improved grey wolf optimization algorithm IGWO to optimize the hyperparameters of the model.

[0017] According to an embodiment of the present invention, the following steps are specifically adopted to train the GWO-XGBoost model: perform data processing on the sample data set, and divide the processed data set into a training set and a test set according to a certain ratio; set the parameters of the IGWO-XGBoost model, and the parameters include the number of grey wolf populations, the maximum number of iterations, the dimension of the position vector of grey wolf individuals, and the parameter ranges of each hyperparameter; initialize the positions of the grey wolf populations, and perform fitness calculation, so as to divide the wolf packs into levels, and determine the positions of the alpha wolf, beta wolf, and delta wolf. Among them, the fitness is set as the mean square error MSE (mean-square error) of the XGBoost model; during the iteration process, traverse the positions of each wolf each time, determine the fitness and perform the division of the alpha wolf, beta wolf, and delta wolf, complete the update of the wolf pack positions, and output the position vector and fitness value of the optimal solution alpha wolf. Compare whether the fitness meets the preset threshold or the number of iterations has reached the maximum value. If it meets, output the model evaluation index. At the same time, after the number of iterations reaches the set threshold, perform mutation on the individual positions of the wolf pack with a certain probability to avoid falling into local optimum; after the iteration terminates, use the obtained optimal hyperparameters to train the XGBoost model, and use the test set samples to test the model after training, and output the evaluation index of the model.

[0018] The present invention has the following beneficial effects:

[0019] By establishing a hoisting dynamics model, studying the influencing factors of the hoisting dynamic load effect, deducing the influence degree of the dynamic load effect on the fatigue life, and giving a control strategy for fatigue failure, the control parameters matching the deduction results are fed back to the physical entity to guide the regulation, improvement and upgrade of the physical entity, realizing the delay and control of the fatigue failure of the crane, making the service life of the crane approach the design life to the greatest extent, which has important engineering value;

[0020] The IGWO-XGBoost model is used to deduce the fatigue remaining life. The global fast search ability of IGWO is utilized to optimize the hyperparameters of XGBoost, ensuring the prediction accuracy of the model while controlling the model complexity.

[0021] Based on the digital twin framework for the deduction and control of the vibration fatigue failure of a crane, and supported by the deduction and control of the vibration fatigue failure of the crane, a twin system platform for the deduction and control of the vibration fatigue failure of the crane is developed to realize the real-time behavior mapping between the physical entity and the virtual entity, complete the deduction of the fatigue life of the crane, trace the influence of dynamic load effects, and thus realize the control of the operating parameters of fatigue failure. Brief Description of the Drawings

[0022] Figure 1 It is a block diagram of a crane with the function of deducing and controlling vibration fatigue failure according to an embodiment of the present invention;

[0023] Figure 2 It is a schematic diagram of the generation principle of the first sample data set according to an embodiment of the present invention. Detailed Embodiments

[0024] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0025] Next, a crane with the function of deducing and controlling vibration fatigue failure proposed in the embodiments of the present invention will be described in conjunction with the accompanying drawings.

[0026] Figure 1 It is a block diagram of a crane with the function of deducing and controlling vibration fatigue failure according to an embodiment of the present invention. As Figure 1 shown, the crane with the function of deducing and controlling vibration fatigue failure includes: a modeling module 1, a first sample generation module 2, a second sample generation module 3, a training module 4, and an optimization module 5.

[0027] Among them, the modeling module 1 is used to construct a three-dimensional virtual model according to the physical model of the crane; the first sample generation module 2 is used to obtain the fatigue remaining life of the crane under different lifting points and load combinations based on the three-dimensional virtual model, and generate a first sample data set with the lifting point and load combination as the input and the fatigue remaining life as the output. The load combination includes: the lifting load and the lifting dynamic load coefficient during lifting operation; the second sample generation module 3 is used to establish a lifting dynamic model of the crane, solve the lifting dynamic model to obtain the lifting dynamic load coefficient under different control parameters, and generate a second sample data set with the control parameters as the input and the lifting dynamic load coefficient as the output. The control parameters include: the mass of the goods, the initial lifting speed, and the wire rope stiffness; the training module 4 is used to train the GWO-XGBoost model using the first sample data set to generate a first prediction model, and train the GWO-XGBoost model using the second sample data set to generate a second prediction model; the optimization module 5 is used to use the trained GWO-XGBoost model, input different control parameters into the second prediction model, set the load combination according to the output result of the second prediction model and then input it into the first prediction model, obtain the fatigue remaining life of the crane under different control parameters, perform the deduction of the fatigue remaining life of the crane, use the difference between the target life and the fatigue remaining life as the objective function, and take the minimum difference as the target orientation to optimize the control parameters, and realize the control of the fatigue failure of the crane according to the optimal control parameters found.

[0028] Specifically, the crane in the present invention can be a gantry crane, which mainly includes: a hoisting mechanism, a trolley traveling mechanism, a crab traveling mechanism, a gantry structure, a lifting device, a sensing system, an electric control system, a communication system, and a safety device. The gantry structure is composed of a main beam (laid with tracks), end beams, legs, and a lower cross beam, which plays a role in supporting and connecting other mechanisms and determines the service life of the whole machine. The trolley traveling mechanism is connected to the lower cross beam of the gantry structure, enabling the whole machine to move longitudinally along the tracks laid on the ground. The crab traveling mechanism is placed on the main beam tracks of the gantry structure, enabling the hoisting trolley to move transversely along the main beam tracks. The cooperation of the trolley and crab traveling mechanisms achieves full coverage of the lifting / unloading points within the operation scenario. The hoisting mechanism is fixed on the hoisting trolley and mainly includes: wire ropes, drums, couplings, stepping motors, and stepping speed reducers, etc. One end of the wire rope is fixed to the trolley frame, and the other end is connected to the movable pulley of the lifting device through winding around the drum, and the goods are lifted and lowered after being driven by the motor. The safety device can prevent the trolley and crab from running out of the tracks due to unexpected situations, and prevent the goods from hitting the fuselage when the lifting height exceeds the limit. The electric control system controls the movement of each mechanism of the whole machine. The sensing system monitors the operating state and performance characteristics of the entire gantry crane through various sensor devices, including surface strain gauges, tension sensors, lifting stroke encoders, lidar rangefinders, etc. The communication system plays a role in transmitting signals between each mechanism and each system, including but not limited to Bluetooth signal transmission during the process of the control system driving each mechanism, and data stream transmission after the sensors collect information.

[0029] According to the material, geometry, shape, position and other attributes of the crane and its service scenario, the SolidWorks software can be used for 1:1 modeling to construct a three-dimensional virtual model.

[0030] Using the three-dimensional virtual model of the crane to obtain the fatigue life cycle times of the crane under different lifting points and load combinations, generating a first sample data set with the lifting point and load combination as the input and the fatigue life cycle times as the output. The load combination includes: the lifting load and the lifting dynamic load coefficient during lifting operation. At the same time, solve the lifting dynamic model of the crane to obtain the lifting dynamic load coefficient under different control parameters (cargo mass, initial lifting speed, and wire rope stiffness), and generate a second sample data set with the control parameters as the input and the lifting dynamic load coefficient as the output. Use the first sample data set to train the GWO-XGBoost model to generate a first prediction model, and use the second sample data set to train the GWO-XGBoost model to generate a second prediction model. The optimization module is used to input different control parameters into the second prediction model using the trained GWO-XGBoost model. The second prediction model outputs the corresponding lifting dynamic load coefficient. According to the output result of the second prediction model, combined with the lifting load, after setting the load combination, input it into the first prediction model. Thus, the first prediction model can output the fatigue life cycle times of the crane under different control parameters. According to the fatigue life cycle times, the remaining fatigue life of the crane can be obtained, and the remaining fatigue life of the crane can be deduced. The optimization module continuously adjusts the control parameters, uses the difference between the target life and the remaining fatigue life as the objective function, and takes the minimum difference as the target orientation to optimize the control parameters. According to the optimal control parameters found, the fatigue damage of the gantry crane can be controlled. The target life can be set according to the crane structure design and its service conditions. Thus, by establishing a lifting dynamics model, studying the influencing factors of the lifting dynamic load effect, deducing the influence degree of the dynamic load effect on the fatigue life, and giving a control strategy for fatigue damage, feedback the control parameters matching the deduction results to the physical entity to guide the regulation, improvement, and upgrade of the physical entity, realize the delay and control of the crane's fatigue damage, make the fatigue life of the crane approach the expected life to the greatest extent, which has important engineering value.

[0031] In a specific embodiment of the present invention, the optimization module 5 can be used to optimize the control parameters by using the particle swarm optimization algorithm. The specific optimization process is as follows: determine the optimized control parameters (initial lifting speed, lifting mass, wire rope stiffness) and set their respective value ranges according to actual needs and service conditions, and determine the fitness as the difference between the predicted life and the target life under this parameter; initialize the particle swarm, determine the number of particles in the particle swarm, and randomly determine the initial positions of each particle; set the number of iterations and the iteration termination conditions; according to the fitness value, determine the running speed and direction of each particle, perform iterations and calculate the new fitness; complete the iteration and output the optimal control parameters when meeting the target life. Adjust the actual operation of the crane according to the obtained optimal control parameters to make the fatigue life of the crane approach the expected life.

[0032] The following describes in detail how to generate the first sample dataset with specific embodiments.

[0033] According to an embodiment of the present invention, the first sample generation module is specifically configured to: discretize the crane main girder, trolley track dimensions, and actual operating conditions into different lifting points in the length direction of the main girder, combine the rated lifting load, consider the influence of the lifting dynamic load coefficient, set the interval value of the actual load combination, take a load gradient every Δm increments, and set k input sample data corresponding to n lifting points, where Δm, n, and k are preset positive integers. Thus, the setting of the lifting point and the load combination is completed. Import the three-dimensional virtual model into Abaqus software to establish a global rough model of the crane gantry structure. Set the classical working conditions, material properties, and boundary conditions, and analyze the stress nephogram and displacement nephogram information of the gantry structure through finite element static simulation. Combine the engineering practice and determine the key parts prone to fatigue failure according to the most unfavorable principle. Cut out the fatigue failure parts from the global rough model and transplant the boundary conditions to establish a local fine model, and obtain the dangerous welds on the key parts through simulation analysis of the local fine model. According to the lifting point and the lifting load, combine the lifting dynamic load coefficient, compile a generalized load spectrum, apply it to the local fine model for finite element analysis, import the simulation results and the local fine model into the FE-SAFE software, and use the VERITY module for simulation analysis to obtain the equivalent structural stress at the dangerous welds. Set the material properties of the dangerous welds, analyze the fatigue life cycle times at the dangerous welds according to the equivalent structural stress at the dangerous welds, and obtain the total damage at the dangerous welds according to the fatigue life cycle times and the Miner linear cumulative damage theory. After conversion, obtain the fatigue remaining life in years, and construct a first sample dataset with the lifting point and load combination as the input and the fatigue remaining life as the output.

[0034] Specifically, the vibration during the lifting process of the goods of the gantry crane will reduce the smoothness of the lifting process, shorten the fatigue remaining life of the metal structure, accelerate the process of fatigue damage, and affect the intrinsic safety of the equipment body. Therefore, it is necessary to consider the influence of the vibration during the lifting process of the goods of the lifting dynamic load crane on the fatigue remaining life, and introduce the lifting dynamic load coefficient, where the lifting dynamic load coefficient is set by technicians according to experience. As Figure 2 shown, after constructing a three-dimensional virtual model of the crane physical entity using SolidWorks software, import it into Abaqus software to establish a global rough model of the gantry structure. After setting the classical working conditions, material properties, and boundary conditions, analyze the stress nephogram and displacement nephogram information of the gantry structure through finite element static simulation. Combine the engineering practice and determine the key parts prone to fatigue failure according to the most unfavorable principle; cut out the fatigue failure parts from the global rough model and transplant the boundary conditions to establish a local fine model, and obtain the dangerous welds on the key parts after simulation analysis.

[0035] According to the lifting points and the lifting load, combined with the dynamic load factor of lifting, a generalized load spectrum is compiled and applied to the local refined model for finite element analysis. The simulation results and the refined model are imported into the FE-SAFE software together; the dangerous welds are defined, and the VERITY module is used for simulation analysis to obtain the equivalent structural stress at the dangerous welds; the material properties of the dangerous welds are set, the fatigue life cycle times at the dangerous welds are analyzed, and the fatigue remaining life in years is obtained through conversion. Thus, a first sample data set with the lifting points and load combinations as inputs and the fatigue remaining life as the output is constructed.

[0036] The following describes how to analyze the fatigue life cycle times at the dangerous welds based on the equivalent structural stress in combination with specific embodiments.

[0037] As a typical welded structural member, once a crack appears at the welded joint of the crane metal structure, the crack will gradually expand under the action of the working load until fracture occurs. Under the action of a constant stress amplitude σ i , the fatigue life cycle times at the welded joint can be calculated using the fracture mechanics Paris formula. According to the ratio of the weld length to the plate thickness a / t, 0 < a / t < 0.1 is defined as a short crack, and 0.1 < a / t < 1 is defined as a long crack. The Paris formula unifies the two crack propagation stages as:

[0038] da / dN = C[f 1 (ΔK) a / t≤0.1 , f 2 (ΔK) a / t＞0.1 (1);

[0039] where C is a constant; ΔK is the change range of the stress intensity factor that conforms to the stress range at the far end, f 1 (ΔK) a / t≤0.1 is the short crack propagation rate function, and f 2 (ΔK) a / t＞0.1 is the long crack propagation rate function.

[0040] Considering the influence of the notch stress concentration exhibited by the self-equilibrating part under the actual stress state, it is assumed that f 1 (ΔK) a / t≤0.1 and f 2 (ΔK) a / t＞0.1 conform to the law of energy conservation, and rewrite the above formula (1) as:

[0041] da / dN = C(M kn ) n (ΔK) m (2);

[0042] In the formula, M knis the notch stress intensity amplification factor; n and m are exponents determined according to the test data of typical short cracks and long cracks.

[0043] Integrating the above formula (2), the fatigue life from the generation of the weld crack to the crack penetrating the plate thickness t can be obtained as:

[0044]

[0045] In the formula, I(r) is a dimensionless function of the load bending ratio r; Δσ s is the structural stress range of variation.

[0046] Based on formula (3) and starting from engineering practice, by using fatigue test data to correct the two constants C and m in the above formula, we can obtain:

[0047] N = (ΔS s / C d ) -1 / h (4);

[0048] In the formula, ΔS s is the equivalent structural stress range of variation, which can be calculated according to the following formula; C d and h are test parameters; N is the number of fatigue life cycles.

[0049]

[0050] In the formula, r is the bending ratio; Δσ s is the structural stress range of variation; Δσ b is the bending stress range of variation; Δσ m is the membrane stress range of variation.

[0051] The following describes how to construct the second sample data set in combination with specific embodiments:

[0052] According to an embodiment of the present invention, the second sample generation module is specifically used for: according to the structural characteristics, load characteristics and service characteristics of the crane, equivalent the crane to a hoisting dynamic model composed of mass / rotational inertia, stiffness and damping; establish the vibration differential equation of the crane according to the hoisting dynamic model; calculate and solve the parameters of the vibration differential equation according to the structural dimensions, cross-sectional dimensions, connection methods and material properties of the physical entity of the crane, and the parameters include the mass of each component, equivalent stiffness, equivalent damping and rotational inertia, and use MATLAB software to establish the hoisting dynamic simulation model of the gantry crane; adjust the control parameters by adjusting the cargo mass, initial hoisting speed, wire rope stiffness and response time, and perform simulation solution on the hoisting dynamic simulation model; use an oscilloscope to obtain the change of the wire rope tension F within the set time, and combine the wire rope tension F and the relationship formula between the hoisting dynamic load factor between Obtain the hoisting dynamic load coefficient under different control parameters n is the pulley block ratio of the crane, m 2 is the mass of the goods.

[0053] Specifically, the problem of the structural fatigue failure mechanism caused by the transient impact during the dynamic service operation of the crane has always been unable to be effectively solved, resulting in the inability to guarantee the accuracy of the prediction of the remaining structural fatigue life. Therefore, it is necessary to study the key factors that cause the transient impact during the crane service, and explore the influence law of the transient load impact on the structural stress cycle index of the crane through the dynamics theory, which can provide a theoretical basis for the accurate prediction of the remaining life of the crane structure. For this purpose, the present invention establishes a hoisting dynamics model to study the influencing factors of the hoisting dynamic load effect, and deduces the influence degree of the hoisting dynamic load coefficient on the fatigue life.

[0054] According to the hoisting power model of the crane, combined with the characteristics of the complex mechanism and structure of the crane, a vibration differential equation that can not only reflect the vibration characteristics of the crane but also is conducive to modeling and solving calculations can be established on the premise of retaining the key factors. In a specific embodiment of the present invention, the vibration differential equation of the crane can be specifically obtained according to the following formula:

[0055]

[0056] Among them, M is the output torque of the reducer, F is the wire rope tension, R is the drum radius, i is the transmission ratio of the commutator, n is the pulley block ratio, J 1 is the moment of inertia of the reducer motor and the coupling, J 2 is the moment of inertia of the commutator, the drum and the coupling, θ 1 is J 1 corresponding angular displacement, θ 2 is J 2 corresponding angular displacement, c 0 、k 0 are the damping and stiffness coefficients of the coupling respectively, c 1 、k 1 、m 1 are the damping, stiffness coefficient and mass of the main girder respectively, c 2 、k 2 、m 2 are the damping, stiffness coefficient and mass of the pulley block wire rope respectively, c 3 、k 3 、m 3 are the damping, stiffness coefficient and mass of the left outrigger respectively, c 4 、k 4 、m 4 are the damping, stiffness coefficient and mass of the right outrigger respectively, x 1 、x 2 、x3 , x 4 represent the displacements of the main beam, the goods, the left outrigger and the right outrigger respectively, represents the first-order differential of θ 1 with respect to time, represents the second-order differential of θ 1 with respect to time, represents the first-order differential of θ 2 with respect to time, represents the second-order differential of θ 2 with respect to time, represents the first-order differential of x 1 with respect to time, represents the second-order differential of x 1 with respect to time, represents the first-order differential of x 2 with respect to time, represents the second-order differential of x 2 with respect to time, represents the first-order differential of x 3 with respect to time, represents the second-order differential of x 3 with respect to time, represents the first-order differential of x 4 with respect to time, represents the second-order differential of x 4 with respect to time.

[0057] According to the structural dimensions, cross-sectional dimensions, connection methods and material properties of the physical entities, combined with theoretical calculations, the parameters in the vibration differential equation are solved, including the masses of each component, equivalent stiffness, equivalent damping, and moment of inertia. Using the Simulink tool and its function modules built into the MATLAB software, such as the integral module, multiplication module, addition and subtraction module, constant module and integral module, etc., a hoisting dynamic simulation model of the crane is established. On this basis, by adjusting the mass m of the goods 2 , the initial hoisting speed of the goods (i.e., ), the stiffness k of the wire rope 2 and the response time, the model is simulated and solved. Using the oscilloscope of Simulink, the change of the target variable (i.e., the wire rope tension F) within the set time is obtained. Combining with the relationship formula between the wire rope tension and the hoisting dynamic load coefficient the hoisting dynamic load coefficients under different working conditions are obtained Taking this hoisting dynamic load coefficient A second sample data set with control parameters as input and hoisting dynamic load coefficient as output is constructed.

[0058] In an embodiment of the present invention, the IGWO-XGBoost model specifically includes: using the XGBoost regression algorithm as the prediction model, and using the improved grey wolf optimization algorithm IGWO to optimize the hyperparameters of the model.

[0059] Specifically, XGBoost (Extreme Gradient Boosting) is an ensemble model based on the Gradient Boosting Decision Tree (GBDT), with the Classification And Regression Tree (CART) as the base classifier, and is established by improving Boosting. During the model training process, the results of multiple Classification And Regression Trees (CART) are integrated to make up for the shortcoming of the insufficient accuracy of the prediction results of a single CART. If the total amount of the CART subtree space is M, the model output result is:

[0060]

[0061] In the formula, is the prediction result of the i-th sample; m is the number of trees; x i represents the i-th sample, and f m (x i ) is the prediction result of the m-th round of samples.

[0062] The objective function of XGBoost includes two parts: a regularization term and a loss function, and the minimum objective function Obj can be expressed as:

[0063]

[0064] In the formula, l is the loss function; C is a constant term; Ω is the regularization term; γ and λ are regularization parameters for controlling the complexity of the model, y i is the true label value of the training sample, Ω(f t ) is the regularization term, T is the number of leaf nodes, and ω j is the leaf node weight.

[0065] The relationship between the regularization term Ω and the number of leaf nodes T and the leaf node weight ω j is as shown in formula (8), then the objective function Obj(t) in the t-th iteration is:

[0066]

[0067] Performing a second-order Taylor formula expansion on the objective function (formula (9)) and deleting the constant term gives:

[0068]

[0069] where g i and h i are the first and second order gradients of the loss function respectively; v j is the sample at the j-th node; ω j represents the node weight.

[0070] Let Take the partial derivative of ω j and set it to 0 to obtain the optimal solution. Substituting the optimal solution into the objective function gives:

[0071]

[0072] ω j * is to minimize the node weight.

[0073] The Grey Wolf Optimizer (GWO) is a swarm intelligence optimization algorithm that simulates the predation behavior of grey wolves. Its principle is mainly based on the social hierarchy and hunting strategy of grey wolves. The optimal solution alpha (α) is used to simulate the position of the lead wolf, the second-best solution beta (β) and the third-best solution delta (δ) are used to simulate the positions of the second-order and third-order wolves, and the remaining candidate solutions omega (ω) are used to simulate the positions of subordinate wolves (i.e., fourth-order wolves). In the GWO algorithm, the hunting (solution seeking) is initiated and guided by the α, β, and δ wolves, and the ω wolves surround the prey. The mathematical model of this behavior can be expressed as:

[0074]

[0075] where D is the distance between an individual grey wolf and the prey; t is the current iteration number; X p is the prey position; X is the current grey wolf position; A and C are cooperation coefficient vectors. Adjusting A and C can change the influence of the prey position on the next position of the grey wolf. To simulate the process of approaching the prey, A can be set as a random number in [-a, a], and a decreases from 2 to 0 during the iteration process; r 1 and r 2 are random numbers in the interval [0, 1].

[0076] When the wolf pack starts to surround the prey, as the position of the target prey changes, the positions of the α, β, and δ wolves are also continuously iteratively changing, and the position update formula is:

[0077]

[0078] X(t + 1) = (X 1 + X 2 + X 3 ) / 3 (16)

[0079] where Xα (t), X β (t) and X δ (t) represent the position vectors of the alpha wolf, beta wolf, and delta wolf after t iterations respectively; D α , D β and D δ represent the distances between the current candidate gray wolf ω and the three optimal wolves after t iterations respectively; X(t + 1) is the position of the gray wolf after the (t + 1)-th iteration.

[0080] To overcome the problem of falling into local optimum during the iteration process, by combining the GWO optimization algorithm with the late interference mutation strategy, when the number of iterations reaches the set threshold (0.7T, where T is the maximum number of iterations), the mutation judgment of the wolf pack positions is started, so that they mutate randomly with a certain probability. At the same time, the length of the mutated position vector is controlled, and the mutated position vector is clipped so that the mutation result is still within the pre-set wolf pack position range, obtaining the mutated individual positions and calculating the fitness values of the corresponding individuals. After all individual positions in the wolf pack are updated, the wolf pack levels are re-divided according to the fitness, and the wolf pack positions are updated and mutated. Through iteration and update until the prey is successfully captured, the optimal solution is obtained. XGBoost has good performance in dealing with highly complex, non-linear, and small-scale data sets, and reduces the risk of overfitting through means such as pruning and column sampling. Therefore, in an embodiment of the present invention, by using the XGBoost regression algorithm as the prediction model and using the improved gray wolf optimization algorithm IGWO to optimize the hyperparameters of the model, the prediction accuracy of the model can be ensured while controlling the model complexity.

[0081] In one embodiment of the present invention, the training module 4 is specifically configured to: perform data processing on the sample data set, and divide the processed data set into a training set and a test set according to a certain ratio. Set the parameters of the IGWO-XGBoost model, where the parameters include the number of gray wolf populations, the maximum number of iterations, the dimension of the position vector of gray wolf individuals, and the parameter ranges of each hyperparameter. Initialize the positions of the gray wolf populations and calculate the fitness, so as to divide the wolf packs into levels and determine the positions of the α-wolf, β-wolf, and δ-wolf. Among them, the fitness is set as the mean square error of the XGBoost model. During the iteration process, the position of each wolf is traversed in each iteration to determine the fitness and perform the division of the α-wolf, β-wolf, and δ-wolf, complete the update of the wolf pack positions, and output the position vector and fitness value of the optimal solution α-wolf. Compare whether the fitness meets the preset threshold or the number of iterations has reached the maximum value. If so, output the model evaluation indicators. At the same time, after the number of iterations reaches the set threshold, mutate the individual positions of the wolf packs with a certain probability to avoid falling into local optimality. After the iteration terminates, use the obtained optimal hyperparameters to train the XGBoost model. After the training is completed, use the test set samples to test the model and output the evaluation indicators of the model. The evaluation indicators include: fitting accuracy, root mean square percentage error, mean absolute percentage error, etc.

[0082] In one embodiment of the present invention, the above crane further includes: a virtual platform building module, and the virtual platform building module is used to: import the three-dimensional virtual model in.STL format into the Unity 3D software for material and color configuration of the crane, and at the same time build a virtual scene in Unity3D. Export the built virtual scene as an HTML file through the WebGL platform of Unity 3D, use the IIS Web server to build a website with the virtual model, and configure the HTML file of the virtual scene into the root directory of the built website to form a virtual space, so that users can open the virtual space by entering the website address in the web page. Export the built virtual scene as an HTML file through the WebGL platform of Unity 3D, use the IIS Web server to build a website with the virtual model, and configure the HTML file of the virtual scene into the root directory of the built website to form a virtual space, so that users can open the virtual space by entering the website address in the web page. Build the architecture of function buttons, data input boxes, and prediction result display areas on the twin system platform. Monitor the user operations and send requests through the twin system platform, and call the GWO-XGBoost model according to the requests to complete the fatigue remaining life deduction of the crane and return the results and display them on the interface of the twin system platform.

[0083] Specifically, based on the digital twin framework for the deduction and control of the vibration fatigue failure of a crane, supported by the deduction and control of the vibration fatigue failure of the crane, with VS code as the platform, Node.js as the operating environment, Python and C# as the development languages, combined with software such as Unity 3D, MySQL, and Matlab Simulink, and combined with the corresponding communication protocols, a data transceiver and transmission method is built. Around the system functions, development processes, and system platforms, a twin system platform for the deduction and control of the vibration fatigue failure of the crane is developed to achieve real-time behavior mapping between physical entities and virtual entities, complete the deduction of the fatigue life of the crane, and trace the influence of dynamic load effects, thereby realizing the control of the operating parameters of fatigue failure.

[0084] The interface of the twin system platform can be divided into six major areas, including the overall machine performance parameter area (Area 1), the virtual-real space window area (Area 2), the overall machine operation information area (Area 3), the fatigue remaining life deduction area (Area 4), the information area for controlling the prevention of fatigue failure (Area 5), and the function button area (Area 6). Among them, the function button area consists of seven function buttons: virtual-real interaction start, transmission communication start, operation information monitoring, fatigue life deduction, fatigue failure control, information data management, and system exit. The different function services and implementation processes are as follows:

[0085] ① The overall machine performance parameter area is used to display the relevant design parameters of the gantry crane, so as to characterize some of the inherent attribute information of the equipment. ② Click the function buttons of "virtual-real interaction start → transmission communication start → operation information monitoring" in sequence, and the synchronous operation of the virtual-real model can be observed in Area 2, and the real-time service status information of the equipment such as the actual lifting weight, cargo lifting position, large / small vehicle running position, and stress at the measuring point can be presented in Area 3. ③ Click the "fatigue life deduction" function button, and use the collected information and some derivative data (historical and real-time) to complete the deduction of the historical / current / future life, and the deduction result information such as the corresponding design life, crack length, service life, and remaining life can be presented in Area 4. ④ Select the parameter to be changed in Area 5 and enter the corresponding value. After clicking the "fatigue failure control" button, the dynamic load coefficient and related life information after failure control can be presented in Area 4. ⑤ Click the buttons of "information data management → system exit" in sequence, and all connections can be disconnected, all channels and functions can be closed, and the system can be exited after the data information is saved and updated.

[0086] Thus, centering on the integrated hoisting power model of the crane structure and based on the deduced results of the fatigue remaining life, around the fatigue damage control strategy, after obtaining the matching dynamic load effect coefficient by dynamically adjusting control parameters such as the cargo mass, initial hoisting speed, and wire rope stiffness, the re-deduction of the fatigue life can be completed on the virtual entity (twin system platform), and the control parameters matching the deduced results are fed back to the physical entity to guide the regulation, improvement, and upgrade of the physical entity, realizing the control of the physical entity with the virtual one.

[0087] In summary, for the crane with the function of vibration fatigue damage deduction and control according to the embodiments of the present invention, by establishing a hoisting dynamics model, studying the influencing factors of the hoisting dynamic load effect, deducing the influence degree of the dynamic load effect on the fatigue life, and giving the control strategy for fatigue damage, the control parameters matching the deduced results are fed back to the physical entity to guide the regulation, improvement, and upgrade of the physical entity, realizing the delay and control of the crane's fatigue damage, making the service life of the crane approach the design life to the greatest extent, which has important engineering value; the IGWO-XGBoost model is used for deducing the fatigue remaining life, and the global fast search ability of IGWO is used to optimize the hyperparameters of XGBoost, ensuring the prediction accuracy of the model while controlling the model complexity; based on the digital twin framework for vibration fatigue damage deduction and control of the crane and supported by the vibration fatigue damage deduction and control of the crane, a twin system platform for vibration fatigue damage deduction and control of the crane is developed to realize the real-time behavior mapping between the physical entity and the virtual entity, complete the deduction of the crane's fatigue life, and trace the influence of the dynamic load effect, so as to realize the control of the operating parameters of fatigue damage.

[0088] In the description of the present invention, the terms "first" and "second" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of such features. The meaning of "a plurality" is two or more, unless otherwise specifically defined.

[0089] In the description of this specification, the description with reference to terms such as "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not have to refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in a suitable manner in any one or more embodiments or examples. In addition, without contradiction, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples.

[0090] Any process or method description shown in the flowchart or described otherwise herein can be understood to represent a module, segment, or portion of code including one or more executable instructions for implementing a customized logical function or process. The scope of the preferred embodiments of the present invention includes additional implementations, where functions can be executed in a substantially simultaneous manner or in a reverse order according to the involved functions, rather than in the order shown or discussed, which should be understood by those skilled in the art to which the embodiments of the present invention pertain.

[0091] It should be understood that various parts of the present invention can be implemented by hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, any one or a combination of the following techniques well known in the art can be used: discrete logic circuits with logic gate circuits for implementing logical functions on data signals, application specific integrated circuits with suitable combinational logic gate circuits, programmable gate arrays (PGAs), field programmable gate arrays (FPGAs), etc.

[0092] Those of ordinary skill in the art of the present technology can understand that all or part of the steps carried by the method of the above embodiments can be completed by instructing relevant hardware through a program, and the program can be stored in a computer-readable storage medium. When the program is executed, it includes one or a combination of the steps of the method embodiments.

[0093] In addition, each functional unit in various embodiments of the present invention can be integrated into a processing module, or each unit can exist physically alone, or two or more units can be integrated into one module. The above integrated module can be implemented in the form of hardware or in the form of a software functional module. When the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0094] The above-mentioned storage medium can be a read-only memory, a magnetic disk, an optical disk, etc.

[0095] Although the embodiments of the present invention have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those of ordinary skill in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present invention.

Claims

1. A crane with vibration fatigue damage deduction and control function, characterized in that: include: A modeling module, wherein the modeling module is used to construct a three-dimensional virtual model according to a physical model of the crane; A first sample generation module, the first sample generation module is used to obtain the fatigue remaining life of the crane under different lifting point and load combinations according to the three-dimensional virtual model, and generate a first sample data set with the generated lifting point and load combination as input and the fatigue remaining life as output, wherein the load combination includes: a lifting load and a lifting dynamic load coefficient during lifting operation; a second sample generation module, the second sample generation module is used to establish a lifting power model of the crane, solve the lifting power model to obtain the lifting dynamic load coefficient under different control parameters, and generate a second sample data set with the control parameters as input and the lifting dynamic load coefficient as output, wherein the control parameters include: cargo mass, initial lifting speed and wire rope stiffness; A training module, wherein the training module is used to train the GWO-XGBoost model using the first sample data set to generate a first prediction model, and to train the GWO-XGBoost model using the second sample data set to generate a second prediction model; The optimization module is used to use the trained GWO-XGBoost model to input different control parameters into the second prediction model, set the load combination according to the output result of the second prediction model, and then input it into the first prediction model to obtain the fatigue remaining life of the crane under different control parameters, and deduce the fatigue remaining life of the crane. The difference between the target life and the fatigue remaining life is used as the objective function, and the minimum difference is used as the goal orientation to optimize the control parameters, and control the fatigue damage of the crane is achieved according to the found optimal control parameters.

2. The crane with vibration fatigue damage deduction and control function according to claim 1 is characterized in that: The optimization module is used to optimize the control parameters using a particle swarm optimization algorithm.

3. The crane with vibration fatigue damage deduction and control function according to claim 1 is characterized in that: Also includes: A virtual platform building module, wherein the virtual platform building module is used to: Import the 3D virtual model into Unity 3D software in .STL format to configure the material and color of the crane, and build the virtual scene in Unity3D; Export the constructed virtual scene to an HTML file through the WebGL platform of Unity 3D, use the IIS Web server virtual model to build a website, and configure the HTML file of the virtual scene to the root directory of the constructed website to form a virtual space, so that users can enter the URL in the web page to open the virtual space; Embed the web page of the virtual space into the twin system platform framework built by Visual Studio Code; Build function buttons, data input boxes, and prediction result display area architecture on the twin system platform; The twin system platform monitors user operations and sends requests, and calls the GWO-XGBoost model according to the requests to complete the fatigue remaining life deduction of the crane, and returns the results and displays them on the twin system platform interface.

4. The crane with vibration fatigue failure simulation and control function according to claim 1 is characterized in that: The first sample generation module is specifically used for: According to the crane main beam, trolley track size and actual operation conditions, the main beam is discretized into different lifting points along the length direction. Combined with the rated lifting load, considering the influence of the lifting dynamic load coefficient, the interval value of the actual load combination is set, and a load gradient is taken for every Δm increments. Set k input sample data corresponding to n lifting points, and Δm, n and k are preset positive integers; Importing the three-dimensional virtual model into Abaqus software to establish a global rough model of the crane gantry structure; Set classic working conditions, material properties and constraints, analyze the stress and displacement cloud diagrams of the portal structure through finite element static simulation, and determine the key parts prone to fatigue failure based on the actual project and the most unfavorable principle; Cut out the fatigue failure part from the global rough mold and transplant the boundary conditions, establish a local fine model, and simulate and analyze the local fine model to obtain the dangerous welds on the key parts; According to the lifting point and lifting load, combined with the lifting dynamic load coefficient, a generalized load spectrum is compiled and applied to the local fine model for finite element analysis. The simulation results and the local fine model are imported into the FE-SAFE software, and the VERITY module is used for simulation analysis to obtain the equivalent structural stress at the dangerous weld. The material properties of dangerous welds are set, and the fatigue life cycles of dangerous welds are analyzed according to the equivalent structural stress at the dangerous welds. According to the fatigue life cycles and combined with Miner's linear cumulative damage theory, the total damage at the dangerous welds is obtained, and the fatigue remaining life in years is obtained after conversion. The first sample data set is constructed with the lifting point and load combination as input and the fatigue remaining life as output.

5. The crane with vibration fatigue failure simulation and control function according to claim 1 is characterized in that: The second sample generation module is specifically used for: According to the structural characteristics, load characteristics and service characteristics of the crane, the crane is equivalent to a lifting dynamic model composed of mass / rotational inertia, stiffness and damping; Establishing a vibration differential equation of the crane according to the lifting power model; According to the structural dimensions, cross-sectional dimensions, connection methods and material properties of the physical entity of the crane, the parameters of the vibration differential equation described in the formula are calculated and solved, and the parameters include the mass of each component, equivalent stiffness, equivalent damping, and moment of inertia. The lifting power simulation model of the gantry crane is established using MATLAB software; The control parameters are adjusted by adjusting the cargo mass, the initial lifting speed, the stiffness of the wire rope and the response time, and the lifting power simulation model is simulated and solved; Use an oscilloscope to obtain the change of wire rope tension F within the set time, combined with the wire rope tension F and the lifting dynamic load coefficient The relationship formula between Obtain the lifting dynamic load coefficient under different control parameters n is the crane pulley ratio, and m2 is the mass of the cargo.

6. The crane with vibration fatigue failure simulation and control function according to claim 5 is characterized in that: The vibration differential equation of the crane is obtained according to the following formula: Wherein, M is the output torque of the reducer, F is the wire rope tension, R is the drum radius, i is the commutator transmission ratio, n is the pulley group ratio, J1 is the moment of inertia of the reducer motor and coupling, J2 is the moment of inertia of the commutator, drum and coupling, θ1 is the angular displacement corresponding to J1, θ2 is the angular displacement corresponding to J2, c0 and k0 are the damping and stiffness coefficients of the coupling respectively, c1, k1 and m1 are the damping, stiffness coefficient and mass of the main beam respectively, c2, k2 and m2 are the damping, stiffness coefficient and mass of the wire rope of the pulley group respectively, c3, k3 and m3 are the damping, stiffness coefficient and mass of the left outrigger respectively, c4, k4 and m4 are the damping, stiffness coefficient and mass of the right outrigger respectively, x1, x2, x3 and x4 are the displacements of the main beam, cargo, left outrigger and right outrigger respectively, represents the first-order differential of θ1 with respect to time, represents the second-order differential of θ1 with respect to time, represents the first-order differential of θ2 with respect to time, represents the second-order differential of θ2 with respect to time, Table x1 first order differential with respect to time, represents the second-order differential of x1 with respect to time, represents the first-order differential of x2 with respect to time, represents the second-order differential of x2 with respect to time, Table x3 is the first-order differential with respect to time, represents the second-order differential of x3 with respect to time, Table x4 is the first-order differential with respect to time, Represents the second-order differential of x4 with respect to time.

7. The crane with vibration fatigue failure simulation and control function according to claim 1 is characterized in that: The IGWO-XGBoost model specifically includes: The XGBoost regression algorithm is used as the prediction model, and the improved grey wolf optimization algorithm IGWO is used to optimize the hyperparameters of the model.

8. The crane with vibration fatigue damage deduction and control function according to claim 7 is characterized in that: The training module is specifically used for: Process the sample data set and divide the processed data set into training set and test set according to a certain ratio; Set the parameters of the IGWO-XGBoost model, including the number of gray wolf populations, the maximum number of iterations, the dimension of the gray wolf individual position vector, and the parameter range of each hyperparameter; Initialize the position of the gray wolf population and calculate the fitness to divide the wolf pack into different levels and determine the positions of α wolf, β wolf, and δ wolf. The fitness is set to the mean square error of the XGBoost model. During the iteration process, each iteration traverses the position of each wolf, determines the fitness and divides it into α wolf, β wolf and δ wolf, completes the update of the wolf pack position, and outputs the position vector and fitness value of the optimal solution α wolf, and compares whether the fitness meets the preset threshold or the number of iterations has reached the maximum value. If it does, the model evaluation index is output. At the same time, after the number of iterations reaches the set threshold, the individual positions of the wolf pack are mutated with a certain probability; After the iteration is terminated, the XGBoost model is trained using the optimal hyperparameters obtained. After the training is completed, the model is tested using the test set samples and the evaluation indicators of the model are output.

Images