Device operation parameter dynamic optimization method and system based on digital twinning

Abstract

The present application relates to the technical field of simulation verification, in particular to a device operation parameter dynamic optimization method and system based on digital twinning; a heterogeneous discrete element particle bed is constructed according to geological survey data, a digital twinning calculation model coupled with multiple physical fields is established in combination with device dynamics parameters; device physical operation data are acquired and input into the digital twinning calculation model; noise is removed through a time-frequency joint analysis algorithm, and medium interaction impedance characteristics are extracted; error compensation is performed on the geological survey data based on the medium interaction impedance characteristics in combination with numerical inversion solution and spatial position anchoring, and a three-dimensional impedance distribution topology set is constructed; the three-dimensional impedance distribution topology set is taken as a simulation input boundary condition, numerical simulation deduction is performed, and a virtual resistance peak value is output; according to the virtual resistance peak value, single cutting geometric pitch within a threshold interval is introduced as rigid data constraint, bivariate differential compensation solution is performed in a virtual space, and a device parameter configuration set oriented to simulation verification is generated.

Description

Technical Field

[0001] This invention relates to the field of simulation verification technology, specifically to a method and system for dynamic optimization of equipment operating parameters based on digital twins. Background Technology

[0002] With the rapid development of computer-aided engineering and digital twin technology, simulating and analyzing the stress state of complex equipment during operation by constructing virtual computing models has become an important means of ensuring the safe operation of equipment. This virtual simulation verification is particularly important in the field of agricultural machinery for tillage and cutting operations. However, existing computer simulation verification methods have certain technical limitations when facing complex and variable unstructured farmland geographical environments.

[0003] Traditional simulation models often assume farmland soil to be an ideal homogeneous material, making it difficult to accurately reflect the objective laws governing the nonlinear changes in geophysical properties with depth. This results in severe distortion of the underlying boundary conditions of the basic computational model. When introducing agricultural machinery physical operation data into the virtual model for interactive mapping, the collected load data contains a large amount of interference from frame vibration and gear resonance. Traditional computer filtering algorithms struggle to completely extract the true soil interactive impedance characteristics, preventing the digital twin model from accurately updating the map using real external resistance.

[0004] Furthermore, when simulations encounter high-energy collisions caused by locally hard soil blocks, conventional fixed-time-step numerical integration calculations are prone to numerical divergence, leading to system crashes. Therefore, it is necessary to address the underlying defects in existing technologies, such as distortion in the underlying medium modeling, feature extraction failures due to complex dynamic noise, and computational divergence caused by high-energy collisions.

[0005] To address this, a method and system for dynamic optimization of equipment operating parameters based on digital twins are proposed. Summary of the Invention

[0006] The purpose of this invention is to provide a method and system for dynamic optimization of equipment operating parameters based on digital twins, which generates a set of equipment parameter configurations for simulation verification by performing bivariate differential compensation calculations in virtual space.

[0007] To achieve the above objectives, the present invention provides the following technical solution: A method for dynamic optimization of equipment operating parameters based on digital twins includes: Based on geological survey data, a heterogeneous discrete element particle bed is constructed, and combined with equipment dynamic parameters, a digital twin computational model with multi-physics field coupling is established. The physical operation data of the equipment is acquired and input into the virtual analysis module in the digital twin computing model; the frame vibration and gear resonance noise are eliminated by the time-frequency joint analysis algorithm, and the medium interaction impedance characteristics are extracted; based on the medium interaction impedance characteristics, combined with numerical inversion solution and spatial location anchoring, error compensation is performed on the geological survey data, and a three-dimensional impedance distribution topology set is constructed. Using the three-dimensional impedance distribution topology set as the simulation input boundary condition, numerical simulation is performed. When impedance change characteristics are identified, the contact calculation time step is locally adaptively refined. The equipment stress state within a preset time period is deduced in the digital twin calculation model, and the virtual resistance peak is output. Based on the virtual resistance peak, the single cutting geometric pitch is introduced as a rigid data constraint within a threshold range. Bivariate differential compensation is performed in the virtual space to generate a set of equipment parameter configurations for simulation verification. The set of equipment parameter configurations includes simulation configuration data for equipment rotation speed and moving speed.

[0008] Preferably, the process of establishing a multiphysics coupled digital twin computational model includes: Obtain geological survey data containing soil stratification, compaction, and composition distribution; based on the geological survey data, extract microscopic contact mechanical characteristics corresponding to different depth dimensions; combine the microscopic contact mechanical characteristics to generate virtual particle sets at corresponding depths in virtual space, and combine them to construct a heterogeneous discrete element particle bed with vertical gradient changes. Obtain the device dynamic parameters, which include the device's three-dimensional geometric topology and the constraints of the rotary joints; based on the device dynamic parameters, establish a kinematic assembly model in virtual space that is spatially mapped to the physical device; import the kinematic assembly model into a heterogeneous discrete element particle bed, and perform physical field binding calculations in combination with the set surface contact friction constitutive boundary conditions to output a multi-physics coupled digital twin calculation model.

[0009] Preferably, the process of extracting dielectric cross-impedance characteristics includes: The system acquires physical operating data of the equipment, including mechanical load values ​​and high-frequency vibration characteristics; inputs the physical operating data into a virtual analysis module; the virtual analysis module performs signal frequency band decomposition analysis on the physical operating data based on a preset time-frequency joint analysis algorithm, identifies and separates low-frequency frame turbulence interference signals and high-frequency gear resonance interference signals, reconstructs and outputs the power shaft load fluctuation sequence; based on the power shaft load fluctuation sequence, and combined with the transmission efficiency attenuation mapping relationship in the digital twin calculation model, it performs inverse dynamic solution, filters out the mechanical loss characteristics inside the model, and extracts the medium cross-impedance characteristics caused by external collisions.

[0010] The process of analyzing signal frequency bands from equipment physical operation data based on the time-frequency joint analysis algorithm includes: The physical operation data of the equipment is used as the initial time-domain sequence input to the time-frequency conversion calculation network, and the initial time-frequency energy spectrum data containing dual features of time and frequency dimensions is output through transformation and calculation. Extract the theoretical equipment rotation speed and theoretical moving speed output by the digital twin calculation model at the current moment; combine the transmission ratio topology matrix in the digital twin calculation model to perform forward kinematics deduction calculation on the theoretical equipment rotation speed, and generate high-frequency gear meshing characteristic trajectory data that dynamically moves with the farmland load; simultaneously combine the geometric features of farmland surface undulation in virtual space to perform low-frequency response solution on the theoretical moving speed, and generate low-frequency frame sway characteristic trajectory data that dynamically fluctuates with the travel speed. The generated high-frequency gear meshing characteristic trajectory data and low-frequency frame turbulence characteristic trajectory data are used as dynamic anchoring references and projected onto the initial time-frequency energy spectrum data to perform spatial intersection matching calculation; the dynamic energy convergence band that overlaps with the characteristic trajectory data in the spectrum data is extracted and identified as high-frequency gear resonance interference signal and low-frequency frame turbulence interference signal. An adaptive energy masking calculation is performed on the regions containing interference signals in the initial time-frequency energy spectrum data, and the output is a power shaft load fluctuation sequence after filtering out dynamic interference energy.

[0011] Preferably, the process of constructing a three-dimensional impedance distribution topology set by performing error compensation on geological survey data includes: The spatial absolute coordinates and heading angles, which are collected synchronously with the physical operation data of the equipment, are used as spatial position anchoring data. The extracted medium interaction impedance characteristics are timestamped and gridded with the spatial position anchoring data to obtain local impedance observations with precise geographic coordinate labels. Geological survey data, serving as the prior background field, is input into a spatial interpolation network. Based on local impedance observations, the impedance deviation matrix between the observations and the prior background field is calculated using a spatial residual computational mechanism. The impedance deviation matrix is ​​then used to perform adaptive weight correction and error compensation processing on the geological survey data of the unexplored areas ahead, outputting a dynamically assimilated and updated three-dimensional impedance distribution topology set.

[0012] Preferably, the process of outputting the virtual resistance peak includes: The spatial gradient change sequence directly in front of the three-dimensional impedance distribution topology set is extracted as the simulation input boundary condition; the spatial gradient change sequence is prospectively scanned and calculated in the digital twin computing model; based on the calculation comparison, when the local impedance value change rate is found to exceed the preset safety threshold, the impedance mutation feature is extracted and output. Based on the impedance mutation characteristics, a local adaptive refinement mechanism for the contact calculation time step is activated to adjust the time span of numerical integration calculation in the three-dimensional mesh space. Driven by the refined time step, the equipment stress state array data within a future preset time period is extrapolated. The collision extreme value is extracted from the equipment stress state array data, and a virtual resistance peak value is output for overload warning.

[0013] Preferably, the process of generating the equipment parameter configuration set includes: comparing the virtual resistance peak with the preset theoretical load threshold of the model to calculate the load overflow data that quantifies the overload risk; obtaining the single-cut geometric pitch value that characterizes the crushing quality of the operation, and limiting the single-cut geometric pitch value within a preset qualified threshold range as a rigid data constraint boundary for optimization calculation; combining the load overflow data and the rigid data constraint boundary, performing bivariate differential compensation calculation for the execution end and traction end in the virtual space; calculating the moving speed data that needs to be adjusted to offset the load overflow, and simultaneously calculating the equipment rotation speed data that needs to be adjusted to meet the constant geometric pitch condition; and generating the equipment parameter configuration set for simulation verification based on the calculated equipment rotation speed data and moving speed data.

[0014] A dynamic optimization system for equipment operating parameters based on digital twins includes: Model building module: Construct a heterogeneous discrete element particle bed based on geological survey data, and establish a multi-physics field coupled digital twin computational model by combining equipment dynamic parameters; Topology Analysis Module: Acquires physical operation data of equipment and inputs it into the virtual analysis module within the digital twin computing model; eliminates frame vibration and gear resonance noise through a time-frequency joint analysis algorithm and extracts the medium interaction impedance characteristics; based on the medium interaction impedance characteristics, combined with numerical inversion calculation and spatial location anchoring, performs error compensation on geological survey data and constructs a three-dimensional impedance distribution topology set; Virtual simulation module: Using the three-dimensional impedance distribution topology set as the simulation input boundary condition, it performs numerical simulation and simulation. When the impedance change feature is identified, it triggers local adaptive refinement of the contact calculation time step, simulates the equipment stress state within a preset time period in the digital twin calculation model, and outputs the virtual resistance peak. Simulation verification module: Based on the virtual resistance peak, the single cutting geometric pitch is introduced as a rigid data constraint within the threshold range. Bivariate differential compensation calculation is performed in the virtual space to generate a set of equipment parameter configurations for simulation verification. The set of equipment parameter configurations includes simulation configuration data of equipment rotation speed and moving speed.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. This invention addresses the problem of distorted bottom-layer boundaries caused by the over-idealization of the working medium in traditional simulation models. It proposes a method for constructing heterogeneous discrete element particle beds based on geological survey data. By extracting survey information such as soil stratification, compaction, and composition distribution, and using a nonlinear mapping neural network model, this information is transformed into microscopic contact mechanical characteristics such as cohesion and internal friction angles corresponding to different depth dimensions. These characteristics are then combined and stacked in virtual space to create a heterogeneous discrete element particle bed with vertically varying gradients. The underlying data structure highly replicates the complex physical properties of the working medium in a real natural environment, where the mechanical properties increase nonlinearly with depth, providing a solid virtual physical verification foundation for the dynamic optimization of the entire equipment's operating parameters.

[0016] 2. This invention transforms the physical operation data of the equipment into a dual-dimensional graph encompassing both time and frequency dimensions. It utilizes a digital twin model to proactively deduce the gear meshing and frame vibration characteristic trajectories as a dynamic anchoring benchmark, accurately identifying and eliminating specific interference signal energy bands. Subsequently, it further combines the transmission efficiency attenuation mapping relationship to perform inverse dynamic calculations, filtering out the mechanical transmission friction loss characteristics within the model. This achieves the extraction of purely external collision-induced dielectric impedance characteristics from noisy, mixed physical signals, reducing the system's misjudgment rate.

[0017] 3. This invention establishes a micro-scheduling and bivariate collaborative optimization mechanism. When impedance abrupt change characteristics are captured during scanning, the contact calculation time step is adaptively refined only in the local three-dimensional mesh space where interference collision is about to occur, avoiding the risk of numerical explosion and crashing, and preventing unnecessary consumption of global computing power. When generating the parameter configuration set, the single-cut geometric pitch is introduced as a rigid constraint boundary to activate the bivariate differential compensation solution. When facing the risk of load overflow, it can simultaneously output a strategy to reduce the moving speed to unload physical pressure, and a strategy to adjust the equipment speed to compensate for the loss of cutting quality. This not only ensures the safe obstacle crossing of the equipment in terms of dynamics theory, but also strictly guarantees high-quality operation in terms of agronomic standards. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating the method for dynamic optimization of equipment operating parameters based on digital twins according to the present invention. Figure 2 This is a data logic diagram of the dynamic optimization method for equipment operating parameters based on digital twins according to the present invention; Figure 3 This is a schematic diagram of the structure of the equipment operation parameter dynamic optimization system based on digital twin of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Example 1: This invention proposes a method for dynamic optimization of equipment operating parameters based on digital twins. The process of the method is as follows: Figure 1 As shown, the data logic of the method is as follows: Figure 2 As shown, it includes: Based on geological survey data, a heterogeneous discrete element particle bed is constructed, and combined with equipment dynamic parameters, a digital twin computational model with multi-physics field coupling is established. The physical operation data of the equipment is acquired and input into the virtual analysis module in the digital twin computing model; the frame vibration and gear resonance noise are eliminated by the time-frequency joint analysis algorithm, and the medium interaction impedance characteristics are extracted; based on the medium interaction impedance characteristics, combined with numerical inversion solution and spatial location anchoring, error compensation is performed on the geological survey data, and a three-dimensional impedance distribution topology set is constructed. Using the three-dimensional impedance distribution topology set as the simulation input boundary condition, numerical simulation is performed. When impedance change characteristics are identified, the contact calculation time step is locally adaptively refined. The equipment stress state within a preset time period is deduced in the digital twin calculation model, and the virtual resistance peak is output. Based on the virtual resistance peak, the single cutting geometric pitch is introduced as a rigid data constraint within a threshold range. Bivariate differential compensation is performed in the virtual space to generate a set of equipment parameter configurations for simulation verification. The set of equipment parameter configurations includes simulation configuration data for equipment rotation speed and moving speed.

[0021] Preferably, the process of establishing a multiphysics coupled digital twin computational model includes: Obtain geological survey data containing soil stratification, compaction, and composition distribution; based on the geological survey data, extract microscopic contact mechanical characteristics corresponding to different depth dimensions; combine the microscopic contact mechanical characteristics to generate virtual particle sets at corresponding depths in virtual space, and combine them to construct a heterogeneous discrete element particle bed with vertical gradient changes. Obtain the device dynamic parameters, which include the device's three-dimensional geometric topology and the constraints of the rotary joints; based on the device dynamic parameters, establish a kinematic assembly model in virtual space that is spatially mapped to the physical device; import the kinematic assembly model into a heterogeneous discrete element particle bed, and perform physical field binding calculations in combination with the set surface contact friction constitutive boundary conditions to output a multi-physics coupled digital twin calculation model.

[0022] Geological survey data was obtained from regional soil surveys and remote sensing data. Soil stratification compaction data included the distribution of soil compaction from the surface to the deeper layers, while compositional distribution included information on moisture content and the proportion of sand and clay particles. After receiving this data, it was transformed into microscopic contact mechanical characteristics, including the cohesion and internal friction angle values ​​between virtual particles. Considering that real farmland soil is not a uniform, single medium but rather exhibits a pattern of loose surface and hard subsurface (such as the plow pan), virtual particle sets with different cohesion and internal friction angles were generated layer by layer based on depth coordinates. These were then stacked to create a heterogeneous discrete element particle bed with gradually changing mechanical properties in the vertical direction.

[0023] The device dynamics parameters include the three-dimensional geometric topology file of the blade shaft and blades of agricultural machinery (such as rotary tillers), as well as the kinematic constraints between various rotating hinge points. Based on this, a kinematic assembly model with realistic mass distribution and rotational inertia is reconstructed in the computer. The virtual assembly model is imported into a heterogeneous discrete element particle bed, and combined with surface contact friction constitutive boundary conditions, physical field binding calculations are performed to output a complete multi-physics coupled digital twin computational model, capable of simulating soil cutting, soil throwing, and force feedback.

[0024] This invention constructs a heterogeneous discrete element particle bed with vertically varying gradients by introducing geological survey data and extracting microscopic contact mechanics characteristics. At the underlying data structure level, it highly replicates the complex mechanical properties of soil compaction increasing nonlinearly with depth in a real natural environment. When virtual devices are simulated within the heterogeneous particle bed, the simulation prediction distortion caused by over-idealization of the virtual environment is eliminated. Simultaneously, multiphysics-based calculations seamlessly integrate granular dynamics and multibody dynamics, providing a robust and reliable virtual physics verification platform for accurately capturing transient resistance peaks, significantly enhancing the guiding value of the digital twin system.

[0025] The process of extracting microscopic contact mechanical features corresponding to different depth dimensions includes: Geological survey data is normalized into standard feature vectors and input into a pre-trained nonlinear mapping neural network model. Spatial mapping calculation is performed through the hidden layer of the nonlinear mapping neural network model, and virtual particle cohesion and internal friction angle values ​​are generated in the output layer at the corresponding depth. By utilizing a depth gradient cyclic solution mechanism, geological survey data from different depth layers are sequentially input into a nonlinear mapping neural network model, and the microscopic contact mechanical characteristics that vary with depth coordinates are summarized and output.

[0026] The specific data processing and execution steps include: Data input and preprocessing: The input data consists of characteristic data extracted from geological survey data at specific depths and layers, including compaction values, moisture content values, and clay content values; data parsing, cleaning, and normalization operations are performed to output a standard feature vector that eliminates dimensional differences.

[0027] The model performs forward computation and feature generation; inputting standard feature vectors to a nonlinear mapping neural network model; performing layer-by-layer nonlinear activation and spatial mapping calculations through weighted connections of hidden layer nodes, outputting virtual particle cohesion and internal friction angle values ​​at a specific depth, which together constitute the microscopic contact mechanics characteristics of a single layer. To prevent the generation of discrete element parameters that violate physical principles, boundary constraint logic based on physical truncation and the Sigmoid activation function is configured in the output layer of the neural network, forcibly constraining the output virtual particle cohesion value to a positive number and strictly limiting the internal friction angle value to a physically reasonable range of 0 to 90 degrees.

[0028] The deep gradient iterative solution involves inputting feature data from multiple continuous profiles from the surface to the deep layers. Using computer iterative control logic, standard feature vectors at each depth are sequentially input into a nonlinear mapping neural network model for continuous batch solution. The output is a microscopic contact mechanics feature matrix corresponding to different depth coordinates.

[0029] The nonlinear mapping neural network model employs a multi-layer feedforward architecture, comprising an input layer, multiple nonlinear hidden layers, and an output layer. For example, the nonlinear mapping neural network model includes one input layer, four hidden layers, and one output layer. The number of neurons in each hidden layer ranges from 32 to 256, for example, 128 neurons. The hidden layers use the ReLU nonlinear activation function, and the output layer uses the Sigmoid activation function to achieve boundary constraints on the virtual particle cohesion and internal friction angle values. During the training phase, the mean squared error is used as the loss function. When the loss function value decreases to the order of 10^-3 to 10^-4, it is considered to have met the preset convergence threshold. In the model calibration and training phase, the input data consists of geological survey data from multiple sets of real soil samples measured in a physics laboratory, as well as the actual macroscopic shear strength limit values ​​measured using a direct shear apparatus.

[0030] In a virtual space, a proportionally scaled virtual soil sample is constructed. The cohesion and internal friction angle values ​​of the virtual particles are continuously combined and adjusted, and a virtual direct shear simulation is performed. The shear strength output by the virtual simulation is compared with the shear strength from the actual physical experiment. When the error is less than a preset convergence threshold, the virtual microscopic parameters are recorded. Geological survey data is used as the feature input, and the successfully inverted virtual microscopic parameters are used as the target output to construct a training dataset. Through supervised deep learning, the connection weights and biases within the network are iterated using the backpropagation algorithm, ultimately solidifying the nonlinear coupling mapping logic from soil physicochemical properties to micromechanical parameters into the weight matrix of the neural network.

[0031] This invention introduces a nonlinear mapping neural network model trained by comparing real physical damage experiments with virtual simulations, transforming complex cross-scale physical mapping relationships into precise matrix calculations. By converting readily available macroscopic geological survey data into hidden microscopic cohesion and internal friction angle values, it achieves automated and intelligent calibration of simulation parameters; ensuring that the generated microscopic contact mechanics feature matrix can characterize the objective reality of the nonlinear changes in soil mechanical properties with depth, providing a reliable underlying data verification environment for the dynamic optimization system of the entire equipment's operating parameters.

[0032] Preferably, the process of extracting dielectric cross-impedance characteristics includes: The system acquires physical operating data of the equipment, including mechanical load values ​​and high-frequency vibration characteristics; inputs the physical operating data into a virtual analysis module; the virtual analysis module performs signal frequency band decomposition analysis on the physical operating data based on a preset time-frequency joint analysis algorithm, identifies and separates low-frequency frame turbulence interference signals and high-frequency gear resonance interference signals, reconstructs and outputs the power shaft load fluctuation sequence; based on the power shaft load fluctuation sequence, and combined with the transmission efficiency attenuation mapping relationship in the digital twin calculation model, it performs inverse dynamic solution, filters out the mechanical loss characteristics inside the model, and extracts the medium cross-impedance characteristics caused by external collisions.

[0033] In complex field operation environments, the physical operation data collected by sensors directly mounted on the equipment contains noisy mixed signals; for example, the mechanical load values ​​obtained by installing a high-frequency torque sensor on the power output shaft of the equipment, and the high-frequency vibration characteristics obtained by installing a triaxial acceleration sensor on the equipment frame and gearbox housing. The physical operation data of the device input to the virtual parsing module not only includes the actual force of the blade cutting the soil, but also the low-frequency frame bumps caused by the tractor chassis running over the soil mounds, as well as the high-frequency gear resonance generated by the high-speed meshing of gears inside the gearbox; if used directly without distinction, it will lead to low recognition accuracy.

[0034] Therefore, a time-frequency joint analytical algorithm is used to decompose the signal on a two-dimensional plane formed by the time axis and the frequency axis. Since the inherent vibration frequencies of different mechanical structures vary significantly, it is possible to accurately locate and eliminate interference signal energy bands originating from the mechanical frame and gears, thereby reconstructing a pure sequence that only represents the actual vibration of the power shaft.

[0035] The reconstructed power shaft load fluctuation sequence cannot be directly equated to soil resistance. The reason is that the power output from the engine inevitably generates internal mechanical losses such as friction and heat during its journey through the clutch, drive shaft, and reduction gearbox to the cutter shaft. By retrieving the pre-stored transmission efficiency attenuation mapping relationship from the digital twin model and performing inverse dynamics calculations, these internal loss characteristics are eliminated, extracting the medium interaction impedance characteristics purely caused by external soil or rock collisions with the cutter.

[0036] The process of obtaining the pre-stored transmission efficiency attenuation mapping relationship specifically includes: The first step is to perform an unloaded bench test. With the equipment in an unloaded state where it is not in contact with the working medium, the equipment is controlled to run at multiple different speeds; the input torque of the drive shaft at this time is measured using sensors. This value represents the basic energy consumption required for the equipment to overcome the friction between its internal gears and bearings, and it is recorded as the baseline data for unloaded loss.

[0037] The second step is to perform a dynamometer load test. Connect the power output end of the equipment to a general-purpose dynamometer. When the equipment is at different speeds, the dynamometer applies multiple load torques of different gradients to simulate real working resistance. Simultaneously, the drive torque at the input end of the equipment and the actual output torque received by the dynamometer are collected. The difference between the two is calculated to obtain the torque loss difference, and the data is summarized to generate a dynamic loss dataset covering different speed and load combinations.

[0038] The third step is to generate a mapping matrix. The two sets of data mentioned above are input into the computer's data fitting module. The equipment's rotational speed and load torque are used as independent variables, and the calculated torque loss difference is used as the dependent variable. Data processing is performed through surface fitting. Feature data points after fitting are extracted to generate a lookup table matrix for direct computer access. This matrix represents the transmission efficiency attenuation mapping relationship and is ultimately pre-stored in the digital twin model.

[0039] Traditional single-frequency domain filtering methods often filter out or retain impact noise that overlaps with the useful signal frequency band, leading to severe distortion of drag data. This invention employs a time-frequency joint analytical algorithm, which can preserve transient drag characteristics while eliminating false fluctuations caused by chassis bumps and gear resonance. Furthermore, by introducing inverse dynamics calculations to filter out internal mechanical loss characteristics, the final extracted dielectric cross-impedance characteristics achieve a leap from superficial data to deep physical reality. This provides a precise numerical benchmark for subsequent error compensation and forward-looking simulations, significantly reducing the misjudgment rate of digital twin systems and improving the reliability of adaptive optimization.

[0040] The process of analyzing signal frequency bands from equipment physical operation data based on the time-frequency joint analysis algorithm includes: The physical operation data of the equipment is used as the initial time-domain sequence input to the time-frequency conversion calculation network, and the initial time-frequency energy spectrum data containing dual features of time and frequency dimensions is output through transformation and calculation. Extract the theoretical equipment rotation speed and theoretical moving speed output by the digital twin calculation model at the current moment; combine the transmission ratio topology matrix in the digital twin calculation model to perform forward kinematics deduction calculation on the theoretical equipment rotation speed, and generate high-frequency gear meshing characteristic trajectory data that dynamically moves with the farmland load; simultaneously combine the geometric features of farmland surface undulation in virtual space to perform low-frequency response solution on the theoretical moving speed, and generate low-frequency frame sway characteristic trajectory data that dynamically fluctuates with the travel speed. The generated high-frequency gear meshing characteristic trajectory data and low-frequency frame turbulence characteristic trajectory data are used as dynamic anchoring references and projected onto the initial time-frequency energy spectrum data to perform spatial intersection matching calculation; the dynamic energy convergence band that overlaps with the characteristic trajectory data in the spectrum data is extracted and identified as high-frequency gear resonance interference signal and low-frequency frame turbulence interference signal. An adaptive energy masking calculation is performed on the regions containing interference signals in the initial time-frequency energy spectrum data, and the output is a power shaft load fluctuation sequence after filtering out dynamic interference energy.

[0041] Specifically, the time-frequency conversion calculation network uses short-time Fourier transform to perform joint time-frequency analysis on the time-domain sequence; wherein, the time window length of the short-time Fourier transform is preferably set to 128 to 1024 sampling points, and the overlap rate between adjacent time windows is preferably set to 50% to 75%; through the above transformation, the initial time-frequency energy spectrum data with time on the horizontal axis, frequency on the vertical axis, and color depth representing energy intensity is output.

[0042] Specifically, for the geometric features of farmland surface undulations, the average spatial wavelength value of the surface undulations is extracted; the current theoretical moving speed is divided by the average spatial wavelength value to calculate the transient turbulence base frequency; based on this frequency value, low-frequency frame turbulence characteristic trajectory data that dynamically fluctuates with the moving speed is generated in the time-frequency spectrum.

[0043] In actual operation, the rotational speed and travel speed of the equipment fluctuate as it encounters soil of varying hardness, meaning that the gear resonance frequency and frame vibration frequency also drift continuously, causing traditional fixed-frequency threshold filters to fail. To solve this problem, the theoretical equipment rotational speed is retrieved in real time from the digital twin model. Based on the transmission ratio topology matrix, the theoretical meshing frequency of the gears in the gearbox at the current rotational speed can be accurately calculated, thus plotting a high-frequency gear meshing characteristic trajectory data that dynamically moves with the fluctuation of the operating load in the time-frequency coordinate system. Similarly, by combining the surface undulation characteristics in virtual space and the theoretical travel speed, a low-frequency frame vibration characteristic trajectory data is generated. These two dynamic trajectories are used as the baseline for mapping and projected onto the initial time-frequency energy spectrum generated by the real physical sensors. As long as the high-energy color band in the spectrum spatially intersects and matches these two trajectory lines, the machine tool's own noise can be determined, thereby accurately identifying it.

[0044] An adaptive energy masking elimination calculation is performed on the locked area, which means that the energy values ​​of these specific coordinates are forcibly set to zero on the mathematical matrix, and a pure power shaft load fluctuation sequence is reconstructed in reverse.

[0045] Furthermore, the specific process for obtaining the transmission ratio topology matrix includes: The system inputs the original mechanical assembly drawings and transmission specifications of the equipment. It then analyzes the power flow from the power input end to the execution end, extracting the tooth count characteristics of each driving and driven wheel and the spatial connection constraints of each transmission shaft. The output is an initial mechanical transmission sequence data containing multi-level fixed reduction ratios and dynamic gear ratios. Next, it inputs the initial mechanical transmission sequence data and performs directed graph matrix construction calculations. Specifically, it maps each physical rotation axis into mathematical topological nodes, maps the physical meshing relationship of gears into topological edges connecting nodes, and assigns multi-level fixed reduction ratios as feature weights to the corresponding topological edges. The output is a transmission ratio topology matrix that accurately reflects the mathematical laws of the gradual change in equipment speed and power splitting. Based on the generated transmission ratio topology matrix and the basic files of the digital twin calculation model, it performs data solidification and addressing association calculations within the model, spatially mapping and binding the input and output nodes in the transmission ratio topology matrix with the virtual power output nodes and virtual execution nodes within the model. Finally, it outputs the underlying dynamic architecture basic data with forward kinematics deduction capabilities.

[0046] This invention utilizes the theoretical velocity state calculated in real time by a digital twin model to proactively deduce the dynamic trajectory of noise frequencies in the time-frequency spectrum. Using the theoretical trajectory in virtual space as an anchoring reference, it cleanses the noisy signals of the real world, achieving deep symbiosis and cross-guidance between virtual and real data. This highly personalized and adaptive signal decomposition method can still stably extract the true load waveform even in extremely harsh and variable mechanical resonance environments, demonstrating excellent robustness.

[0047] Preferably, the process of constructing a three-dimensional impedance distribution topology set by performing error compensation on geological survey data includes: The spatial absolute coordinates and heading angles, which are collected synchronously with the physical operation data of the equipment, are used as spatial position anchoring data. The extracted medium interaction impedance characteristics are timestamped and gridded with the spatial position anchoring data to obtain local impedance observations with precise geographic coordinate labels. Geological survey data, serving as the prior background field, is input into a spatial interpolation network. Based on local impedance observations, the impedance deviation matrix between the observations and the prior background field is calculated using a spatial residual computational mechanism. The impedance deviation matrix is ​​then used to perform adaptive weight correction and error compensation processing on the geological survey data of the unexplored areas ahead, outputting a dynamically assimilated and updated three-dimensional impedance distribution topology set.

[0048] The dielectric inter-impedance characteristics are timestamped with the spatial absolute coordinates transmitted from the positioning satellite receiver and anchored to specific grid points on the virtual map to form local impedance observations. The high-precision local impedance observation value accurately sensed by the current blade is subtracted from the estimated value at the same coordinate point on the prior map, and an impedance deviation matrix is ​​calculated using spatial residual calculation. This calculated impedance deviation matrix is ​​then used as a correction factor to adaptively weight the prior map data for the unworked area directly ahead. Areas closer to the current point receive a greater weight from this deviation correction. This real-time rolling error compensation process outputs a three-dimensional impedance distribution topology set that integrates macroscopic surveying and microscopic measurements.

[0049] This invention integrates high-precision impedance observations with macroscopic geological survey maps, utilizes spatial residual calculation mechanisms to extract deviations, and combines spatial interpolation networks to perform weighted diffusion corrections for previously unreached areas, thus giving historical survey data timeliness. This error compensation strategy based on Bayesian correction allows the three-dimensional impedance distribution topology set to continuously roll and approximate the actual underground physical state as the equipment moves, providing a data support base with continuous predictive capabilities for subsequent advanced identification of impedance abrupt changes.

[0050] Preferably, the process of outputting the virtual resistance peak includes: The spatial gradient change sequence directly in front of the three-dimensional impedance distribution topology set is extracted as the simulation input boundary condition; the spatial gradient change sequence is prospectively scanned and calculated in the digital twin computing model; based on the calculation comparison, when the local impedance value change rate is found to exceed the preset safety threshold, the impedance mutation feature is extracted and output. Based on the impedance mutation characteristics, a local adaptive refinement mechanism for the contact calculation time step is activated to adjust the time span of numerical integration calculation in the three-dimensional mesh space. Driven by the refined time step, the equipment stress state array data within a future preset time period is extrapolated. The collision extreme value is extracted from the equipment stress state array data, and a virtual resistance peak value is output for overload warning.

[0051] The input is a 3D impedance distribution topology set. Starting from the device's current virtual spatial coordinates, the impedance distribution data stored in each consecutively arranged 3D grid node is extracted sequentially along the virtual device's expected trajectory, according to their spatial physical distance. Since simple absolute impedance values ​​cannot intuitively reflect the degree of environmental abrupt changes, spatial difference calculation is performed on the extracted impedance distribution data. Specifically, the impedance value of the next adjacent grid node is subtracted from the impedance value of the next grid node, and the difference is divided by the absolute physical distance between the two grid nodes to calculate an index representing the rate of change of local resistance, i.e., the gradient value. All gradient values ​​calculated along the path are arranged strictly in spatial order, and the final combined output data set is the spatial gradient change sequence.

[0052] The spatial gradient change sequence directly in front of the updated three-dimensional impedance distribution topology set is continuously extracted; when the expected resistance value of two connected grid points is found to rise sharply and exceed the preset safety threshold, the coordinate area is immediately marked and the impedance change feature is extracted.

[0053] In computer simulations, time progresses frame by frame (i.e., a conventional time step). If the impact of a blade striking an extremely hard, abrupt region is directly simulated using the normal time step, the simulation values ​​would explode and collapse due to the high velocity and stiffness at the moment of contact. To avoid this problem and ensure accurate capture of the maximum force at the moment of impact, a local adaptive refinement mechanism for the contact calculation time step is activated upon receiving the impedance abrupt change characteristic. Specifically, to maintain the synchronization of global time integration, a local sub-step loop is activated; the global model maintains the conventional time step, while within the local mesh space where the collision is about to occur, multiple refinement of the target time step is nested within a conventional time step cycle, and the mechanical balance between the local and global processes is achieved through boundary force interpolation. In essence, only within the three-dimensional space where the geometric interference collision is about to occur is the calculation frequency increased, breaking the originally large time step into smaller, more fragmented segments.

[0054] Driven by minute temporal resolution, array data of the device's force state at various minute time points is acquired. Subsequently, the maximum force value is selected from the array data and output as the virtual resistance peak.

[0055] This scheme identifies abrupt changes through forward scanning, shortening the time step to a minute scale only within local mesh regions where high-energy collisions are imminent. This not only avoids the risk of numerical explosion and ensures the accuracy of peak value calculations, but also makes the entire simulation run faster than the actual elapsed time in physical reality. Thus, it enables highly accurate virtual drag peak value warnings to be issued milliseconds in advance, before catastrophic physical damage occurs.

[0056] Further, the process of refining the determination of the time step includes: Obtain the dielectric equivalent stiffness limit value contained in the triggered impedance change characteristics, and simultaneously extract the local equivalent mass value and instantaneous relative collision velocity value of the operation execution end that is about to collide in the digital twin calculation model. Using the equivalent stiffness limit value of the medium and the local equivalent mass value as input variables, the high-frequency vibration boundary solution of the contact system is performed in virtual space, and the critical time step threshold that maintains the absolute stability of the local numerical integral is output. Based on the instantaneous relative collision velocity value, dynamic penalty calculation is performed in combination with the critical time step threshold, and a preset numerical safety factor is introduced for multiplication and degradation processing, and the refined target time step value is output. The target time step value is used to replace the conventional step size in the local three-dimensional mesh space, driving the high-frequency discrete numerical integration derivation of the local region.

[0057] The first step is to extract the physical characteristics and state parameters of the collision; input impedance mutation characteristics and the three-dimensional spatial operation state of the current digital twin computing model, and perform data decomposition and mapping extraction; extract the medium equivalent stiffness limit value representing the true physical hardness of the block in front from the impedance mutation characteristics; at the same time, extract the local equivalent mass value of the blade part that is about to collide with the block from the virtual device model, as well as the instantaneous relative collision velocity value of the blade and the block at the moment of contact.

[0058] The second step involves boundary calculation of the critical time step threshold. The equivalent stiffness limit of the medium and the local equivalent mass value are input into the contact resonance frequency calculation module. In explicit dynamic simulation theory, the instant of contact can be equivalent to a spring oscillator system; the greater the stiffness and the smaller the mass, the higher the vibration frequency. These two values ​​are used to perform high-frequency vibration boundary calculation of the contact system, finding the critical point at which the contact system experiences numerical oscillation collapse. The output data is the critical time step threshold for maintaining absolute stability of the local numerical integral, i.e., the maximum time step limit that the computer can tolerate.

[0059] The third step involves the dynamic convergence and degradation generation of the target time step value. Based on the critical time step threshold and the instantaneous relative collision velocity value, dynamic convergence constraints and safe dimensionality reduction are performed. The magnitude of the instantaneous relative collision velocity value is determined; a faster velocity means a deeper geometric depth of the blade intrusion into the hard block within a single time step, which is more likely to cause calculation errors. Therefore, dynamic penalty calculation is performed on the critical time step threshold in conjunction with the velocity value to further reduce the step size. Subsequently, to leave sufficient absolute safety redundancy space, a preset numerical safety factor (e.g., a parameter between 0.1 and 0.3) is introduced and multiplied with the penalized value for degradation processing. Finally, the refined target time step value is output, and this time value is used to replace the regular step size of the region where a collision is about to occur, ensuring the smooth progress of high-frequency numerical integration.

[0060] This invention introduces three core physical quantities: contact stiffness, local mass, and instantaneous collision velocity. By calculating the contact resonance frequency, it determines the baseline of numerical stability in real time and dynamically generates target time step values. This personalized micro-scheduling mechanism of computing power effectively captures the millisecond-level peak values ​​of real resistance through fine-grained time steps.

[0061] Preferably, the process of generating a device parameter configuration set includes: The virtual resistance peak is compared with the preset theoretical load threshold of the model to calculate the load overflow data that quantifies the overload risk; the single-cut geometric pitch value characterizing the crushing quality is obtained, and the single-cut geometric pitch value is limited to a preset qualified threshold range as a rigid data constraint boundary for optimization calculation; combining the load overflow data and the rigid data constraint boundary, a bivariate differential compensation calculation for the execution end and traction end is performed in the virtual space; the moving speed data that needs to be adjusted is calculated to offset the load overflow, and the equipment rotation speed data that needs to be adjusted is simultaneously calculated to meet the constant geometric pitch condition; based on the calculated equipment rotation speed data and moving speed data, a set of equipment parameter configurations for simulation verification is generated.

[0062] The predicted peak virtual resistance is subtracted from the theoretical load threshold of the limit model that the equipment material can withstand to obtain the amount of load overflow that needs to be reduced. In the principles of agricultural machinery, simply reducing the tractor's traction speed can indeed reduce resistance, but it will cause the soil clods cut by the blades to become finer; while simply reducing the cutter shaft speed will stall the engine. Therefore, a single-cut geometric pitch value representing the quality of soil crushing is introduced (its principle is proportional to the traction speed divided by the equipment speed), and it is forcibly locked within a qualified threshold range as a rigid data constraint boundary for the optimization solution. Among them, the single-cut geometric pitch is used to characterize the quality of crushing, and it is defined as the geometric pitch parameter obtained by the ratio of the equipment's travel distance per unit time to the cutter shaft speed.

[0063] Under the constraints of the boundary conditions, based on bivariate differential compensation, the required reduction in the equipment moving speed is calculated to reduce the overall pushing load on the cutter shaft and offset the aforementioned load overflow. Simultaneously, to compensate for the over-crushing of soil caused by the reduced moving speed, the required adjustment of the cutter shaft rotation speed is calculated to maintain the proportional balance between the numerator and denominator, thereby ensuring the rigid condition of a constant cutting pitch. The obtained equipment rotation speed data and moving speed data are encapsulated and fused to generate a set of equipment parameter configurations that balances overload avoidance with ensuring high-quality crushing.

[0064] The bivariate differential compensation solution includes: a load sensitivity mapping module and a kinematic pitch constant constraint module.

[0065] Load sensitivity mapping module: It contains a load speed sensitivity relationship matrix generated through previous physical bench experiments; this matrix records the corresponding numerical mapping relationship of how much torque load can be reduced by reducing the unit movement speed under different geological compaction conditions.

[0066] The kinematic pitch constant constraint module internally incorporates the single-cutting quantity equation logic from agricultural machinery, namely, the calculation equation for the single-cutting geometric pitch is the quotient of the equipment's moving speed value divided by the equipment's rotational speed value of the cutter shaft.

[0067] Once the load overflow data is identified and bivariate differential compensation calculation is triggered, the data flow and operation steps are as follows: Perform initial compensation calculation for movement speed; input load overflow data and current equipment status parameters; call the load sensitivity mapping module, use the current load overflow data as the target retrieval feature, perform table lookup and linear interpolation calculation in the load speed sensitivity relationship matrix, find and output the movement speed reduction that must be corresponding to offset the load overflow; then, subtract the movement speed reduction from the currently running movement speed data, and calculate and output the movement speed data that needs to be adjusted in the first stage.

[0068] Extract rigid constraint features and construct equations; input the qualified threshold range of the single-cut geometric pitch value that characterizes the crushing quality of the operation; the operation is to extract the target pitch constant (usually the midpoint of the range) from the threshold range as the rigid data constraint boundary, and activate the kinematic pitch constant constraint module based on the target pitch constant to construct the single-cut geometric equation in memory with the equipment rotation speed data as the unknown solution variable.

[0069] Perform differential calculation of the tool axis speed; input the obtained data on the required adjustment of the traverse speed and the constructed single-cutting geometry equation; substitute the reduced traverse speed data as a known numerator parameter into the single-cutting geometry equation; since the quotient of the equation (i.e., the target pitch constant) is rigidly locked, the denominator value that must be reduced proportionally is directly calculated by solving this linear equation; the output data is the tool axis speed data that needs to be reduced synchronously.

[0070] Encapsulation and output of parameter configuration sets; input the calculated moving speed data that needs adjustment and the tool axis rotation speed data that needs to be reduced synchronously; perform time-series alignment and data packet encapsulation on the adjustment data of these two spatial and rotational dimensions; convert and output the device parameter configuration set for simulation verification, and send it to the digital twin model for security verification.

[0071] This solution employs a bivariate differential compensation logic by rigidly embedding the single-cut geometric pitch as the underlying constraint in the virtual solution. When faced with load overflow, it comprehensively considers both factors, simultaneously calculating a strategy to reduce the moving speed to unload pressure and a strategy to increase the equipment speed to compensate for cutting losses. This multi-dimensional collaborative parameter configuration generation mechanism ensures that the output optimized solution not only guarantees the safe obstacle crossing of the physical equipment in terms of dynamics theory but also guarantees high-quality operation in terms of agronomic standards, possessing extremely high industrial transformation and engineering application value.

[0072] Example 2: This invention proposes a dynamic optimization system for equipment operating parameters based on digital twins. The structure of the system is as follows: Figure 3 As shown, it specifically includes: a model building module, a topology analysis module, a virtual simulation module, and a simulation verification module; Model building module: Construct a heterogeneous discrete element particle bed based on geological survey data, and establish a multi-physics field coupled digital twin computational model by combining equipment dynamic parameters; Preferably, the process of establishing a multiphysics coupled digital twin computational model includes: Obtain geological survey data containing soil stratification, compaction, and composition distribution; based on the geological survey data, extract microscopic contact mechanical characteristics corresponding to different depth dimensions; combine the microscopic contact mechanical characteristics to generate virtual particle sets at corresponding depths in virtual space, and combine them to construct a heterogeneous discrete element particle bed with vertical gradient changes. Obtain the device dynamic parameters, which include the device's three-dimensional geometric topology and the constraints of the rotary joints; based on the device dynamic parameters, establish a kinematic assembly model in virtual space that is spatially mapped to the physical device; import the kinematic assembly model into a heterogeneous discrete element particle bed, and perform physical field binding calculations in combination with the set surface contact friction constitutive boundary conditions to output a multi-physics coupled digital twin calculation model.

[0073] The process of obtaining the constitutive boundary conditions for surface contact friction includes: Obtain physical interaction benchmark data between the actual working medium and the corresponding material of the physical equipment. The physical interaction benchmark data includes the macroscopic static friction critical state value and the collision energy recovery ratio value. A virtual constitutive computational framework is constructed in the digital twin computational model for the contact interface between discrete element particles and kinematic assembly. The virtual constitutive computational framework contains the virtual collision restitution coefficient and virtual friction coefficient matrix to be calibrated. A virtual interactive test scenario is established that maps to the physical interaction benchmark data acquisition environment. The virtual collision recovery coefficient and virtual friction coefficient matrix to be calibrated are input into the virtual constitutive calculation framework to drive the discrete element particles and kinematic assemblies to perform virtual collision and sliding deduction calculations. The system compares the virtual interaction state features output from the simulation with the physical interaction benchmark data in real time. It then uses an error-backward optimization mechanism to iteratively update the matrix parameters to be calibrated. When the comparison error decreases to within a preset convergence threshold, the current parameter matrix is ​​locked, and the surface contact friction constitutive boundary conditions, calibrated using real physical data, are output. In other words, the error-backward optimization mechanism, based on gradient descent or genetic algorithms, guides the direction of the next parameter generation, continuously approximating the true values.

[0074] Obtain realistic physical interaction benchmark data: Acquire real soil samples collected from agricultural operation sites, and standard metal templates with material and surface roughness completely consistent with actual agricultural machinery blades. Construct a physical interaction test platform in a physics laboratory; first, perform an inclined plane sliding experiment, placing the real soil sample on an adjustable-angle standard metal template, slowly raising the angle, and recording the critical angle value at which the soil just begins to slide relative to the metal template; then, perform a drop collision experiment, allowing a standard soil block to fall freely from a specific height onto a fixed standard metal template, recording the rebound height after the collision. Output realistic macroscopic physical interaction benchmark values, specifically the macroscopic static friction critical state value and the collision energy recovery ratio value.

[0075] A virtual constitutive computation framework and test scenario are constructed: Virtual particle properties from the digital twin computational model and surface geometric mesh properties of the kinematic assembly are input, and a nonlinear elastic contact algorithm framework is assigned to their contact interface. This framework reserves undetermined variables characterizing energy loss and sliding resistance. Simultaneously, virtual inclined plane sliding test benches and drop collision test benches are built to scale within the computer, constructing a virtual constitutive computation framework containing the virtual collision restitution coefficient and virtual friction coefficient matrices to be calibrated, as well as a virtual interactive test scenario.

[0076] Virtual inversion and parameter solidification are performed: The input parameters are the matrix to be calibrated and the actual macroscopic interaction physical benchmark values. In a virtual interaction test scenario, an optimization algorithm continuously generates multiple sets of different virtual collision recovery coefficients and virtual friction coefficients, driving virtual particles to perform sliding and rebound simulations on a virtual metal template. The sliding angle and rebound height calculated in the virtual simulation are extracted in real time as virtual interaction state features and compared with the actual macroscopic interaction physical benchmark values. An error-backward optimization mechanism guides the direction of the next parameter generation, continuously approaching the true values. When the performance of the virtual simulation is completely consistent with the performance of the real physical experiment (i.e., the error is less than the preset convergence threshold), iteration is immediately stopped and the current parameter combination is locked. The final output data is an extremely accurate surface contact friction constitutive boundary condition, which is directly imported into the digital twin model in the form of a solidified contact matrix to participate in subsequent multiphysics binding calculations.

[0077] This solution is based on real-world physics experiments involving inclined plane sliding and drop collisions. By utilizing virtual scene reproduction and error-backward optimization algorithms, the computer autonomously explores and locks in constitutive parameters that are closest to physical reality. This endows the virtual blade with realistic surface roughness and physical elasticity, ensuring that the output force fluctuation data when calculating the virtual resistance peak matches the compressive strength limit and energy dissipation process of physical machinery in the real world. This significantly improves the scientific basis and engineering reliability of dynamic optimization of equipment parameters.

[0078] Topology Analysis Module: Acquires physical operation data of equipment and inputs it into the virtual analysis module within the digital twin computing model; eliminates frame vibration and gear resonance noise through a time-frequency joint analysis algorithm and extracts the medium interaction impedance characteristics; based on the medium interaction impedance characteristics, combined with numerical inversion calculation and spatial location anchoring, performs error compensation on geological survey data and constructs a three-dimensional impedance distribution topology set; The process of compensating for errors in geological survey data and constructing a three-dimensional impedance distribution topology set includes: The spatial absolute coordinates and heading angles, which are collected synchronously with the physical operation data of the equipment, are used as spatial position anchoring data. The extracted medium interaction impedance characteristics are timestamped and gridded with the spatial position anchoring data to obtain local impedance observations with precise geographic coordinate labels. Geological survey data, serving as the prior background field, is input into a spatial interpolation network. Based on local impedance observations, the impedance deviation matrix between the observations and the prior background field is calculated using a spatial residual computational mechanism. The impedance deviation matrix is ​​then used to perform adaptive weight correction and error compensation processing on the geological survey data of the unexplored areas ahead, outputting a dynamically assimilated and updated three-dimensional impedance distribution topology set.

[0079] The process of determining adaptive weights during adaptive weight correction and error compensation includes: Acquire the precise geographic coordinates of the local impedance observations and the spatial coordinates of each grid node to be corrected in the unoperated area ahead; perform spatial distance calculation to calculate the physical spatial distance between the current observation point and each grid node to be corrected, and output the distance sequence data containing the spatial location gradient; Input the physical space distance value into the preset spatial correlation attenuation module; use the physical space distance value as an independent variable to perform nonlinear attenuation calculation based on the reciprocal of the distance, and calculate and output the set of initial compensation weight parameters corresponding to each grid node to be corrected in front; The initial set of compensation weight parameters is input into the normalization processing module. Within the limited local influence area, global mapping and proportional compression calculations are performed on the initial compensation weight parameters of all grid nodes to be corrected, and the output is a standard adaptive weight allocation matrix with a smooth decrease in spatial gradient.

[0080] Specifically, the calculation and extraction of spatial distance gradients involves inputting the precise geographic coordinates corresponding to the current local impedance observation value, as well as the spatial coordinates of each grid node to be corrected within the untouched area ahead. Spatial Euclidean distance calculation is performed within the spatial interpolation network to quantify and calculate the physical spatial distance between the current observation point and each grid node to be corrected ahead. The output data is a distance sequence containing the spatial location gradient.

[0081] Nonlinear Attenuation Feature Calculation: Input the calculated physical spatial distance value as described above into the preset spatial correlation attenuation module. Perform nonlinear attenuation calculation based on the reciprocal of the distance, using the physical spatial distance value as an independent variable. The underlying logic is as follows: the smaller the physical spatial distance value, the larger the generated initial compensation weight parameter; when the physical spatial distance value exceeds the preset effective influence radius, the set initial compensation weight parameter rapidly approaches zero according to the attenuation curve. The output data is the set of initial compensation weight parameters corresponding to each grid node in front.

[0082] Global weight normalization and allocation matrix generation: The initial compensation weight parameter set generated above is input into the normalization module. Within a defined local influence region, global mapping and proportional compression calculations are performed on the initial compensation weight parameters of all grid nodes. This operation eliminates the magnitude difference in absolute values, ensuring that the sum of the weights of all nodes within this region conforms to predetermined mathematical lower and upper limits. The final output data is a standard adaptive weight allocation matrix with a smooth, decreasing spatial gradient.

[0083] The standard adaptive weight allocation matrix is ​​used as a multiplication factor and calculated with the previously obtained impedance deviation matrix. This ensures that virtual grids closer to the current agricultural machinery blade receive a very high correction ratio, while grids farther away gradually maintain their original survey values. This allows for high-precision assimilation and correction of data in unknown areas ahead, all within the computer. Specifically, the standard adaptive weight allocation matrix is ​​multiplied by the previously obtained impedance deviation matrix to obtain the absolute impedance correction for each grid node. Then, using spatial weighted superposition, this absolute impedance correction is accumulated into the prior map data of the unexplored area directly ahead, completing the adaptive weight correction.

[0084] For example, the spatial interpolation network employs an inverse distance weighted interpolation algorithm. For nonlinear attenuation calculation based on the reciprocal of distance, the underlying quantitative calculation logic is as follows: the specific value of the initial compensation weight parameter is set to a constant 1 divided by a specific power of the physical spatial distance value, wherein the value of the specific power is preferably between 1 and 3; when the physical spatial distance value exceeds the preset effective influence physical radius, the corresponding initial compensation weight parameter is directly set to zero. After normalization, the set of initial compensation weight parameters obtained through this quantitative calculation method can form a standard adaptive weight allocation matrix with a smooth decreasing spatial gradient.

[0085] This invention introduces a nonlinear attenuation calculation mechanism based on the reciprocal of distance and a normalization processing module, which enables the correction of measured deviations to spread smoothly forward in a form that conforms to the law of correlation of natural geographic space. This not only ensures the continuity and smooth transition of the generated three-dimensional impedance distribution topology set in the spatial dimension, but also provides stable, coherent and high-fidelity forward-looking boundary conditions for the subsequent digital twin model to perform advanced overload prevention simulation, thereby improving the control stability of the overall optimization method.

[0086] Virtual simulation module: Using the three-dimensional impedance distribution topology set as the simulation input boundary condition, it performs numerical simulation and simulation. When the impedance change feature is identified, it triggers local adaptive refinement of the contact calculation time step, simulates the equipment stress state within a preset time period in the digital twin calculation model, and outputs the virtual resistance peak. Simulation verification module: Based on the virtual resistance peak, the single cutting geometric pitch is introduced as a rigid data constraint within the threshold range. Bivariate differential compensation calculation is performed in the virtual space to generate a set of equipment parameter configurations for simulation verification. The set of equipment parameter configurations includes simulation configuration data of equipment rotation speed and moving speed.

[0087] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for dynamic optimization of equipment operating parameters based on digital twins, characterized in that, include: Based on geological survey data, a heterogeneous discrete element particle bed is constructed, and combined with equipment dynamic parameters, a digital twin computational model with multi-physics field coupling is established. The process of establishing a multiphysics coupled digital twin computational model includes: Obtain geological survey data containing soil stratification, compaction, and composition distribution; based on the geological survey data, extract microscopic contact mechanical characteristics corresponding to different depth dimensions; combine the microscopic contact mechanical characteristics to generate virtual particle sets at corresponding depths in virtual space, and combine them to construct a heterogeneous discrete element particle bed with vertical gradient changes. Obtain the device dynamic parameters, which include the device's three-dimensional geometric topology and the constraints of the rotary joints; based on the device dynamic parameters, establish a kinematic assembly model in virtual space that is spatially mapped to the physical device; import the kinematic assembly model into a heterogeneous discrete element particle bed, and perform physical field binding calculations in combination with the set surface contact friction constitutive boundary conditions to output a multi-physics coupled digital twin calculation model. The physical operation data of the equipment is acquired and input into the virtual analysis module in the digital twin computing model; the frame vibration and gear resonance noise are eliminated by the time-frequency joint analysis algorithm, and the medium interaction impedance characteristics are extracted; based on the medium interaction impedance characteristics, combined with numerical inversion solution and spatial location anchoring, error compensation is performed on the geological survey data, and a three-dimensional impedance distribution topology set is constructed. The process of extracting the dielectric cross-impedance characteristics includes: acquiring equipment physical operation data containing mechanical load values ​​and high-frequency vibration characteristics; inputting the equipment physical operation data into a virtual analysis module; the virtual analysis module performing signal frequency band decomposition analysis on the equipment physical operation data based on a preset time-frequency joint analysis algorithm, identifying and separating low-frequency rack turbulence interference signals and high-frequency gear resonance interference signals, reconstructing and outputting the power shaft load fluctuation sequence; based on the power shaft load fluctuation sequence, combined with the transmission efficiency attenuation mapping relationship in the digital twin calculation model, performing inverse dynamic solution, filtering out the mechanical loss characteristics inside the model, and extracting the dielectric cross-impedance characteristics caused by external collisions; Using the three-dimensional impedance distribution topology set as the simulation input boundary condition, numerical simulation is performed. When impedance change characteristics are identified, the contact calculation time step is locally adaptively refined. The equipment stress state within a preset time period is deduced in the digital twin calculation model, and the virtual resistance peak is output. Based on the virtual resistance peak, the single cutting geometric pitch is introduced as a rigid data constraint within a threshold range. Bivariate differential compensation is performed in the virtual space to generate a set of equipment parameter configurations for simulation verification. The set of equipment parameter configurations includes simulation configuration data for equipment rotation speed and moving speed.

2. The method for dynamic optimization of equipment operating parameters based on digital twins according to claim 1, characterized in that: The process of analyzing signal frequency bands from equipment physical operation data based on the time-frequency joint analysis algorithm includes: The physical operation data of the equipment is used as the initial time-domain sequence input to the time-frequency conversion calculation network, and the initial time-frequency energy spectrum data containing dual features of time and frequency dimensions is output through transformation and calculation. Extract the theoretical equipment rotation speed and theoretical moving speed output by the digital twin calculation model at the current moment; combine the transmission ratio topology matrix in the digital twin calculation model to perform forward kinematics deduction calculation on the theoretical equipment rotation speed, and generate high-frequency gear meshing characteristic trajectory data that dynamically moves with the farmland load; simultaneously combine the geometric features of farmland surface undulation in virtual space to perform low-frequency response solution on the theoretical moving speed, and generate low-frequency frame sway characteristic trajectory data that dynamically fluctuates with the travel speed. The generated high-frequency gear meshing characteristic trajectory data and low-frequency frame turbulence characteristic trajectory data are used as dynamic anchoring references and projected onto the initial time-frequency energy spectrum data to perform spatial intersection matching calculation; the dynamic energy convergence band that overlaps with the characteristic trajectory data in the spectrum data is extracted and identified as high-frequency gear resonance interference signal and low-frequency frame turbulence interference signal. An adaptive energy masking calculation is performed on the regions containing interference signals in the initial time-frequency energy spectrum data, and the output is a power shaft load fluctuation sequence after filtering out dynamic interference energy.

3. The method for dynamic optimization of equipment operating parameters based on digital twins according to claim 1, characterized in that: The process of compensating for errors in geological survey data and constructing a three-dimensional impedance distribution topology set includes: The spatial absolute coordinates and heading angles, which are collected synchronously with the physical operation data of the equipment, are used as spatial position anchoring data. The extracted medium interaction impedance characteristics are timestamped and gridded with the spatial position anchoring data to obtain local impedance observations with precise geographic coordinate labels. Geological survey data, serving as the prior background field, is input into a spatial interpolation network. Based on local impedance observations, the impedance deviation matrix between the observations and the prior background field is calculated using a spatial residual computational mechanism. The impedance deviation matrix is ​​then used to perform adaptive weight correction and error compensation processing on the geological survey data of the unexplored areas ahead, outputting a dynamically assimilated and updated three-dimensional impedance distribution topology set.

4. The method for dynamic optimization of equipment operating parameters based on digital twins according to claim 1, characterized in that: The process of outputting the virtual resistance peak includes: The spatial gradient change sequence directly in front of the three-dimensional impedance distribution topology set is extracted as the simulation input boundary condition; the spatial gradient change sequence is prospectively scanned and calculated in the digital twin computing model; based on the calculation comparison, when the local impedance value change rate is found to exceed the preset safety threshold, the impedance mutation feature is extracted and output. Based on the impedance mutation characteristics, a local adaptive refinement mechanism for the contact calculation time step is activated to adjust the time span of numerical integration calculation in the three-dimensional mesh space. Driven by the refined time step, the equipment stress state array data within a future preset time period is extrapolated. The collision extreme value is extracted from the equipment stress state array data, and a virtual resistance peak value is output for overload warning.

5. The method for dynamic optimization of equipment operating parameters based on digital twins according to claim 1, characterized in that: The process of outputting the virtual resistance peak includes: The spatial gradient change sequence directly in front of the three-dimensional impedance distribution topology set is extracted as the simulation input boundary condition; the spatial gradient change sequence is prospectively scanned and calculated in the digital twin computing model; based on the calculation comparison, when the local impedance value change rate is found to exceed the preset safety threshold, the impedance mutation feature is extracted and output. Based on the impedance mutation characteristics, a local adaptive refinement mechanism for the contact calculation time step is activated to adjust the time span of numerical integration calculation in the three-dimensional mesh space. Driven by the refined time step, the equipment stress state array data within a future preset time period is extrapolated. The collision extreme value is extracted from the equipment stress state array data, and a virtual resistance peak value is output for overload warning.

6. A dynamic optimization system for equipment operating parameters based on digital twins, characterized in that, include: Model building module: Construct a heterogeneous discrete element particle bed based on geological survey data, and establish a multi-physics field coupled digital twin computational model by combining equipment dynamic parameters; The process of establishing a multiphysics coupled digital twin computational model includes: Obtain geological survey data containing soil stratification, compaction, and composition distribution; based on the geological survey data, extract microscopic contact mechanical characteristics corresponding to different depth dimensions; combine the microscopic contact mechanical characteristics to generate virtual particle sets at corresponding depths in virtual space, and combine them to construct a heterogeneous discrete element particle bed with vertical gradient changes. Obtain the device dynamic parameters, which include the device's three-dimensional geometric topology and the constraints of the rotary joints; based on the device dynamic parameters, establish a kinematic assembly model in virtual space that is spatially mapped to the physical device; import the kinematic assembly model into a heterogeneous discrete element particle bed, and perform physical field binding calculations in combination with the set surface contact friction constitutive boundary conditions to output a multi-physics coupled digital twin calculation model. Topology Analysis Module: Acquires physical operation data of equipment and inputs it into the virtual analysis module within the digital twin computing model; eliminates frame vibration and gear resonance noise through a time-frequency joint analysis algorithm and extracts the medium interaction impedance characteristics; based on the medium interaction impedance characteristics, combined with numerical inversion calculation and spatial location anchoring, performs error compensation on geological survey data and constructs a three-dimensional impedance distribution topology set; The process of extracting the dielectric cross-impedance characteristics includes: acquiring equipment physical operation data containing mechanical load values ​​and high-frequency vibration characteristics; inputting the equipment physical operation data into a virtual analysis module; the virtual analysis module performing signal frequency band decomposition analysis on the equipment physical operation data based on a preset time-frequency joint analysis algorithm, identifying and separating low-frequency rack turbulence interference signals and high-frequency gear resonance interference signals, reconstructing and outputting the power shaft load fluctuation sequence; based on the power shaft load fluctuation sequence, combined with the transmission efficiency attenuation mapping relationship in the digital twin calculation model, performing inverse dynamic solution, filtering out the mechanical loss characteristics inside the model, and extracting the dielectric cross-impedance characteristics caused by external collisions; Virtual simulation module: Using the three-dimensional impedance distribution topology set as the simulation input boundary condition, it performs numerical simulation and simulation. When the impedance change feature is identified, it triggers local adaptive refinement of the contact calculation time step, simulates the equipment stress state within a preset time period in the digital twin calculation model, and outputs the virtual resistance peak. Simulation verification module: Based on the virtual resistance peak, the single cutting geometric pitch is introduced as a rigid data constraint within the threshold range. Bivariate differential compensation calculation is performed in the virtual space to generate a set of equipment parameter configurations for simulation verification. The set of equipment parameter configurations includes simulation configuration data of equipment rotation speed and moving speed.

Images