A load calculation method for excavators based on digital twinning and Kalman filtering

Abstract

The application provides a kind of based on digital twin and Kalman filtering's excavator load calculation method, it is related to engineering machinery technical field, the method includes: obtaining the multi-source sensor data of target excavator, the multi-source sensor data includes the drive pressure of each hydraulic cylinder, displacement data and body attitude data;Based on the displacement data, the joint motion state is solved, and the measured joint torque is calculated according to the drive pressure and the current joint configuration;Based on the joint motion state and the preset dynamics parameters, the inertia joint torque of system itself is calculated, and the measured joint torque is subtracted from the inertia joint torque, to obtain the net load torque;A state space model with load mass and lumped disturbance as state vector is constructed, and the net load torque is used as observation variable, recursive optimal estimation is carried out by using Kalman filter, to obtain the target load mass estimation value of the target excavator.

Description

Technical Field

[0001] This invention relates to the field of engineering machinery technology, and in particular to a method for calculating excavator load based on digital twin and Kalman filtering. Background Technology

[0002] With the development of industrial processes, excavators are widely used in various high-risk environments and complex construction sites, and their level of intelligence is closely related to their operational efficiency. Current technologies typically employ methods such as directly installing sensors or using models based on radial basis function neural networks and gated recurrent unit neural networks to collect signals such as pressure and flow for indirect prediction. In addition, hydraulic load simulation systems based on PID control are also used in experimental setups to reproduce load pressure.

[0003] However, the aforementioned existing technologies still have many limitations. First, the system closed loop is fragmented; existing sensing and prediction outputs are mostly isolated signals, failing to form a real-time control closed loop with bidirectional interaction with the overall machine control. Second, the model's generalization and adaptive capabilities are weak; data-driven models heavily rely on training samples, while mechanistic models depend on fixed parameters, both of which struggle to cope with time-varying characteristics such as unknown materials, complex postures, and system wear. Third, there is a lack of multidisciplinary collaborative digital simulation support; high-fidelity digital twin models have not been established, making it impossible to conduct accurate working condition simulations and strategy verification in virtual space. Finally, the load decoupling accuracy is insufficient; traditional algorithms cannot fundamentally separate the material gravity effect from the dynamic resistance of excavation, resulting in high noise, high latency, and poor robustness in load estimates during dynamic operations. Summary of the Invention

[0004] To address the aforementioned technical problems, according to a first aspect of the embodiments of this application, a method for calculating excavator load based on digital twin and Kalman filtering is provided. The method includes: Acquire multi-source sensor data of the target excavator, including the driving pressure and displacement data of each hydraulic cylinder and the body posture data; The joint motion state is calculated based on the displacement data, and the measured joint torque is calculated based on the driving pressure and the current joint configuration. Based on the joint motion state and preset dynamic parameters, the system's own inertial joint torque is calculated, and the measured joint torque is subtracted from the inertial joint torque to obtain the net load torque; A state-space model with load mass and lumped disturbance as state vectors is constructed, and the net load torque is used as the observation variable. A Kalman filter is used to perform recursive optimal estimation to obtain the target load mass estimate of the target excavator.

[0005] This solution enables direct, accurate, and online perception of the actual load weight inside the bucket, overcoming the limitations of traditional fixed-parameter models and achieving a high-precision load estimation closed loop under all working conditions.

[0006] In one embodiment, the step of calculating the joint motion state based on the displacement data and calculating the measured joint torque according to the driving pressure and the current joint configuration includes: Using a pre-defined inverse kinematics model of the working device, the displacement data is converted into joint angle vectors to determine the real-time configuration of the robotic arm; Numerical difference and filtering are performed on the joint angle vector sequence to obtain the joint angular velocity and angular acceleration; The pressure difference between the two chambers of each hydraulic cylinder and the effective area of ​​the piston are obtained, and the instantaneous lever arm determined by the real-time configuration is used to calculate the measured joint torque that drives each joint.

[0007] This solution can accurately convert the underlying basic displacement and pressure physical quantities into torque characteristic information required for real-time dynamic calculations of the system.

[0008] In one embodiment, calculating the system's own inertial joint torque based on the joint motion state and preset dynamic parameters includes: The recursive Newton-Euler algorithm is used to calculate the angular velocity, angular acceleration and center of mass acceleration of each link from the base to the end of the target excavator. The internal forces are recursively calculated from the end to the base, and the inertial force and inertial torque of each link are calculated. Based on the balance relationship between force and torque, the inertial joint torque used purely to overcome its own inertia and gravity is calculated.

[0009] This solution enables efficient and real-time isolation of the system's own no-load dynamic effects, providing a high-quality net torque reference for subsequent accurate decoupling of external loads.

[0010] In one implementation, constructing a state-space model with load quality and lumped disturbance as state vectors includes: The resistance effect during the material excavation process is modeled as the lumped disturbance acting on each joint; The lumped disturbance and the load mass are jointly defined as the state vector to construct the state equation; Based on the kinematic Jacobian matrix, the gravitational torque caused by the load mass is mapped to the joint space and superimposed with the lumped perturbation to construct the observation equation corresponding to the net load torque.

[0011] This solution effectively isolates and constructs a system from the mechanical mechanism of pure material gravity and environmental shear friction resistance.

[0012] In one implementation, prior to constructing the observation equation corresponding to the net load torque, the method further includes: Based on the fuselage attitude data and the current joint configuration, the kinematic Jacobian matrix is ​​dynamically updated to adaptively adjust the gravity mapping relationship according to the excavator's three-dimensional spatial attitude changes.

[0013] This solution enables accurate calculation of gravity mapping in any posture, such as horizontal, inclined, or rotating, achieving full-attitude adaptive perception without the need for external compensation.

[0014] In one implementation, the step of using a Kalman filter to perform recursive optimal estimation to obtain the target load mass estimate of the target excavator includes: Within each calculation cycle, based on the state estimate from the previous time step, the prior estimate of the current state and the prior error covariance matrix are predicted. Calculate the observation residual between the measured net load torque and the predicted observation value for the current period; Calculate the Kalman gain matrix based on the prior error covariance matrix and the preset observation noise covariance matrix; The prior estimate is corrected using the Kalman gain matrix and the observation residual to obtain the posterior estimate of the state vector; The first element is extracted from the posterior estimate and used as the target load quality estimate after full-process correction and compensation.

[0015] This solution can act as an intelligent decoupler, dynamically separating the gravity effect within the mixed signal in the optimal way, and outputting load estimation results with extremely strong anti-interference capabilities.

[0016] In one embodiment, after obtaining the target load mass estimate of the target excavator, the method further includes: The multi-source sensor data and the target load mass estimate are synchronously transmitted to a pre-constructed three-dimensional digital twin model through a preset data interface; The virtual excavator is driven to respond synchronously in the three-dimensional digital twin model, and the estimated value of the target load quality and the corresponding risk warning information are superimposed and displayed on the interactive interface of the virtual scene.

[0017] This solution integrates physical sensing with virtual simulation environments, providing operators with intuitive, high-fidelity real-time status feedback and assisting in data-driven operational insights and decision-making.

[0018] In one implementation, after obtaining the posterior estimate of the state vector, the method further includes: Obtain the posterior estimation error covariance matrix of the Kalman filter output; Based on the posterior estimation error covariance matrix, the standard deviation of the target load quality estimate is calculated in real time, and a dynamic confidence interval is constructed. When the upper limit of the dynamic confidence interval reaches a preset warning threshold, a first-level warning signal is triggered; When the lower limit of the dynamic confidence interval exceeds the preset safety limit threshold, an overload condition is determined and a second-level alarm signal is triggered.

[0019] This solution introduces a quantitative assessment of safety risk boundaries for excavator systems, greatly enhancing proactive safety assurance capabilities under complex working conditions.

[0020] In one embodiment, the method further includes: Continuously monitor the observation residual sequence of the Kalman filter within a continuous sliding window; Calculate the statistical characteristics of the observed residual sequence within the sliding window, the statistical characteristics including the residual mean and covariance; In response to the residual mean deviating from zero by more than a first threshold, or the increase in covariance exceeding a second threshold, the system determines that there is a model mismatch and generates an early fault diagnosis prompt.

[0021] This solution can identify weak signals of underlying model mismatch before traditional failures occur, providing high-level quantitative basis for predictive maintenance of the entire system.

[0022] In one embodiment, acquiring the multi-source sensor data of the target excavator further includes: Acquire high-density 3D point cloud data of the work scene collected by a depth camera at a fixed frequency; The three-dimensional terrain of the work area is reconstructed in real time by integrating the three-dimensional point cloud data, and the terrain undulation, material accumulation outline and spatial distance information are dynamically updated in the corresponding virtual scene.

[0023] This solution can enhance the excavator's comprehensive understanding of the macroscopic morphology of the working environment and improve the realism and spatial matching of the environment simulation in the digital space.

[0024] In a second aspect, embodiments of this specification provide an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method as described in any of the first aspects.

[0025] Thirdly, embodiments of this specification provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method as described in any of the first aspects. Attached Figure Description

[0026] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of this application. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0027] Figure 1 This is a flowchart illustrating the excavator load calculation method based on digital twin and Kalman filtering according to Embodiment 1 of this application. Figure 2 This is a flowchart illustrating the excavator load calculation method based on digital twin and Kalman filtering according to Embodiment 2 of this application; Figure 3 An overall framework diagram of an excavator intelligent integrated system based on multidisciplinary digital twins provided for embodiments of this application; Figure 4 This is a diagram showing the arrangement of multi-source sensors provided in an embodiment of this application. Figure 5 This is a schematic diagram of the load calculation process provided in the embodiments of this application; Figure 6 A multidisciplinary integrated system architecture diagram provided for embodiments of this application; Figure 7 A schematic diagram illustrating the architecture and processing flow of the VR module provided in this application embodiment; Figure 8 This is a schematic diagram of the structure of an electronic device provided for the implementation of this specification. Detailed Implementation

[0028] Unless otherwise defined, the technical or scientific terms used in the embodiments of this specification shall have the ordinary meaning understood by one skilled in the art to which this specification pertains. The terms "first," "second," and similar terms used in the embodiments of this specification do not indicate any order, quantity, or importance, but are merely used to avoid confusion of constituent elements.

[0029] Unless the context otherwise requires, throughout this specification, "a plurality of" means "at least two," and "including" is interpreted as open-ended or encompassing, that is, "including, but not limited to." In the description of this specification, terms such as "one embodiment," "some embodiments," "exemplary embodiment," "example," "specific example," or "some examples" are intended to indicate that a particular feature, structure, material, or characteristic associated with that embodiment or example is included in at least one embodiment or example of this specification. The illustrative representations of the above terms do not necessarily refer to the same embodiment or example.

[0030] The technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. Based on the embodiments in this specification, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this specification.

[0031] As mentioned in the background section, with the development of industrial processes, excavators are widely used in various high-risk environments and complex construction sites, and their level of intelligence is closely related to their operational efficiency. Existing technologies typically employ methods such as directly installing sensors or using models based on radial basis function neural networks, gated recurrent unit neural networks, etc., to collect signals such as pressure and flow for indirect prediction. In addition, hydraulic load simulation systems based on PID control are also used in experimental setups to reproduce load pressure.

[0032] However, the aforementioned existing technologies still have the following technical problems: First, the system closed loop is fragmented. Existing predictive or PID simulation outputs are mostly isolated values, lacking a real-time, bidirectional intelligent decision-making closed loop with the overall machine control. Second, the model generalization and adaptability are limited. Data-driven methods rely on training data and have weak generalization ability, while mechanistic models rely on fixed parameters and are difficult to cope with time-varying characteristics such as system wear and oil temperature. Third, there is a lack of high-fidelity, multi-disciplinary collaborative digital support, making it impossible to accurately reproduce and verify optimization potential in virtual space in advance. Finally, traditional methods cannot fundamentally and optimally separate the material gravity effect from dynamic resistance, resulting in high noise and high latency in the estimated values, and insufficient robustness during severe operations.

[0033] Based on the above inventive concept, the excavator load calculation method based on digital twin and Kalman filtering provided in this specification is described below as an example.

[0034] This embodiment provides a method for calculating excavator load based on digital twin and Kalman filtering, such as... Figure 1 As shown, it includes: S101: Acquire multi-source sensor data of the target excavator.

[0035] According to the embodiments of this application, the multi-source sensor data refers to any signal source that can acquire the physical state of the excavator, without being limited to a specific type, and can broadly cover sensor signals that acquire the driving pressure, displacement data and body posture data of each hydraulic cylinder.

[0036] S102: Calculate the joint motion state based on the displacement data, and calculate the measured joint torque according to the driving pressure and the current joint configuration.

[0037] S103: Based on the joint motion state and preset dynamic parameters, calculate the system's own inertial joint torque, and subtract the measured joint torque from the inertial joint torque to obtain the net load torque.

[0038] According to the embodiments of this application, the net load torque refers to the pure external torque after stripping the excavator system's own no-load dynamic effects, and mainly includes two physical characteristics: the effect caused by the gravity of the material in the bucket and the complex interactive resistance such as shearing and friction generated during the material excavation process.

[0039] S104: Construct a state-space model with load mass and lumped disturbance as state vectors, and use the net load torque as the observation variable to perform recursive optimal estimation using a Kalman filter to obtain the target load mass estimate of the target excavator.

[0040] According to the embodiments of this application, the lumped disturbance refers to the sum of all nonlinear resistances that are difficult to express with precise mathematical mechanism formulas, and can cover comprehensive disturbance terms such as soil shear force, friction force and friction within the unmodeled system; while the obtained target load mass estimate is the pure material gravity component separated by Kalman filtering mathematical dynamics. This estimate does not produce significant noise or error drift with changes in the excavator's horizontal, slope or slewing postures.

[0041] like Figure 2 The diagram shown is a flowchart illustrating the excavator load calculation method based on digital twin and Kalman filtering provided in this embodiment. Based on Embodiment 1, and considering a real-world engineering scenario, the detailed steps of this embodiment include: S201: Acquire multi-source sensor data and environmental point cloud.

[0042] Specifically, the multi-source sensor data can be implemented as pressure sensor data from the boom / stick / bucket cylinders, magnetostrictive displacement sensor data, three-axis data from the inertial navigation system (IMU), and dual-axis tilt sensor data, etc. (Refer to...) Figure 4In the actual system, pressure sensors are installed in the inlet and return oil circuits of the hydraulic cylinders of the boom, stick, and bucket to monitor the drive pressure in real time. Simultaneously, magnetostrictive displacement sensors are integrated at the piston rod ends of each cylinder to synchronously measure the extension and retraction of the hydraulic rod. Dynamic and kinematic formulas are used to calculate the forces and poses of each joint. An IMU is installed on the upper frame to output three-axis acceleration and angular velocity in real time. Dual-axis tilt sensors are placed at key locations on the chassis to continuously acquire the longitudinal and lateral tilt angles of the machine body relative to the horizontal plane. The fusion of data from the two sensors reconstructs the three-dimensional attitude of the machine body in the geodetic coordinate system. Furthermore, a depth camera is installed on the top of the cab to collect high-density three-dimensional point cloud data of the working scene at a fixed frequency, acquiring real-time information on terrain undulations, material accumulation contours, and spatial distances. All sensors are synchronously acquired and transmitted via a high-speed industrial bus, forming a complete, low-latency dynamic state perception system.

[0043] S202: Solve for motion state and measured torque.

[0044] Reference Figure 5 The load solving module will collect the pressure of the two chambers of the hydraulic cylinder in real time. p i,A , p i,B With piston rod displacement l i ( i =1, 2, 3) constitute the original input of the algorithm. The inverse kinematics model of the working device is then used. , displacement l i Convert to joint angle vector q=[ q 1, q 2, q [3] is used to describe the real-time configuration of the robotic arm. The joint angular velocities are obtained by numerically differencing and filtering the joint angle sequence. With angular acceleration This allows for a complete representation of the system's motion state. Simultaneously, the cylinder pressure difference and the effective piston area... A i The product of these factors, multiplied by the instantaneous lever arm determined by the current configuration q. r i (q) gives the measured torque driving each joint. .

[0045] S203: Calculate the system's own inertial joint torque and extract the net load torque.

[0046] Specifically, the net load torque can be realized as a pure external torque after stripping away the excavator system's own no-load dynamic effects. However, the measured torque... τ measIt is the resultant torque that drives the machine's movement and overcomes the external load. To extract the component related only to the external load, the dynamic effects of the system itself (under no-load) must be isolated. (Continue referring to...) Figure 5 The solution process employs the efficient recursive Newton-Euler algorithm. This process recursively calculates the angular velocities of each link from the base to the ends. ω i angular acceleration and center of mass acceleration Its recursive formula depends closely on the calculated q, , and the mass of each link m i Inertial Tensor I i Parameters such as these are then calculated. Subsequently, the internal forces are recursively calculated from the end to the base, determining the inertial force and moment of each link. Through the balance of forces and moments, the joint torque required to generate the current motion, purely for overcoming its own inertia and gravity, is finally determined. Subtracting this portion from the measured total torque yields the net load torque. This net torque... It is the direct signal source for subsequent load estimation. It mainly consists of two parts: one is the effect caused by the gravity of the material in the bucket, and the other is the resistance generated by the complex interaction of shearing, friction and other factors during the material excavation process.

[0047] S204: State-space model construction and Jacobian matrix update.

[0048] Specifically, the lumped disturbance can be implemented as a comprehensive disturbance term encompassing soil shear force, material friction, and unmodeled internal system friction. To address this, [further details are needed]. The pure gravity component of the material is accurately separated, and the material resistance effect is modeled as a "lumped disturbance" acting on each joint. and with load quality m They are collectively defined as state vectors. A state-space model is constructed to describe the relationships between these variables. The state equation models the state as slowly changing within adjacent periods. The observation equation establishes the relationship between the state and the measurements. The physical relationship between them is as follows: the observed net torque equals the theoretical joint torque generated by the load gravity mapped through the Jacobian matrix, plus the lumped perturbation. Where, v k For observation noise, R is its covariance matrix. The gravitational torque is represented by the kinematic Jacobian matrix. Mapping calculation. Due to the Jacobian matrix The gravity mapping is automatically updated as the excavator's posture changes (horizontal, inclined, or slewing), ensuring consistently accurate mapping and achieving full posture adaptation without any external posture compensation. Therefore, the following state-space model can be constructed to describe the relationships between these variables. The state equations and observation equations are modeled as follows: ; ; Among them, W k For process noise; v k For observation noise, R is its covariance matrix; H k The observation matrix is ​​composed of the adaptively updated kinematic Jacobian matrix and the identity matrix, i.e. Due to the Jacobian matrix The algorithm automatically updates as the excavator's posture changes (horizontal, ramp, or slewing), ensuring that the gravity mapping remains accurate and achieving full posture adaptation without any external posture compensation.

[0049] S205: Recursive optimal estimation using a Kalman filter.

[0050] Specifically, the estimated target load mass can be realized as the mass value corresponding to the pure material gravity component, which is mathematically and dynamically separated through Kalman filtering. This value does not exhibit significant noise or error drift with changes in the excavator's posture (such as on a slope or during rotation). For this linear system, a Kalman filter is used for recursive optimal estimation. The filter performs prediction and updating within each calculation cycle: the prediction step extrapolates the current state based on the previous estimate. The update step then utilizes the currently measured net torque. τ load,k The residual between the observed and the predicted values ​​is obtained through the optimal gain K. k The state estimate is corrected. Gain K k Based on the uncertainty covariance P of the state estimate k|k-1 The observation noise covariance R is dynamically calculated. Specifically, the core recursive solution formula for the Kalman filter is as follows: Prediction Step: ;; Update steps: ; ; ; Among them, K k The Kalman gain matrix determines the weight of the observations in the state update. For the prior estimate (predicted value) and posterior estimate (updated value) of the state; P k|k-1 Pk R represents the prior and posterior values ​​of the state estimation error covariance matrix; V is the observation noise. k The covariance matrix. In this closed-loop process, the filter acts as an intelligent "real-time decoupler." It continuously utilizes the latest sensor information to collaboratively adjust the load quality... m and disturbance d i This estimation mathematically achieves the dynamic and optimal separation of gravitational effects from the mixed net torque signal. Finally, at the end of each millisecond-level computation cycle, the updated state vector x is used to estimate the gravitational effects. k Extracting the first element yields the optimal estimate of the bucket load weight after full-process correction and compensation. .

[0051] S206: VR Twin Interaction and Health Early Warning Closed Loop.

[0052] Reference Figure 6 and Figure 7 Specifically, the three-dimensional digital twin model includes a VR module built on Unity3D software and a multidisciplinary simulation module built on Sysplorer software.

[0053] In the construction of the multidisciplinary simulation module, based on the Modelica standardized modeling language and object-oriented non-causal modeling methods, seamless integration and collaborative solution of models from various disciplines at the energy flow and signal flow levels are achieved. The model is constructed by deeply integrating mechanistic models from four major fields: establishing a mechanical model with mass, center of mass, and inertia that supports rigid-flexible coupling analysis; establishing a nonlinear hydraulic system model including a main pump, main valve, cylinder, and motor based on the hydraulic schematic diagram, and physically coupling it with the mechanical model; constructing an engine dynamic model based on the vehicle power library that can dynamically calculate the output speed and torque according to the throttle opening; and developing a control model that receives sensor signals and runs multi-level algorithms.

[0054] The VR module not only integrates mechanical kinematic constraints and the PhysX physics engine (including mass and collider properties), but also employs high-resolution PBR materials, dynamic lighting and shadows, and surface shaders to enhance visual realism. At the dynamic scene level, the VR module integrates depth camera point clouds to reconstruct terrain undulations and earthwork volume in real time, utilizes particle systems to simulate excavation dust and soil splashes, and uses texture displacement technology to present real-time changes in ground indentations and trench morphology. Combined with weather simulation, it achieves a realistic presentation under multiple working conditions. At the interaction and data flow level, the interactive interface integrates a dashboard, a 3D floating panel, historical curves, and a core control panel. Its underlying implementation features a virtual-real synchronization interface based on the System.IO.Ports.SerialPort serial port for interaction with the physical excavator, and a co-simulation interface based on the MQTT protocol to establish a bidirectional data channel with Sysplorer.

[0055] In the specific early warning and feedback process, the dynamic confidence interval can be implemented as a reliability range for quantifying the weight estimate. Standard deviation σ m,k The covariance matrix output by the Kalman filter is calculated in real time. This is a dynamic risk quantification mechanism based on probability statistics, rather than a static threshold. Specifically, the warning / alarm signal can enable the linkage control system to perform active safety control actions such as power limiting, automatic unloading of the hydraulic system, or forced shutdown, integrating electromechanical and hydraulic systems. The estimated error covariance matrix P output by the filter is... k Includes estimated values Confidence level information. This is the load weight value. Estimation error covariance matrix P k and observation residuals An intelligent analysis system that transmits data in real-time to the VR module. This intelligent analysis system utilizes the estimation error covariance matrix P. k Calculate the standard deviation of the estimated load weight in real time. σ m,k And construct dynamic confidence intervals. When the upper limit of the confidence interval A yellow alert will be triggered when the preset warning threshold is reached; when the lower limit of the confidence interval is reached... Exceeding safety limits M max If the system detects an overload with a high degree of confidence, a red alarm will be triggered, and the control system may be activated to limit power as needed. During this process, risk warnings will be provided through visual prompts such as color flashing of specific parts of the 3D model in the VR module and the overlay of warning icons.

[0056] Furthermore, the intelligent analysis system embedded in the VR module performs backtesting analysis on data within a time window to estimate key performance indicators such as single-bucket load, cycle time, and unit fuel consumption in real time, generating a concise work efficiency dashboard that is overlaid on the interactive interface of the 3D scene. Simultaneously, based on the current work stage, the system dynamically calculates and provides immediate operational suggestions and energy efficiency optimization prompts based on the current state through a built-in rule engine and lightweight optimization algorithms. On the other hand, the intelligent analysis system continuously monitors the observation residual sequence of the Kalman filter. υ k The statistical characteristics enable early fault diagnosis. When the system is healthy, υ k It should be zero-mean white noise. The system calculates the mean and covariance within its sliding window. If the residual mean is detected to be continuously deviating from zero or the covariance is abnormally increasing, it can be determined as model mismatch, thus indicating potential early faults or abnormal operating conditions such as sensor zero-point drift, hydraulic system internal leakage, or unforeseen load impacts, providing a quantitative basis for predictive maintenance (early fault diagnosis can be implemented as potential sensor zero-point drift, hydraulic system internal leakage, or unforeseen load impact indications).

[0057] In one exemplary embodiment of this specification, an electronic device is also provided, such as Figure 8 As shown, the electronic device may include: a processor 310, a communication interface 320, a memory 330, and a communication bus 340, wherein the processor 310, the communication interface 320, and the memory 330 communicate with each other via the communication bus 340. The processor 310 can call logical instructions in the memory 330 to execute an excavator load calculation method based on digital twin and Kalman filtering, the method including: Acquire multi-source sensor data of the target excavator, including the driving pressure and displacement data of each hydraulic cylinder and the body posture data; The joint motion state is calculated based on the displacement data, and the measured joint torque is calculated based on the driving pressure and the current joint configuration. Based on the joint motion state and preset dynamic parameters, the system's own inertial joint torque is calculated, and the measured joint torque is subtracted from the inertial joint torque to obtain the net load torque; A state-space model with load mass and lumped disturbance as state vectors is constructed, and the net load torque is used as the observation variable. A Kalman filter is used to perform recursive optimal estimation to obtain the target load mass estimate of the target excavator.

[0058] Furthermore, the logical instructions in the aforementioned memory 330 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0059] In addition to the methods, apparatus, and devices described above, the excavator load calculation method based on digital twin and Kalman filtering provided in the embodiments of this specification can also be a computer program product, which includes computer program instructions that, when executed by a processor, cause the processor to perform the steps in the excavator load calculation method based on digital twin and Kalman filtering according to various embodiments of this specification as described in the "Exemplary Methods" section above.

[0060] The computer program product can be written in any combination of one or more programming languages ​​to perform the operations of the embodiments of this specification. The programming languages ​​include object-oriented programming languages ​​such as Java and C++, as well as conventional procedural programming languages ​​such as the "C" language or similar programming languages.

[0061] Furthermore, embodiments of this specification also provide a computer-readable storage medium having a computer program stored thereon, the computer program being executed by a processor of the steps in the excavator load calculation method based on digital twin and Kalman filtering according to various embodiments of this specification as described in the "Exemplary Methods" section above.

[0062] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this specification can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0063] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0064] The embodiments described above are merely illustrative of several implementation methods outlined in this specification. While the descriptions are specific and detailed, they should not be construed as limiting the scope of the solutions provided in this specification. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this specification, and these all fall within the scope of protection of this specification. Therefore, the scope of protection for this patent should be determined by the appended claims.

Claims

1. A method for calculating excavator load based on digital twin and Kalman filtering, characterized in that, include: Acquire multi-source sensor data of the target excavator, including the driving pressure and displacement data of each hydraulic cylinder and the body posture data; The joint motion state is calculated based on the displacement data, and the measured joint torque is calculated based on the driving pressure and the current joint configuration. Based on the joint motion state and preset dynamic parameters, the system's own inertial joint torque is calculated, and the measured joint torque is subtracted from the inertial joint torque to obtain the net load torque; A state-space model with load mass and lumped disturbance as state vectors is constructed, and the net load torque is used as the observation variable. A Kalman filter is used to perform recursive optimal estimation to obtain the target load mass estimate of the target excavator.

2. The method according to claim 1, characterized in that, The step of calculating the joint motion state based on the displacement data and calculating the measured joint torque according to the driving pressure and the current joint configuration includes: Using a pre-defined inverse kinematics model of the working device, the displacement data is converted into joint angle vectors to determine the real-time configuration of the robotic arm; Numerical difference and filtering are performed on the joint angle vector sequence to obtain the joint angular velocity and angular acceleration; The pressure difference between the two chambers of each hydraulic cylinder and the effective area of ​​the piston are obtained, and the instantaneous lever arm determined by the real-time configuration is used to calculate the measured joint torque that drives each joint.

3. The method according to claim 1, characterized in that, The calculation of the system's own inertial joint torque based on the joint motion state and preset dynamic parameters includes: The recursive Newton-Euler algorithm is used to calculate the angular velocity, angular acceleration and center of mass acceleration of each link from the base to the end of the target excavator. The internal forces are recursively calculated from the end to the base, and the inertial force and inertial torque of each link are calculated. Based on the balance relationship between force and torque, the inertial joint torque used purely to overcome its own inertia and gravity is calculated.

4. The method according to claim 1, characterized in that, The construction of the state-space model with load quality and lumped disturbance as state vectors includes: The resistance effect during the material excavation process is modeled as the lumped disturbance acting on each joint; The lumped disturbance and the load mass are jointly defined as the state vector to construct the state equation; Based on the kinematic Jacobian matrix, the gravitational torque caused by the load mass is mapped to the joint space and superimposed with the lumped perturbation to construct the observation equation corresponding to the net load torque.

5. The method according to claim 4, characterized in that, Before constructing the observation equation corresponding to the net load torque, the method further includes: Based on the fuselage attitude data and the current joint configuration, the kinematic Jacobian matrix is ​​dynamically updated to adaptively adjust the gravity mapping relationship according to the excavator's three-dimensional spatial attitude changes.

6. The method according to claim 1, characterized in that, The recursive optimal estimation using a Kalman filter to obtain the target load mass estimate of the target excavator includes: Within each calculation cycle, based on the state estimate from the previous time step, the prior estimate of the current state and the prior error covariance matrix are predicted. Calculate the observation residual between the measured net load torque and the predicted observation value for the current period; Calculate the Kalman gain matrix based on the prior error covariance matrix and the preset observation noise covariance matrix; The prior estimate is corrected using the Kalman gain matrix and the observation residual to obtain the posterior estimate of the state vector; The first element is extracted from the posterior estimate and used as the target load quality estimate after full-process correction and compensation.

7. The method according to claim 1, characterized in that, After obtaining the target load mass estimate of the target excavator, the method further includes: The multi-source sensor data and the target load mass estimate are synchronously transmitted to a pre-constructed three-dimensional digital twin model through a preset data interface; The virtual excavator is driven to respond synchronously in the three-dimensional digital twin model, and the estimated value of the target load quality and the corresponding risk warning information are superimposed and displayed on the interactive interface of the virtual scene.

8. The method according to claim 6, characterized in that, After obtaining the posterior estimate of the state vector, the method further includes: Obtain the posterior estimation error covariance matrix of the Kalman filter output; Based on the posterior estimation error covariance matrix, the standard deviation of the target load quality estimate is calculated in real time, and a dynamic confidence interval is constructed. When the upper limit of the dynamic confidence interval reaches a preset warning threshold, a first-level warning signal is triggered; When the lower limit of the dynamic confidence interval exceeds the preset safety limit threshold, an overload condition is determined and a second-level alarm signal is triggered.

9. The method according to claim 6, characterized in that, The method further includes: Continuously monitor the observation residual sequence of the Kalman filter within a continuous sliding window; Calculate the statistical characteristics of the observed residual sequence within the sliding window, the statistical characteristics including the residual mean and covariance; In response to the residual mean deviating from zero by more than a first threshold, or the increase in covariance exceeding a second threshold, the system determines that there is a model mismatch and generates an early fault diagnosis prompt.

10. The method according to claim 1, characterized in that, The acquisition of multi-source sensor data of the target excavator also includes: Acquire high-density 3D point cloud data of the work scene collected by a depth camera at a fixed frequency; The three-dimensional terrain of the work area is reconstructed in real time by integrating the three-dimensional point cloud data, and the terrain undulation, material accumulation outline and spatial distance information are dynamically updated in the corresponding virtual scene.

Images