A method and system for modeling work conditions of a construction machine

Abstract

This invention discloses a method and system for modeling the working conditions of construction machinery, comprising: receiving a user-input work task description, identifying key parameters through preliminary analysis, and converting them into standardized task parameter objects; performing deep analysis on the standardized task parameter objects based on atomic action data in a pre-built atomic action-load library, and converting them into atomic action sequences, their temporal relationships, and working condition context parameters; generating instruction layer data; performing load matching and calculation based on the instruction layer data and working condition context parameters to obtain the actual load curves of each joint; and performing working condition synthesis to obtain a directly usable working condition model file. This invention can automatically convert unstructured work task descriptions of construction machinery into usable digital working condition models containing accurate load information, achieving one-click automatic generation from task description to working condition model, with consistent and repeatable results, and significantly improved efficiency.

Description

Technical Field

[0001] This invention belongs to the field of digital design and R&D technology of engineering machinery, specifically relating to a method and system for modeling the working conditions of engineering machinery. Background Technology

[0002] In the research and development of new energy construction machinery, in order to design products with superior performance and reasonable cost, it is necessary to conduct research on the overall power performance, economy, and energy management strategies based on accurate and comprehensive product operating conditions. With the rapid development and application of new energy construction machinery, research on operating condition modeling of new energy construction machinery is becoming increasingly important.

[0003] Modeling the operating conditions of new energy construction machinery is more complex than that of passenger vehicles. While the modeling of its walking system is similar to that of passenger vehicles, a core challenge in modeling the operating conditions of construction machinery is how to effectively and reasonably define the "load" in its operating condition model. A complete operating condition model not only needs to describe the motion commands of the construction machinery during operation but also must include the load information borne during operation. Due to the complex physical characteristics of the load on construction machinery, the final point of application of the load in the operating system is usually on the end effector. Different operating actions will generate different power requirements for overcoming the end effector load. Therefore, when modeling the operating conditions, it is also necessary to establish an accurate mapping from the "end effect load" to the "load of each action".

[0004] Patent application No. 202511599942.0 discloses a method, apparatus, equipment, and medium for constructing a bucket load spectrum generation model. The method includes: controlling a physical test bench to apply multi-dimensional dynamic loads to various stress points of a hydraulic excavator, and controlling the hydraulic excavator to perform simulated operations according to the working condition type; recording in real time the instantaneous loads acting on the bucket after load transfer to each stress point, and summarizing all instantaneous loads to form a three-dimensional load spectrum of the bucket; collecting dynamic response data generated at each stress point, associating the three-dimensional load spectrum with the dynamic response data to form training samples, aggregating the training samples to form a training dataset and performing feature processing, inputting it into the original deep learning model for iterative training, and using the deep learning model that meets the iteration termination condition as the hydraulic excavator bucket load spectrum generation model.

[0005] Patent application number 201811043486.1 discloses a method for constructing and identifying a loader's operating condition. First, corresponding sensors are deployed on the loader to collect multi-source signals such as torque, pressure, gear position, and braking. The data is then standardized, zero-drift signals are removed, missing values ​​are interpolated, and the collected signals undergo noise reduction and filtering. Second, principal component analysis is used to select high-contribution feature attributes from the loader's multi-attribute data, and statistical analysis is used to extract the features of the principal components. Then, loader operating condition samples are established, and a supervised learning data mining algorithm is used to establish a correlation mapping between load signals and pre-classified operating condition patterns. A working condition identification model is formed through training with a large number of data samples. The feature extraction method of principal component analysis is combined with the KNN algorithm, and the distance formula in the KNN algorithm is improved to better suit operating condition identification, thereby improving the accuracy and efficiency of the working condition identification algorithm.

[0006] However, the existing technologies, including those disclosed above, have the following main drawbacks when implementing the working condition modeling of engineering machinery: (1) The working condition model fails to achieve deep integration of load information with action instructions in a digital and calculable form; (2) When using the working condition model to carry out simulation practice, it is usually necessary to manually set the load. This method relies heavily on the personal experience of engineers and lacks load details, making it difficult to accurately reflect the dynamic characteristics of load changes with action; (3) Traditional load spectra usually only record the total load at the end and cannot directly use the working condition data to analyze the power contribution of each action. At the same time, there is no good method for load decomposition of composite actions; (4) Existing technologies are mostly isolated methods or tools, lacking a complete automated method from task description to usable working condition model files, making it difficult to quickly respond to the working condition modeling needs of multiple scenarios. Summary of the Invention

[0007] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method and system for modeling the working conditions of construction machinery. This method can automatically transform the unstructured work task description of construction machinery into a usable digital working condition model containing accurate load information. It achieves "one-click" automatic generation from task description to working condition model, with consistent and repeatable results, and significantly improved efficiency.

[0008] This invention provides the following technical solution:

[0009] Firstly, a method for modeling the operating conditions of engineering machinery is provided, including:

[0010] Receive the job task description input by the user, identify the key parameters through preliminary parsing, and convert them into a standardized task parameter object;

[0011] Based on the atomic action data in the pre-built atomic action-load library, the standardized task parameter objects are deeply parsed and transformed into atomic action sequences, their temporal relationships, and working condition context parameters.

[0012] Based on atomic action data in the atomic action-load library, instruction layer data is generated according to the atomic action sequence, its timing relationship, and working context parameters.

[0013] Based on the end-load model in the atomic motion-load library and the mechanism and coupling parameters, load matching and calculation are performed according to command layer data and working context parameters to obtain the actual load curve of each joint.

[0014] The atomic action sequence and its timing relationship, instruction layer data, and the actual load curves of each joint are synthesized to obtain a working condition model file that can be used directly.

[0015] Furthermore, the task description received from the user is initially parsed to identify key parameters, and then converted into a standardized task parameter object, including:

[0016] Receive job descriptions from users via a graphical user interface that supports structured forms and restricted natural language input boxes;

[0017] The input job description is initially parsed using a lexical analyzer and a domain-based syntax rule parser, and key parameters, including job stage, environmental conditions, material properties, and target action parameters, are automatically identified.

[0018] The identified key parameters are converted into standardized task parameter objects.

[0019] Furthermore, the atomic action-load library adopts a three-layer architecture design, including an atomic action definition layer, an end-load model layer, and a mechanism and coupling parameter layer;

[0020] The atomic motion definition layer is used to store atomic motion data, including all possible basic motion units of the engineering machinery operation system stored in the form of database tables, i.e., atomic motions, each atomic motion corresponding to a single degree of freedom motion.

[0021] The end-load model layer is used to store end-load models, including load models applied to the end actuator under different operating conditions.

[0022] The mechanism and coupling parameter layer is used to store mechanism and coupling parameters, including a kinematic parameter table of the engineering machinery operation system and a coefficient matrix describing the coupling effect of compound actions. The coupling coefficient matrix is ​​used to reflect the mutual influence of loads caused by hydraulic system flow distribution and mechanical coupling when multiple atomic actions are executed simultaneously.

[0023] Furthermore, based on the atomic action data in the pre-built atomic action-load library, the standardized task parameter objects are deeply parsed and transformed into atomic action sequences, their temporal relationships, and working condition context parameters, including:

[0024] Call atomic action data from the atomic action - workload library;

[0025] Based on atomic action data, the task stages are identified from standardized task parameter objects, the corresponding atomic action sequence is matched for each task stage, and the temporal relationship between each action is clarified.

[0026] Extract the operating context parameters from the task description, including environmental conditions, material properties, and target parameters.

[0027] Furthermore, the generation of instruction layer data based on atomic action data in the atomic action-load library, according to the atomic action sequence, its timing relationship, and operating context parameters, includes:

[0028] Based on the working context parameters, the amplitude and time base of the standard action template corresponding to the atomic action sequence are scaled to make it conform to the specific requirements of the current task.

[0029] The scaled action templates are spliced ​​together according to the parsed stage order and parallel relationship. The dynamic time warping algorithm is used to process the overlapping parts of parallel actions to ensure phase alignment of key actions.

[0030] Generate multi-channel time series data, with each channel corresponding to the instruction value of an atomic action, and sample and output the data at fixed time intervals.

[0031] Furthermore, based on the end-load model in the atomic motion-load library and the mechanism and coupling parameters, load matching and calculation are performed according to command layer data and working context parameters to obtain the actual load curves of each joint, including:

[0032] The system queries the atomic action-load library for end load models that match the operating condition context parameters. If the operating condition context parameters do not perfectly match the discrete conditions stored in the atomic action-load library, the system generates an end load curve suitable for the current operating condition using a multidimensional parameter interpolation algorithm.

[0033] Based on the forward kinematics model of the engineering machinery operation system, the joint angle vector at each moment is calculated according to the displacement curves of each actuator in the command layer data.

[0034] Based on the current joint angle vector, the velocity Jacobian matrix describing the linear mapping relationship between the end-effector velocity and the joint velocity is analytically derived.

[0035] Real-time monitoring of parallel action flags in instruction layer data; when multiple atomic actions are detected to be executed simultaneously, the corresponding coupling coefficient matrix is ​​queried from the mechanism and coupling parameters in the atomic action-load library based on the action combination.

[0036] The theoretical joint load, without considering coupling, is calculated based on the end-load curve, joint angle vector, and velocity Jacobian matrix. If parallel actions exist and the coupling coefficient matrix is ​​found, coupling correction is applied to the actions involved, and the actual load curves of each joint are finally output.

[0037] Furthermore, the theoretical joint load is calculated using the following formula:

[0038] ;

[0039] In the formula, Indicates the theoretical joint load. Representing the velocity Jacobian matrix The transpose of (*), Represents the joint angle vector. Indicates the end load curve;

[0040] The formula for coupling correction of the involved actions is as follows:

[0041] ;

[0042] In the formula, This represents the joint load after coupling correction, and K represents the coupling coefficient matrix.

[0043] Furthermore, the process of synthesizing the atomic action sequence and its timing relationship, instruction layer data, and the actual load curves of each joint to obtain a directly usable working condition model file includes:

[0044] The data aligner aligns all time series on the same time axis. Then, the file generator integrates all data, including header information, job phase division, and time series data, into a complete working condition model file according to a predefined data format.

[0045] The header information includes task description, machine model, and timestamp data; the job phase division records the start and end times of each job phase and the atomic actions it contains; the time series data includes the instruction value, end load value, and joint load value at each time point; and the working condition model file includes three parts: instruction layer, end load layer, and joint load layer data.

[0046] Secondly, a system for modeling the working conditions of engineering machinery is provided, including:

[0047] The task input module is used to receive the job task description input by the user, identify the key parameters through preliminary parsing, and convert them into standardized task parameter objects;

[0048] The task parsing module is used to perform deep parsing of standardized task parameter objects based on atomic action data in the pre-built atomic action-load library, and to transform them into atomic action sequences, their temporal relationships, and working context parameters.

[0049] The instruction generation module is used to generate instruction layer data based on atomic action data in the atomic action-load library, according to the atomic action sequence, its timing relationship, and working context parameters.

[0050] The load matching and calculation module is used to perform load matching and calculation based on the end load model in the atomic motion-load library and the mechanism and coupling parameters, according to the command layer data and working context parameters, to obtain the actual load curve of each joint.

[0051] The working condition synthesis module is used to synthesize the atomic action sequence and its timing relationship, instruction layer data, and the actual load curve of each joint to obtain a working condition model file that can be used directly.

[0052] Furthermore, the load matching and calculation module includes:

[0053] The load matching subunit is used to query the atomic action-load library for the end load model that matches the operating condition context parameters. If the operating condition context parameters do not completely match the discrete conditions stored in the atomic action-load library, the end load curve suitable for the current operating condition is generated by a multi-dimensional parameter interpolation algorithm.

[0054] The attitude calculation subunit is used to calculate the joint angle vector at each moment based on the kinematic forward solution model of the engineering machinery operation system and the displacement curves of each actuator in the command layer data.

[0055] The Jacobian matrix calculation sub-unit is used to analytically derive the velocity Jacobian matrix, which describes the linear mapping relationship between the end-effector velocity and the joint velocity, based on the current joint angle vector.

[0056] The coupling coefficient query subunit is used to monitor the parallel action flags in the instruction layer data in real time. When multiple atomic actions are detected to be executed simultaneously, the corresponding coupling coefficient matrix is ​​queried from the mechanism and coupling parameter in the atomic action-load library according to the action combination.

[0057] The load decomposition calculation subunit is used to calculate the theoretical joint load without considering coupling based on the end load curve, joint angle vector, and velocity Jacobian matrix. If there are parallel actions and the coupling coefficient matrix is ​​found, the coupling correction is performed on the actions involved, and the actual load curve of each joint is finally output.

[0058] Compared with the prior art, the beneficial effects of the present invention are:

[0059] (1) By constructing a multi-level, updatable atomic action-load library, this invention establishes a precise physical mapping from the end load to the load of each atomic action, making the load of each action calculable and traceable, and supporting refined power and energy consumption analysis.

[0060] (2) This invention performs in-depth analysis on standardized task parameter objects and transforms them into atomic action sequences, their temporal relationships, and working condition context parameters. It further generates instruction layer data, performs load matching and calculation based on instruction layer data and working condition context parameters, obtains the actual load curves of each joint, and obtains a working condition model file that can be used directly through working condition synthesis. This invention realizes the automatic, efficient, and standardized transformation of unstructured engineering machinery operation tasks that rely on natural language or expert experience descriptions into digital working condition models containing accurate load information.

[0061] (3) When performing load matching and calculation, this invention combines the Jacobian matrix with the coupling coefficient matrix of the composite action, which solves the problem of load distribution when excavators and other engineering machinery are operating in parallel. It realizes accurate load decomposition of the special characteristics of engineering machinery, so that the generated working condition model can truly reflect the force situation during the operation.

[0062] (4) This invention achieves standardization and automation of working condition modeling: it solidifies the implicit knowledge that relies on expert experience into an explicit and iteratively optimized digital library, realizing the "one-click" automatic generation from task description to working condition model, with consistent and repeatable results, and greatly improved efficiency;

[0063] (5) The invention significantly improves the R&D level of new energy power systems: it provides a high-precision and high-reliability input benchmark for battery matching, motor selection and energy management strategy development of new energy engineering machinery, which helps to achieve product differentiation competitiveness. Attached Figure Description

[0064] Figure 1 This is a flowchart illustrating the engineering machinery operation condition modeling method in Embodiment 1 of the present invention;

[0065] Figure 2 This is a schematic diagram of the atomic action-load library structure in Embodiment 1 of the present invention;

[0066] Figure 3 This is a schematic diagram of the engineering machinery operation condition modeling system in Embodiment 2 of the present invention;

[0067] Figure 4 This is a schematic diagram of the internal structure of the load matching and calculation module in Embodiment 2 of the present invention;

[0068] Figure 5 This is a flowchart illustrating the engineering machinery operation condition modeling method in Embodiment 3 of the present invention. Detailed Implementation

[0069] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and should not be used to limit the scope of protection of the present invention.

[0070] The term "and / or" simply describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. Additionally, the character " / " generally indicates that the preceding and following related objects have an "or" relationship.

[0071] Example 1

[0072] like Figure 1 As shown in the figure, this embodiment provides a method for modeling the working conditions of engineering machinery, and the steps are as follows:

[0073] Step 1: Receive the job description input by the user, identify the key parameters through preliminary parsing, and convert them into a standardized task parameter object.

[0074] Users can input job descriptions through a graphical user interface. This graphical user interface supports structured forms, such as drop-down menus for selecting job scenarios, soil grades, material types, target filling rates, etc. It also supports restricted natural language input boxes. For example, for the digging and loading cycle of a backhoe hydraulic excavator, users can input "dig and fill the bucket to 90% capacity in Class III soil, rotate 90° and unload cycle".

[0075] Upon receiving the job description input by the user, the system uses a lexical analyzer and a domain-based syntax rule parser to initially parse the job description and automatically identify key parameters, including job stage, environmental conditions, material properties, and target action parameters. Then, the identified key parameters are converted into standardized task parameter objects.

[0076] The task parameter object output in this step serves as the initial data source and is passed to subsequent steps for processing. This step eliminates the need for users to possess specialized knowledge of work condition modeling; they can initiate the work condition modeling process simply by describing the task in industry-standard language, significantly lowering the barrier to entry.

[0077] Step 2: Based on the atomic action data in the pre-built atomic action-load library, perform in-depth parsing of the standardized task parameter objects and transform them into executable atomic action sequences, their timing relationships, and working context parameters.

[0078] The atomic action-load library adopts a three-layer architecture design, including an atomic action definition layer, an end-load model layer, and a mechanism and coupling parameter layer, mainly used to store all basic data related to working condition modeling. For example... Figure 2 As shown, this embodiment uses a backhoe hydraulic excavator operating system as an example for introduction.

[0079] (1) First layer: Atomic motion definition layer, used to store atomic motion data, including all possible basic motion units of the engineering machinery operation system stored in the form of database tables, i.e., atomic motions, each atomic motion corresponds to a single degree of freedom motion. For example, the extension / retraction of a certain joint (such as boom joint, stick joint, bucket joint) or the forward / reverse rotation of a certain motor. Each atomic motion record contains a unique identifier, the joint to which it belongs, the type of actuator (such as hydraulic cylinder, motor) and the associated standard motion template index.

[0080] (2) Second layer: End load model layer, used to store end load models, including load models acting on the end effector under different working conditions. Among them, different working conditions include soil grade, material type, filling rate, operation stage, etc., and end effectors include bucket, grab bucket, etc. These load models correspond to the operation stage and can be parametric statistical models, simplified physical and mechanical formulas, or typical load spectrum segments extracted from measured data, with data source and confidence level labels; or they can be parameterizable and extensible templates, such as waveform templates and modulation coefficients normalized to digging resistance for the digging stage, so as to generate load models for more working conditions by combination.

[0081] The end-load model layer is the basic input for subsequent load decomposition, and its accuracy directly affects the quality of the final generated working condition model. Taking the backhoe hydraulic excavator operating system as an example, the load model can be obtained and assigned corresponding confidence labels through various means:

[0082] ①Based on measured data statistics, that is, under typical working conditions, the bucket force data is collected by six force sensors, the load spectrum under the same working conditions is clustered and normalized to form a typical load template and assigned a high confidence label.

[0083] ② Calculation based on physical theory, that is, deriving a parameterized load model based on soil mechanics formulas, such as the excavation resistance formula, and assigning a medium confidence label;

[0084] ③ Based on high-fidelity simulation generation, that is, using the verified simulation model to perform virtual scanning of different working conditions, generating a load database covering the entire scenario, and assigning medium to high confidence labels;

[0085] ④ Refer to industry standards, that is, directly adopt the typical load values ​​given in national standards or industry specifications, and assign corresponding confidence labels.

[0086] By integrating multi-source data and assessing confidence levels, the most suitable load model can be selected based on actual needs, or a weighted fusion of multiple models can be performed.

[0087] (3) The third layer: mechanism and coupling parameter layer, which is used to store mechanism and coupling parameters, including the kinematic parameter table of the engineering machinery operation system and the coefficient matrix describing the coupling effect of the compound action.

[0088] The kinematic parameter table includes parameters such as link lengths, hinge coordinates, and joint limits. Taking a backhoe hydraulic excavator operating system as an example, the kinematic parameter table stores the complete geometric parameters of the backhoe hydraulic excavator's working device, including but not limited to: the lengths of each link (e.g., boom length 5.6m, stick length 2.9m, bucket connecting rod length 1.4m, etc.), the coordinates of each hinge point (e.g., boom-turntable hinge coordinates, boom-stick hinge coordinates, stick-bucket hinge coordinates, etc.), the range of motion of each joint (e.g., boom joint angle range -50° to +30°), and the cylinder installation dimensions (e.g., cylinder hinge position, initial length, etc.). These parameters are the basis for deriving the forward kinematics and the Jacobian matrix.

[0089] The coupling coefficient matrix is ​​used to reflect the load interaction relationship caused by the flow distribution and mechanical coupling of the hydraulic system when multiple atomic actions are executed simultaneously. The coupling coefficient matrix is ​​a diagonally dominant matrix. The diagonal elements reflect the self-transmission characteristics of each action and are usually taken as 1.0; the off-diagonal elements reflect the degree of mutual influence between different actions and are usually taken as 0 to 0.3.

[0090] The coupling coefficient can be obtained through one or a combination of the following methods:

[0091] ① Theoretical analysis: Based on the principle of flow distribution in hydraulic systems and the dynamic equations of mechanisms, the analytical expression of the coupling coefficient is derived;

[0092] ② Actual measurement and calibration: Pressure and displacement sensors were installed on each cylinder of the prototype, and tests were conducted on single actions and compound actions. The coupling coefficient was obtained by comparison and calculation.

[0093] ③ Simulation fitting: Using a validated high-fidelity simulation model, the coupling coefficient is automatically optimized through a parameter identification algorithm;

[0094] ④ Machine learning: Collect a large amount of measured data on complex actions, train a regression model to fit the coupling coefficients, and as more data accumulates, the coupling coefficients can be continuously optimized through adaptive update algorithms, such as Bayesian update and Kalman filtering.

[0095] This step involves in-depth analysis of the standardized task parameter object, as detailed below:

[0096] Step 2.1: Call the atomic action data from the atomic action - load library;

[0097] Step 2.2: Based on atomic motion data, identify the operation stage from the standardized task parameter object, match the corresponding atomic motion sequence for each operation stage, and clarify the temporal relationship between each action; among them, the operation stage, such as the digging and loading cycle of a backhoe hydraulic excavator, generally includes several operation stages such as approaching the material, digging, lifting, rotating, unloading, and returning; the temporal relationship is, for example, serial or parallel.

[0098] Step 2.3: Extract the working context parameters from the task description, including environmental conditions, material properties, target parameters, etc.

[0099] This step is a crucial bridge between natural language task descriptions and computer-understandable instructions; the completeness and accuracy of its parsing rules directly determine the quality of subsequent operational condition modeling.

[0100] Step 3: Based on the atomic action data in the atomic action-load library, generate instruction layer data according to the atomic action sequence, its timing relationship, and the working condition context parameters.

[0101] This step automates the conversion from high-level task descriptions to specific machine instructions, avoiding the tediousness and errors of manual programming. The specific steps are as follows:

[0102] Step 3.1: Pre-store the normalized time-displacement / velocity curve standard motion templates for each atomic action. These standard motion templates reflect the motion law of the action under typical operation, such as the S-shaped curve of joint acceleration-uniform speed-deceleration.

[0103] Step 3.2: Based on the working context parameters, such as the amplitude of the action and the expected speed, scale the amplitude and time base of the standard action template corresponding to the atomic action sequence to make it conform to the specific requirements of the current task.

[0104] Step 3.3: The scaled action templates are spliced ​​together according to the parsed stage order and parallel relationship. The Dynamic Time Warping (DTW) algorithm is used to process the overlapping parts of parallel actions to ensure the phase alignment of key actions.

[0105] Step 3.4: Generate multi-channel time series data, with each channel corresponding to the instruction value of an atomic action, and sample and output at fixed time intervals.

[0106] Step 4: Based on the end load model in the atomic motion-load library and the mechanism and coupling parameters, perform load matching and calculation according to the command layer data and working context parameters to obtain the actual load curve of each joint.

[0107] Step 4.1: Based on the operating condition context parameters, query the atomic action-load library for the end-load model that matches the operating condition context parameters. If the operating condition context parameters do not completely match the discrete conditions stored in the atomic action-load library, then generate an end-load curve suitable for the current operating condition using a multi-dimensional parameter interpolation algorithm. This curve describes the change of six-dimensional force / torque acting on an end effector over time, such as the tip of a bucket tooth or the center of gravity.

[0108] Step 4.2: Based on the kinematic forward kinematics model of the engineering machinery operation system, calculate the joint angle vector at each moment according to the displacement curves of each actuator in the command layer data. Forward kinematics is derived based on the geometric parameters of the mechanism, such as the length of the links and the coordinates of the hinge points. It has a closed-form analytical expression, which can guarantee computational efficiency and accuracy.

[0109] Step 4.3: Based on the current joint angle vector, the velocity Jacobian matrix describing the linear mapping relationship between the end effector velocity and the joint velocity is analytically derived. Its transpose It is used to map end-effector forces / torques to joint equivalent forces / torques. The Jacobian matrix is ​​calculated based on partial derivatives in the kinematic model and can be implemented analytically or numerically.

[0110] Step 4.4: Monitor the parallel action flags in the instruction layer data in real time. When multiple atomic actions are detected to be executing simultaneously, query the corresponding coupling coefficient matrix K from the mechanism and coupling parameters in the atomic action-load library based on the action combination. If there are no parallel actions or no corresponding combination in the library, output the identity matrix. The coupling coefficient matrix reflects the mutual influence of loads caused by the flow distribution and mechanical coupling of the hydraulic system.

[0111] Step 4.5: Calculate the theoretical joint load without considering coupling based on the end-load curve, joint angle vector, and velocity Jacobian matrix: This yields the theoretical equivalent force / torque for each joint; if parallel actions exist and the coupling coefficient matrix is ​​found, then coupling corrections are applied to the involved actions. , This represents the joint load after coupling correction. Here, matrix multiplication is only performed on parallel joints, while other joints remain unchanged. The final output is the actual load curve for each joint. These loads directly correspond to the force conditions of each atomic action and form the basis for subsequent power calculations and energy consumption analysis.

[0112] Step 5: Combine the atomic action sequence and its timing relationship, instruction layer data, and the actual load curves of each joint to obtain a working condition model file that can be used directly.

[0113] First, the system receives the atomic action sequence and its timing relationship, including the job stage division information, instruction layer data, actual load curves of each joint, and optional end load data. Then, it uses a data aligner to align all time series on the same time axis. Since all data is output at the same sampling frequency, it is only necessary to ensure that the timestamps are consistent. Finally, the system uses a file generator to integrate all data, including header information, job stage division, and time series data, into a complete working condition model file according to a predefined data format.

[0114] The predefined data formats include Excel, JSON, HDF5, or MAT; the header information includes task description, machine model, and timestamp data; the job stage division records the start and end times of each job stage and the atomic actions it contains; the time series data includes the instruction value, end load value, and joint load value at each time point.

[0115] The final generated working condition model file includes three parts: the command layer, the end-load layer, and the joint load layer, and is a structured digital asset. This file can be directly imported into various mainstream simulation software, such as multibody dynamics software for structural strength analysis and electromechanical-hydraulic multiphysics joint simulation platforms for dynamic and economic analysis. It can also be widely used in power system matching, control strategy development, product benchmarking and performance evaluation, and data asset accumulation for new energy engineering machinery.

[0116] Example 2

[0117] Based on the same inventive concept as Embodiment 1, such as Figure 3 As shown, this embodiment provides a working condition modeling system for engineering machinery, including: a task input module, a task parsing module, an instruction generation module, a load matching and calculation module, a working condition synthesis module, and an atomic action-load library.

[0118] The task input module serves as the system's interface with the user, receiving the task description from the user, initially parsing and identifying key parameters, and converting them into standardized task parameter objects. This module eliminates the need for users to possess specialized knowledge of work condition modeling; they can initiate the work condition modeling process simply by describing the task in industry-standard language, significantly lowering the barrier to entry.

[0119] The atomic action-load library interacts with the task parsing module, instruction generation module, and load matching and calculation module: the task parsing module obtains atomic action definitions to aid parsing; the instruction generation module obtains atomic actions to generate instruction curves; and the load matching and calculation module queries the end-load model and coupling coefficients for load decomposition calculation. Through this layered design, the library achieves modular storage and flexible retrieval of knowledge, and can be continuously optimized and updated based on subsequent experimental data or simulation results.

[0120] The task parsing module performs deep parsing of standardized task parameter objects based on atomic action data from a pre-built atomic action-load library, transforming them into atomic action sequences, their temporal relationships, and load context parameters. The parsing results are output in structured data format, passing the atomic action sequences, their temporal relationships, and load context parameters to both the instruction generation module and the load matching and calculation module. This module serves as a crucial bridge between natural language task descriptions and computer-understandable instructions; the completeness and accuracy of its parsing rules directly determine the quality of subsequent load condition modeling.

[0121] The instruction generation module is used to generate instruction-level data based on atomic action data in the atomic action-load library, according to the atomic action sequence, its timing relationship, and operating context parameters. The module contains three core sub-units:

[0122] (1) Standard motion template unit, used to pre-store the normalized time-displacement / velocity curve standard motion template for each atomic motion;

[0123] (2) Template scaling unit, used to scale the amplitude and time base of the standard action template corresponding to the atomic action sequence according to the working context parameters, such as action amplitude and expected speed, so that it meets the specific requirements of the current task;

[0124] (3) Timing synthesis unit, which splices the scaled action templates according to the parsed stage order and parallel relationship, uses the dynamic time warping (DTW) algorithm to process the overlapping parts of parallel actions to ensure the phase alignment of key actions; and generates multi-channel time series data, each channel corresponding to the instruction value of an atomic action, and samples and outputs at fixed time intervals.

[0125] The instruction-level data generated by this module serves two purposes: firstly, it is passed as input to the load matching and calculation module for subsequent attitude calculation and load decomposition; secondly, it is also part of the final working condition file, forming part of the working condition model, which can be directly used by the simulation software. Through this module, the system achieves automated conversion from high-level task description to specific mechanical instructions, avoiding the tediousness and errors of manual programming.

[0126] The load matching and calculation module is the core algorithm of this invention. It performs load matching and calculation based on the end-load model in the atomic motion-load library, as well as mechanism and coupling parameters, according to command layer data and operating context parameters, to obtain the actual load curves for each joint. This module consists of five cooperating sub-units, with the specific internal structure as follows: Figure 4 As shown:

[0127] (1) Load matching subunit, used to query the end load model that matches the working condition context parameters from the atomic action-load library. If the working condition context parameters do not completely match the discrete conditions stored in the atomic action-load library, the end load curve suitable for the current working condition is generated by the multidimensional parameter interpolation algorithm.

[0128] (2) Attitude calculation subunit, used to calculate the joint angle vector at each moment based on the kinematic forward solution model of the engineering machinery operation system and the displacement curve of each actuator in the command layer data;

[0129] (3) Jacobian matrix calculation sub-unit, used to analytically derive the velocity Jacobian matrix describing the linear mapping relationship between the end-effector velocity and the joint velocity based on the current joint angle vector;

[0130] (4) Coupling coefficient query subunit, used to monitor the parallel action flags in the instruction layer data in real time. When multiple atomic actions are detected to be executed at the same time, the corresponding coupling coefficient matrix is ​​queried from the mechanism and coupling parameters in the atomic action-load library according to the action combination.

[0131] (5) Load decomposition calculation subunit, used to calculate the theoretical joint load without considering coupling based on the end load curve, joint angle vector and velocity Jacobian matrix; if there are parallel actions and the coupling coefficient matrix is ​​found, the coupling correction is performed on the actions involved, and the actual load curve of each joint is finally output.

[0132] The load matching and calculation module, through close interaction with the instruction generation module, task parsing module, and atomic motion-load library, achieves a physically consistent decomposition from end load to joint load, and takes into account the unique compound motion coupling effect of engineering machinery, ensuring the accuracy and physical interpretability of the calculation results.

[0133] The working condition synthesis module is used to synthesize the atomic action sequence and its timing relationship, instruction layer data, and the actual load curve of each joint to obtain a working condition model file that can be used directly.

[0134] The specific functional implementation of each module and subunit mentioned above is described in the relevant content of the method in Embodiment 1, and will not be repeated here.

[0135] Example 3

[0136] This embodiment uses the excavation and loading cycle modeling of a backhoe hydraulic excavator as an example. Before carrying out specific working condition modeling, it is assumed that the atomic action-load library has been completed, such as... Figure 5 As shown, the specific steps are as follows:

[0137] S1. Receive the job task description input by the user.

[0138] Users input the following information through the graphical interface of the task input module:

[0139] (1) Structured forms:

[0140] Excavator tonnage: 20 tons; Operation type: excavation and loading cycle; Soil grade: Class III soil; Filling rate: 90%; Rotation angle: 90°.

[0141] (2) Natural Language (optional):

[0142] For a 20-ton excavator, dig in Class III soil and fill the bucket to 90% capacity, then rotate 90° to unload and cycle.

[0143] The lexical analyzer and syntax parser inside the task input module automatically identify key parameters and generate standardized task parameter objects, which are then passed to the task parsing module for further processing.

[0144] S2, Call atomic action - load library basic data.

[0145] Based on the key parameters identified in step S1, and according to the task requiring in-depth analysis in step S3, the basic data in the atomic action-load library is called.

[0146] S3. Parse the task description.

[0147] The task parsing module parses the above task parameters into the following structured data, as shown in Tables 1 to 3 below.

[0148] Table 1. Work Phase Sequence

[0149]

[0150] Table 2. Sequence of atomic actions (with "+" indicating action effects)

[0151]

[0152] Table 3 Operating Context Parameters

[0153]

[0154] S4. Call atomic actions - load library basic data.

[0155] Based on the structured data parsed in S3, and according to the instruction data to be generated in S5, the basic data in the atomic action-load library is called.

[0156] S5: Generate instruction data.

[0157] The instruction generation module calls basic data from the atomic action-load library and performs scaling and timing synthesis based on the operating context parameters.

[0158] (1) Standard action template query, as shown in Table 4 below.

[0159] Table 4 Standard Action Template Lookup Table

[0160]

[0161] (2) Template scaling calculation, as shown in Table 5 below.

[0162] Table 5 Template Scaling Calculation Table

[0163]

[0164] (3) Temporal synthesis.

[0165] The timing synthesis unit employs the DTW algorithm to handle overlapping parallel actions. Taking stage 2 (material excavation) as an example, ARM_CROWD and BUCKET_CURL are executed simultaneously, and the DTW algorithm ensures that the critical phases (such as peak load moments) of the two actions are aligned.

[0166] The generated instruction layer data is shown in Table 6 below. The data is sampled at some time points, and the instruction values ​​are normalized relative values ​​between 0 and 1.

[0167] Table 6 Generated Instruction Layer Data Table

[0168]

[0169] S6: Call atomic action - load library basic data.

[0170] Based on the instruction layer data generated in S5, and according to the data required for end-load matching in S7, the basic data in the atomic action-load library is called.

[0171] S7: End-load matching.

[0172] The load matching subunit queries relevant end-load information from the end-load model layer in the atomic action-load library and performs related calculations. A specific example is as follows:

[0173] For the material excavation stage, assuming the silo contains Class III soil and has a 100% filling rate, the excavation resistance waveform (a pulse that first increases and then decreases, with a peak value of 150kN) is obtained by scaling the amplitude by 90% to get the tooth tip force curve during the excavation stage.

[0174] During the full-load lifting phase, the load is mainly the weight of the material: 90% bucket capacity corresponds to a material volume of 0.9m³, a soil density of 1.8t / m³, a weight of approximately 1.62t, and a gravity of approximately 15.9kN, which acts on the center of gravity of the bucket.

[0175] During the rotation phase, the load is mainly the inertial torque: calculated based on the rotational speed and moment of inertia, it is approximately 5 kNm.

[0176] Based on the above results, the end load curve is generated. Table 7 shows examples of some time points.

[0177] Table 7. Examples of partial time points in the terminal load curve.

[0178]

[0179] S8: Attitude and Jacobian calculation.

[0180] By using the attitude calculation subunit and the Jacobian matrix calculation subunit, and based on the instruction layer data generated in step S5, the joint angles at each moment are calculated using forward kinematics, as shown in Table 8 below.

[0181] Table 8. Examples of joint angles calculated at each moment using forward kinematics.

[0182]

[0183] Calculate the Jacobian matrix based on the current posture. This is a 6×4 matrix, with the first three rows corresponding to the end-effector linear velocity and the last three rows corresponding to the end-effector angular velocity. A simplified illustration is provided using the peak excavation time point t=3.5s as an example; in reality, the matrix elements change continuously with the attitude.

[0184] .

[0185] S9: Coupling coefficient query.

[0186] In Phase 2 - Material Extraction, ARM_CROWD and BUCKET_CURL are executed in parallel. The parallel action combination (ARM_CROWD, BUCKET_CURL) is detected by querying the coupling coefficient sub-unit, and the corresponding coupling matrix is ​​retrieved from the coupling coefficient table of the atomic action-load library.

[0187] .

[0188] The meaning of this matrix:

[0189] (1) The diagonal element 1.0 indicates that the load of the action itself mainly comes from itself;

[0190] (2) The off-diagonal element 0.2 indicates that BUCKET_CURL contributes 20% of the load to ARM_CROWD;

[0191] (3) The off-diagonal element 0.15 indicates that ARM_CROWD contributes 15% of the load to BUCKET_CURL.

[0192] In stage 6 - Return, the three actions are performed in parallel, querying the corresponding 3×3 coupling matrix, as shown in the following example:

[0193] .

[0194] S10: Load decomposition calculation.

[0195] The relevant calculations are performed by decomposing the load into sub-units, as shown in the following example:

[0196] (1) Calculation of theoretical joint load.

[0197] For each time point, calculate the theoretical joint load. .

[0198] Taking t=3.5s as an example, the end load Jacobian matrix As described in S8, the calculation yields:

[0199] ;

[0200] in:

[0201] The boom cylinder force;

[0202] The force of the boom cylinder;

[0203] Bucket cylinder force;

[0204] The rotational torque has not yet been applied.

[0205] (2) Coupling correction.

[0206] At t=3.5s, there are parallel actions ARM_CROWD and BUCKET_CURL, and a coupling matrix is ​​applied. Make corrections:

[0207] .

[0208] Other joints, such as the boom and slewing joints, remain unchanged: , .

[0209] For t=6.0s, during the rotation phase, there are no parallel mining operations, and the theoretical load value is used directly;

[0210] For t=13.0s, in the return phase, the three actions are performed in parallel, and a similar correction is made using a 3×3 coupling matrix.

[0211] (3) Load curve of complete joint.

[0212] Repeat the above calculations for all time points to obtain the complete joint load curves. Table 9 shows some examples.

[0213] Table 9. Example of Joint Load Curve

[0214]

[0215] S11: Combining operating condition documents.

[0216] The command layer, end load layer, and joint load layer are aligned by time using the working condition synthesis module to generate a structured working condition model file.

[0217] All data were sampled at 0.1-second intervals, with strictly aligned timestamps. For example, at t=3.5s, the data from all three layers corresponded completely at the same time point. An Excel file was generated, containing the following three worksheets:

[0218] (1) Worksheet “Header Information”: Stores metadata such as task description, machine model, and timestamp.

[0219] (2) Worksheet “Phase Division”: Record the start and end times of each work phase and the atomic actions it contains.

[0220] (3) Worksheet "Time Series": Stores the instruction layer, end load layer and joint load layer data at each time point in tabular form.

[0221] The following table 10 shows an example of the time series data in the worksheet.

[0222] Table 10: Example of partial data from a time series worksheet.

[0223]

[0224] In summary, this invention establishes a precise physical mapping from end-point load to the load of each atomic action by constructing a multi-layered, updatable atomic action-load library. This makes the load of each action calculable and traceable, supporting refined power and energy consumption analysis. By deeply analyzing standardized task parameter objects and transforming them into atomic action sequences, their temporal relationships, and operating condition context parameters, instruction layer data is further generated. Based on the instruction layer data and operating condition context parameters, load matching and calculation are performed to obtain the actual load curves of each joint. Through operating condition synthesis, a directly usable operating condition model file is obtained. This enables the automatic, efficient, and standardized transformation of unstructured engineering machinery operation tasks that rely on natural language or expert experience descriptions into digital operating condition models containing precise load information.

[0225] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0226] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0227] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0228] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0229] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for modeling the working conditions of engineering machinery, characterized in that, include: Receive the job task description input by the user, identify the key parameters through preliminary parsing, and convert them into a standardized task parameter object; Based on the atomic action data in the pre-built atomic action-load library, the standardized task parameter objects are deeply parsed and transformed into atomic action sequences, their temporal relationships, and working condition context parameters. Based on atomic action data in the atomic action-load library, instruction layer data is generated according to the atomic action sequence, its timing relationship, and working context parameters. Based on the end-load model in the atomic motion-load library and the mechanism and coupling parameters, load matching and calculation are performed according to command layer data and working context parameters to obtain the actual load curve of each joint. The atomic action sequence and its timing relationship, instruction layer data, and the actual load curves of each joint are synthesized to obtain a working condition model file that can be used directly.

2. The method for modeling the working conditions of engineering machinery according to claim 1, characterized in that, The task description received from the user is initially parsed to identify key parameters, which are then converted into a standardized task parameter object, including: Receive job descriptions from users via a graphical user interface that supports structured forms and restricted natural language input boxes; The input job description is initially parsed using a lexical analyzer and a domain-based syntax rule parser, and key parameters, including job stage, environmental conditions, material properties, and target action parameters, are automatically identified. The identified key parameters are converted into standardized task parameter objects.

3. The method for modeling the working conditions of engineering machinery according to claim 1, characterized in that, The atomic action-load library adopts a three-layer architecture design, including an atomic action definition layer, an end-load model layer, and a mechanism and coupling parameter layer. The atomic motion definition layer is used to store atomic motion data, including all possible basic motion units of the engineering machinery operation system stored in the form of database tables, i.e., atomic motions, each atomic motion corresponding to a single degree of freedom motion. The end-load model layer is used to store end-load models, including load models applied to the end actuator under different operating conditions. The mechanism and coupling parameter layer is used to store mechanism and coupling parameters, including a kinematic parameter table of the engineering machinery operation system and a coefficient matrix describing the coupling effect of compound actions. The coupling coefficient matrix is ​​used to reflect the mutual influence of loads caused by hydraulic system flow distribution and mechanical coupling when multiple atomic actions are executed simultaneously.

4. The method for modeling the working conditions of engineering machinery according to claim 1, characterized in that, The atomic action data based on the pre-built atomic action-load library is used to perform deep parsing of standardized task parameter objects and transform them into atomic action sequences, their temporal relationships, and working condition context parameters, including: Call atomic action data from the atomic action - workload library; Based on atomic action data, the task stages are identified from standardized task parameter objects, the corresponding atomic action sequence is matched for each task stage, and the temporal relationship between each action is clarified. Extract the operating context parameters from the task description, including environmental conditions, material properties, and target parameters.

5. The method for modeling the working conditions of engineering machinery according to claim 1, characterized in that, The generation of instruction layer data based on atomic action data in the atomic action-load library, according to the atomic action sequence, its timing relationship, and operating context parameters, includes: Based on the working context parameters, the amplitude and time base of the standard action template corresponding to the atomic action sequence are scaled to make it conform to the specific requirements of the current task. The scaled action templates are spliced ​​together according to the parsed stage order and parallel relationship. The dynamic time warping algorithm is used to process the overlapping parts of parallel actions to ensure phase alignment of key actions. Generate multi-channel time series data, with each channel corresponding to the instruction value of an atomic action, and sample and output the data at fixed time intervals.

6. The method for modeling the working conditions of engineering machinery according to claim 1, characterized in that, The end-effector load model based on the atomic motion-load library, along with mechanism and coupling parameters, performs load matching and calculation based on command layer data and operating context parameters to obtain the actual load curves for each joint, including: The system queries the atomic action-load library for end load models that match the operating condition context parameters. If the operating condition context parameters do not perfectly match the discrete conditions stored in the atomic action-load library, the system generates an end load curve suitable for the current operating condition using a multidimensional parameter interpolation algorithm. Based on the forward kinematics model of the engineering machinery operation system, the joint angle vector at each moment is calculated according to the displacement curves of each actuator in the command layer data. Based on the current joint angle vector, the velocity Jacobian matrix describing the linear mapping relationship between the end-effector velocity and the joint velocity is analytically derived. Real-time monitoring of parallel action flags in instruction layer data; when multiple atomic actions are detected to be executed simultaneously, the corresponding coupling coefficient matrix is ​​queried from the mechanism and coupling parameters in the atomic action-load library based on the action combination. The theoretical joint load, without considering coupling, is calculated based on the end-load curve, joint angle vector, and velocity Jacobian matrix. If parallel actions exist and the coupling coefficient matrix is ​​found, coupling correction is applied to the actions involved, and the actual load curves of each joint are finally output.

7. The method for modeling the working conditions of engineering machinery according to claim 6, characterized in that, The theoretical joint load is calculated using the following formula: ； In the formula, Indicates the theoretical joint load. Representing the velocity Jacobian matrix The transpose of (*), Represents the joint angle vector. Indicates the end load curve; The formula for coupling correction of the involved actions is as follows: ； In the formula, This represents the joint load after coupling correction, and K represents the coupling coefficient matrix.

8. The method for modeling the working conditions of engineering machinery according to claim 1, characterized in that, The process of synthesizing the atomic action sequence and its timing relationship, instruction layer data, and the actual load curves of each joint to obtain a directly usable working condition model file includes: The data aligner aligns all time series on the same time axis. Then, the file generator integrates all data, including header information, job phase division, and time series data, into a complete working condition model file according to a predefined data format. The header information includes task description, machine model, and timestamp data; the job phase division records the start and end times of each job phase and the atomic actions it contains; the time series data includes the instruction value, end load value, and joint load value at each time point; and the working condition model file includes three parts: instruction layer, end load layer, and joint load layer data.

9. A system for modeling the working conditions of engineering machinery, characterized in that, include: The task input module is used to receive the job task description input by the user, identify the key parameters through preliminary parsing, and convert them into standardized task parameter objects; The task parsing module is used to perform deep parsing of standardized task parameter objects based on atomic action data in the pre-built atomic action-load library, and to transform them into atomic action sequences, their temporal relationships, and working context parameters. The instruction generation module is used to generate instruction layer data based on atomic action data in the atomic action-load library, according to the atomic action sequence, its timing relationship, and working context parameters. The load matching and calculation module is used to perform load matching and calculation based on the end load model in the atomic motion-load library and the mechanism and coupling parameters, according to the command layer data and working context parameters, to obtain the actual load curve of each joint. The working condition synthesis module is used to synthesize the atomic action sequence and its timing relationship, instruction layer data, and the actual load curve of each joint to obtain a working condition model file that can be used directly.

10. The engineering machinery operating condition modeling system according to claim 9, characterized in that, The load matching and calculation module includes: The load matching subunit is used to query the atomic action-load library for the end load model that matches the operating condition context parameters. If the operating condition context parameters do not completely match the discrete conditions stored in the atomic action-load library, the end load curve suitable for the current operating condition is generated by a multi-dimensional parameter interpolation algorithm. The attitude calculation subunit is used to calculate the joint angle vector at each moment based on the kinematic forward solution model of the engineering machinery operation system and the displacement curves of each actuator in the command layer data. The Jacobian matrix calculation sub-unit is used to analytically derive the velocity Jacobian matrix, which describes the linear mapping relationship between the end-effector velocity and the joint velocity, based on the current joint angle vector. The coupling coefficient query subunit is used to monitor the parallel action flags in the instruction layer data in real time. When multiple atomic actions are detected to be executed simultaneously, the corresponding coupling coefficient matrix is ​​queried from the mechanism and coupling parameter in the atomic action-load library according to the action combination. The load decomposition calculation subunit is used to calculate the theoretical joint load without considering coupling based on the end load curve, joint angle vector, and velocity Jacobian matrix. If there are parallel actions and the coupling coefficient matrix is ​​found, the coupling correction is performed on the actions involved, and the actual load curve of each joint is finally output.

Images