Method and system for digital twinborn construction and multi-performance real-time evaluation of equipment

Abstract

The invention provides a method and a system for digital twinborn construction and multi-performance real-time evaluation of equipment, the core of the method is to construct a digital twinborn body of the equipment and develop a multi-performance real-time evaluation agent model, and the method comprises the following steps: constructing a multi-field simulation virtual prototype of the equipment based on a joint simulation technology; various performance indexes of the equipment are comprehensively presented through simulation analysis; the model of the virtual prototype is combined with sensing data, and after visualization processing, a digital twinborn body capable of reflecting direct observation information of equipment is constructed; a robust agent model construction strategy is provided, and batch joint simulation data is adopted to train the model, so that the indirect performance index of the equipment can be accurately evaluated; sensor data for collecting working conditions in real time and an agent model for evaluating performance in real time are integrated in a digital twinborn body, and a model-data dual-drive performance evaluation mechanism is formed.

Description

Technical Field

[0001] The present invention relates to the technical field of digital twins and automation technology for equipment. Specifically, it relates to a method and system for constructing a digital twin of equipment and real-time multi-performance evaluation. Background Art

[0002] With the rapid development of industrialization, many major equipment are widely deployed in extreme environment fields such as ore mining, tunnel boring, and metallurgical forming. In these harsh scenarios, the equipment is long-term exposed to the impact of harsh working conditions such as extreme temperatures, huge pressures, and strong vibrations. Compared with ordinary equipment, they are more likely to suffer from complex conditions such as structural damage and functional failure. It is not only necessary to ensure efficient operation, but also to achieve high performance robustness and long-term durability of the structure. Therefore, realizing the timely monitoring and accurate evaluation of various performances of the equipment has become the core element to ensure the efficient operation of the equipment and improve the operation reliability.

[0003] At present, the real-time evaluation of equipment performance mainly relies on monitoring devices such as sensors. Its principle is to install sensors at relevant parts of the equipment to collect working state data in real time. However, in extreme temperature and pressure environments, sensors face many difficulties. On the one hand, the installation difficulty is large; on the other hand, the problem of frequent failure is prominent, resulting in serious limitations in its normal use function and inability to monitor the performance changes of the equipment in real time. In addition, traditional sensor technology has inherent deficiencies and cannot comprehensively display the comprehensive performance of the equipment. For indirect performance indicators such as equipment execution accuracy, structural strength, and wear consumption, it is difficult to conduct real-time and accurate evaluations relying on the monitoring data of a single sensor.

[0004] As an emerging digital analysis method, digital twin technology has made certain progress in the field of equipment automation technology. It combines the digital model of the physical equipment with the real-time collected data to construct a digital copy of the equipment, and realizes the real-time performance evaluation of the equipment based on the self-update mechanism. In this process, the digital twin needs to fuse the model data and the collected data in real time, and ensure the reliability and real-time of the system in extreme environments. But currently this technical route still faces many challenges. The most intractable one is how to accurately evaluate various performance indicators of the equipment under the limited use of sensors.

[0005] During the operation of equipment, many indirect performance indicators that cannot be directly monitored and obtained are precisely the key to evaluating its operating conditions. In previous studies, the method of establishing simulation models was mostly used to analyze these performance indicators offline. For example, finite element simulation can be used to evaluate structural strength indicators such as stress and strain, discrete element simulation can evaluate the working efficiency when the equipment processes discrete media, and hydraulic control simulation can evaluate the action execution accuracy of the equipment. However, these methods have obvious defects. On the one hand, it is difficult for multiple simulation models to form an organic whole and cannot truly reflect the operating mechanism of the equipment and the interaction process with the environment under extreme conditions; on the other hand, the results of the simulation model are restricted by the input conditions and cannot flexibly adapt to the dynamic changes of extreme working conditions.

[0006] Through the retrieval of patent documents, it is found that the invention patent with the publication number CN 115432634 A discloses a method for constructing a digital twin of a forklift based on a flight control system, including the following steps: importing a three-dimensional model, an electrical model, and a hydraulic model into simulation software to establish a digital twin of the forklift; obtaining the operating data of the forklift through the flight control system and multiple sensors; exporting the digital twin of the forklift from the simulation software to generate a digital twin application program; writing the real-time operating data of the forklift into the parameter file of the digital twin through the ground station of the flight control system; running the digital twin application program, loading the parameter file, and generating a result file; obtaining the recommended value of the driving speed of the existing load capacity of the forklift according to the evaluation report and the result file. This patent only targets forklifts, with limited application scenarios and single evaluation content, only involving the recommendation of driving speed.

[0007] In summary, aiming at the problems of the above-mentioned existing technologies, researching a method and system for constructing a digital twin of equipment and real-time multi-performance evaluation has become a key task that needs to be solved urgently at present. Summary of the Invention

[0008] Aiming at the defects in the existing technologies, the purpose of the present invention is to provide a method and system for constructing a digital twin of equipment and real-time multi-performance evaluation.

[0009] According to the present invention, 1. A method for constructing a digital twin of equipment and real-time multi-performance evaluation includes the following steps:

[0010] Step S1, building multiple domain simulation models: decomposing the functional modules of the equipment according to the actual working conditions, and building multiple domain simulation models that can analyze the working mechanism and indirect performance indicators of the equipment;

[0011] Step S2, constructing a simulation virtual prototype: integrating multiple domain simulation models by using a variety of co-simulation technologies to construct a simulation virtual prototype that can simulate the entire working process of the equipment and perform batch simulations;

[0012] Step S3, constructing the digital twin: Visualize the appearance design and internal physical model of the equipment in the simulation virtual prototype, integrate the real-time data collected by sensors, initially construct the digital twin, and dynamically update the state of the digital twin by continuously integrating the real-time data;

[0013] Step S4, batch simulation and adaptive construction of the surrogate model: Based on the set input conditions, the simulation virtual prototype conducts batch simulations, generates simulation results, and trains a surrogate model using an adaptive construction strategy with the simulation results, and outputs the trained surrogate model;

[0014] Step S5, integrated learning and improvement of the surrogate model: Adopt an integrated learning strategy to integrate multiple trained surrogate models optimized by hyperparameters to form an integrated learner that can perform offline evaluations of multiple performances of the equipment under different working conditions;

[0015] Step S6, forming a performance evaluation mechanism for the digital twin, combining the integrated learner with the real-time data of the sensors in the digital twin to form a real-time evaluation method for various indirect performances of the equipment.

[0016] Preferably, step S1 includes the following sub-steps:

[0017] Step S1.1, in combination with the environmental working conditions and actual conditions, conduct a detailed analysis of the working principle and operation mechanism of the equipment to form a three-dimensional geometric model diagram, a hydraulic transmission schematic diagram, and a working mechanism specification;

[0018] Step S1.2, according to the functional analysis of the equipment, decompose the equipment modularly, divide the equipment into multiple different domains, so as to select professional software in each domain respectively for simulation modeling;

[0019] Step S1.3, according to the characteristics of each module domain, select an appropriate simulation method and establish a corresponding simulation model, and the simulation model reflects the working mechanism and various performance indicators of the equipment in the domain;

[0020] Step S1.4, based on the design parameters of the equipment, the use environment, and the design requirements in the working environment, adjust the simulation models of multiple domains to ensure that indirect performance indicators that cannot be directly obtained through sensors can be analyzed through the simulation models of multiple domains;

[0021] Step S1.5, verify the simulation models of multiple domains according to experimental data or theoretical analysis.

[0022] Preferably, step S2 includes the following sub-steps:

[0023] Step S2.1: Establish a "machine - hydraulics - control" co - simulation model to describe the execution of the equipment's own actions. Use model interchange technology to unify multiple - domain simulation models into a feedback control model.

[0024] Step S2.2: Establish an "institution - environment" co - simulation model to describe the real - time interaction between the equipment and the environment. Conduct "machine - hydraulics - control" coupled modeling for the institution, perform finite - element or discrete - element modeling for the environment, and use co - simulation technology to couple the two to form a preliminary simulation virtual prototype.

[0025] Step S2.3: Based on the design parameters of the equipment, the usage environment, and the design requirements in the working environment, adjust the preliminary simulation virtual prototype to obtain an adjusted simulation virtual prototype.

[0026] Step S2.4: Verify the model of the adjusted simulation virtual prototype. After passing the verification, clarify the indirect performance indicators of the equipment that need to be obtained through simulation analysis. The indirect performance indicators of the equipment are used to measure the working efficiency, performance robustness, and structural durability of the equipment.

[0027] Step S2.5: Take the indirect performance indicators of the equipment as the simulation target, set the influencing factors of the indirect performance indicators of the equipment as the simulation input conditions, and conduct Latin - hypercube experimental design for the simulation input conditions to form a simulation experimental design table.

[0028] Preferably, in Step S3, the digital twin is a digital model that describes the working state of the equipment, constructed by using visualization technology in the virtual engine Unity3D, based on the multi - domain simulation virtual prototype formed in Step S2 and the sensors originally installed at the necessary parts of the equipment.

[0029] Preferably, Step S3 includes the following sub - steps:

[0030] Step S3.1: Import the 3D geometric model and hydraulic schematic diagram of the equipment in the simulation virtual prototype verified in Step S2 into the Unity3D platform. Based on the scene - building function of Unity3D, create the external form, internal structure, and composition of each functional module of the equipment, and visualize the created equipment in the 3D scene through Mesh rendering to obtain a preliminary digital twin.

[0031] Step S3.2: Combine the dynamic simulation in the simulation virtual prototype with the preliminary digital twin. Use C# scripts to integrate the structural dynamics model of the equipment into the Rigidbody of the physical engine in Unity3D to obtain a digital twin with dynamic motion, enabling the digital twin to dynamically reflect the motion mechanism of the equipment.

[0032] Step S3.3: On the digital twin with dynamic motion, real-time sensor data of the equipment is received in real time through Socket communication of Unity3D, and the real-time sensor data is mapped to each corresponding component of the digital twin with dynamic motion to obtain a digital twin integrated with real-time sensing data;

[0033] Step S3.4: Use the real-time sensor data to update the state of the digital twin, and display it using the real-time rendering function of Unity3D.

[0034] Preferably, in step S4, for multiple performance indicators of the equipment, based on the dataset of batch simulations, an adaptive construction strategy is adopted for training and hyperparameter optimization. The surrogate model is provided with batch training data by the simulation virtual prototype, avoiding time-consuming and costly experimental analysis of extreme environment equipment, clarifying multiple performance indicators to be evaluated for the equipment and their respective influencing factors. In the simulation virtual prototype, the performance indicators are used as analysis results, and the influencing factors are used as input conditions. Based on the experimental design of the input conditions, batch simulation analysis is carried out, and the results of batch simulations of the virtual prototype are used for training for subsequent training of the surrogate model.

[0035] Preferably, step S4 includes the following sub-steps:

[0036] Step S4.1: Based on the simulation experimental design table formed in step S2, batch simulations are carried out in the simulation virtual prototype, and the simulation results are exported and sorted to obtain a simulation dataset;

[0037] Step S4.2: Construct an adaptive surrogate model to build a robust mapping relationship between multiple performance indicators of the equipment and their influencing factors; the surrogate model includes four base learners: Support Vector Regression (SVR), Radial Basis Neural Network (RBF), Random Forest (RF), and K-Nearest Neighbor Regression (KNN);

[0038] Step S4.3: On the simulation dataset corresponding to each performance indicator, the four base learners are trained, and the particle swarm optimization algorithm (PSO) is used to optimize the hyperparameters of the four base learners with the goal of the accuracy of model evaluation. Among them, the hyperparameters of SVR are the penalty parameter c and the tolerance error limit ε, the hyperparameters of RBF are the radial basis influence range γ and the radial basis standard deviation σ, the hyperparameters of RF are the number of random trees n and the maximum depth m, and the hyperparameters of KNN are the k value and the weight w to obtain the optimized base learners;

[0039] Step S4.4: Based on the optimized base learners, a trained surrogate model is obtained.

[0040] Preferably, step S5 includes the following sub-steps:

[0041] Step S5.1: For each performance metric, initially select a base learner from the trained surrogate models, use the base learner to conduct a preliminary evaluation of the performance, and calculate the residual between the evaluation result and the actual value.

[0042] Step S5.2: Based on the residuals, adjust the training parameters of the remaining base learners, and retrain the remaining base learners. The training objective is to minimize the squared error of the residuals.

[0043] Step S5.3: Combine the initially selected base learner and the remaining adjusted base learners in a weighted manner to optimize the overall prediction result through weighting. The weight of each base learner depends on its improvement degree of the residuals.

[0044] Step S5.4: Iteratively execute Steps S5.1 to S5.3. Each round of training focuses on the residuals generated in the previous round of training and improves the residuals, finally forming an ensemble learner.

[0045] Preferably, Step S6 includes the following sub-steps:

[0046] Step S6.1: Package the ensemble learner formed in Step S5 using C# code and ensure that it can be called in the code of the digital twin.

[0047] Step S6.2: Based on the sensor data interface established by the digital twin, receive real-time sensor data. The digital twin performs timeliness identification and adjustment on the data and uses it as the input of the ensemble learner.

[0048] Step S6.3: Deeply process and extract features from the sensor data through the ensemble learner, and evaluate various indirect performance metrics of the equipment in real time, including working efficiency, structural strength, and durability.

[0049] Step S6.4: Through visualization technology, display the evaluation results of the digital twin on the multi-performance metrics of the equipment in real time.

[0050] The present invention also provides a digital twin construction and multi-performance real-time evaluation system for equipment, including:

[0051] Module M1: Build multiple domain simulation models. Decompose the equipment into functional modules according to the actual working conditions, and build multiple domain simulation models that can analyze the working mechanism and indirect performance metrics of the equipment.

[0052] Module M2: Construct a simulation virtual prototype. Integrate multiple domain simulation models using various co-simulation technologies to construct a simulation virtual prototype that can simulate the entire working process of the equipment and conduct batch simulations.

[0053] Module M3, constructing a digital twin: Visualize the appearance design and internal physical model of the equipment in the simulation virtual prototype, integrate the real-time data collected by sensors, initially construct a digital twin, and dynamically update the state of the digital twin by continuously integrating real-time data;

[0054] Module M4, batch simulation and adaptive construction of surrogate models: Based on the set input conditions, the simulation virtual prototype conducts batch simulations, generates simulation results, and trains a surrogate model using an adaptive construction strategy with the simulation results, and outputs the trained surrogate model;

[0055] Module M5, integrated learning and improvement of surrogate models: Adopt an integrated learning strategy to integrate multiple trained surrogate models optimized by hyperparameters to form an integrated learner that can perform offline evaluations of multiple performances of the equipment under different working conditions;

[0056] Module M6, forming a performance evaluation mechanism for the digital twin, combining the integrated learner with the real-time data of the sensors in the digital twin to form a real-time evaluation method for various indirect performances of the equipment.

[0057] Compared with the prior art, the present invention has the following beneficial effects:

[0058] 1. This method is applicable to equipment in extreme fields such as ore mining, tunnel boring, and metallurgical forming. With only a small amount of sensor data, it can quickly evaluate indirect analysis results such as the working effectiveness, performance reliability, and structural durability of major equipment, promoting the health monitoring, performance optimization, and fault prevention of major equipment in harsh environments. It effectively solves the problems of sensor failure and incomplete performance data, significantly enhances the ability of the equipment to operate healthily for a long time in extreme environments, and provides a practical and innovative solution for equipment performance evaluation.

[0059] 2. The present invention uses joint simulation and visualization technologies to construct a simulation virtual prototype of the equipment, and then builds a digital twin. On this basis, it trains an adaptive robust surrogate model based on integrated learning. This model is constructed based on the data analyzed in batches by the simulation virtual prototype, and organically combines multi-field offline simulation analysis with digital twin performance monitoring, and can quickly and accurately evaluate the equipment performance under specific working conditions. During the operation of the equipment, this surrogate model is fused with the sensing data in the digital twin, and through the "model-data" dual-drive mechanism, it can evaluate indirect performance indicators such as the working efficiency, structural strength, and durability of the equipment in real time.

[0060] 3. The present invention adopts a robust surrogate model, which also improves the accuracy and reliability of the evaluation, overcomes the difficulties in performance evaluation caused by environmental uncertainties, and further promotes the improvement of the overall performance of the equipment. Description of the Drawings

[0061] Other features, objects, and advantages of the present invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0062] Figure 1 is the overall framework of the equipment multi-performance real-time evaluation method in the embodiments of the present invention;

[0063] Figure 2 is the schematic block diagram of the simulation analysis of multiple indirect performance indicators in the embodiments of the present invention;

[0064] Figure 3 is the schematic block diagram of the establishment of the simulation virtual prototype in the embodiments of the present invention;

[0065] Figure 4 is the schematic block diagram of the preliminary establishment of the digital twin in the embodiments of the present invention;

[0066] Figure 5 is the schematic block diagram of the adaptive construction strategy of the surrogate model in the embodiments of the present invention;

[0067] Figure 6 is the schematic block diagram of the self-enhancing integration mechanism of the surrogate model in the embodiments of the present invention;

[0068] Figure 7 is the schematic block diagram of the dual-drive evaluation mechanism formed by the surrogate model and sensing data in the embodiments of the present invention. Detailed Embodiments

[0069] The present invention will be described in detail below with reference to specific embodiments. A specific embodiment of the present invention is the cutting of complex intercalated gangue coal seams by a roadheader. In the patent drawings, elements having similar structures or functions will be denoted by the same reference numerals. The drawings and embodiments are provided for reference and illustration only and are not intended to limit the present invention. The dimensions shown in the drawings and embodiments are only for clear description and do not limit the proportional relationship or provide an exhaustive description of the present invention, nor do they limit the scope of the present invention.

[0070] The present invention discloses a method and system for constructing a digital twin of an equipment and real-time multi-performance evaluation, which is mainly applied to major equipment operating in extreme environments. In extreme environments, sensors cannot work properly, resulting in problems such as structural damage and functional failure of the equipment. Therefore, while ensuring the high efficiency of the equipment operation, it is extremely crucial to achieve performance robustness and structural durability, and construct a digital twin for real-time performance evaluation. The core of the present invention lies in constructing a digital twin of the equipment and developing a real-time multi-performance evaluation surrogate model, including the following steps: constructing a multi-domain simulation virtual prototype of the equipment based on co-simulation technology, and comprehensively presenting various performance indicators of the equipment through simulation analysis; combining the model of the virtual prototype with sensing data, and after visualization processing, constructing a digital twin that can reflect the direct observation information of the equipment; proposing a robust surrogate model construction strategy, training the model with batch co-simulation data to accurately evaluate the indirect performance indicators of the equipment; integrating sensor data for real-time acquisition of working conditions and the surrogate model for real-time performance evaluation in the digital twin to form a "model-data" dual-driven performance evaluation mechanism.

[0071] Embodiment 1:

[0072] Taking the roadheader, a major equipment facing the complex extreme environment of rock breaking, as a specific embodiment for elaboration.

[0073] Figure 1 It is the overall framework of the method for real-time multi-performance evaluation of the equipment in the embodiment of the present invention.

[0074] As Figure 1 , the embodiment of the present invention provides a method for constructing a digital twin of an equipment and real-time multi-performance evaluation, including the following steps:

[0075] Step S1, building multiple domain simulation models: decomposing the functional modules of the equipment according to the actual working conditions, and building multiple domain simulation models that can analyze the working mechanism and indirect performance indicators of the equipment.

[0076] Specifically, since the extreme environments faced by major equipment are different, their working principles will also vary greatly. The multiple domain simulation models are precisely for simulating the operation of each component of the equipment under real working conditions in the simulation environment. In this embodiment, the multiple domain simulation models first combine the extreme environment working conditions with the equipment operation mechanism, modularize the functions of the equipment into multiple domains, and then use professional simulation software in each domain to successively build simulation models. The simulation models can truly reflect the mechanisms such as the appearance design, joint movement, hydraulic transmission, and structural strength of the equipment; and accurately analyze the indirect performance indicators such as the working efficiency, multi-physical fields, and loss conditions of the equipment.

[0077] Figure 2 It is the schematic block diagram of the simulation analysis of various indirect performance indicators in the embodiment of the present invention.

[0078] As shown Figure 2 below, step S1 includes the following sub-steps:

[0079] Step S1.1, in combination with the environmental conditions and actual conditions, conduct a detailed analysis of the working principle and operation mechanism of the equipment to form necessary preliminary preparation materials such as a three-dimensional geometric model diagram, a hydraulic transmission schematic diagram, and a working mechanism specification;

[0080] Step S1.2, according to the functional analysis of the equipment, decompose the equipment modularly, divide the equipment into multiple different fields, so as to select professional software in each field to conduct simulation modeling respectively;

[0081] Step S1.3, according to the characteristics of each module field, select an appropriate simulation method, establish a corresponding simulation model, and the simulation model reflects the working mechanism and various performance indicators of the equipment in the field to which it belongs.

[0082] In this embodiment, a cantilever roadheader is used as an example. The instruction control software SIMULINK is used to simulate the feedback module of the equipment, the multi-field system software AMESIM is used to simulate the hydraulic module of the equipment, the multi-body dynamics software ADAMS is used to simulate the power module of the equipment, and the discrete element software EDEM is used to simulate the execution module of the equipment, etc. By establishing simulation models in their respective fields, analyze indicators such as the working efficiency, wear and tear, load conditions, and execution accuracy of the tunneling robotic arm.

[0083] Step S1.4, to ensure the applicability and accuracy of the simulation model, based on the design parameters of the equipment, the use environment, and the design requirements in the working environment, adjust the simulation models in multiple fields to ensure that indirect performance indicators that cannot be directly obtained through sensors can be analyzed through the simulation models in multiple fields;

[0084] Step S1.5, verify the simulation models in multiple fields according to experimental data or theoretical analysis.

[0085] In the actual working process, the functional modules such as the structural dynamics, hydraulic transmission, and feedback control of the equipment cooperate with each other and influence each other, and at the same time the equipment also interacts with the extreme environment it faces in real time. Therefore, it is necessary to construct an overall simulation virtual prototype.

[0086] Step S2, construct a simulation virtual prototype: use a variety of co-simulation technologies to integrate simulation models in multiple fields to construct a simulation virtual prototype that can simulate the entire working process of the equipment and conduct batch simulations.

[0087] In this embodiment, the simulation platform is a platform for conducting simulation modeling relying on the functions of professional simulation software, and the simulation virtual prototype is a co-simulation model built by the user with the help of professional simulation software, and this model can conduct performance index analysis.

[0088] In this embodiment, considering the extreme environment where the equipment is located and the common working modes under this working condition, a simulation virtual prototype that can cover performance indicators in multiple fields is built. The simulation models in multiple fields include simulation models in the fields of structural dynamics, fluid transmission, and planning control. On this basis, the simulation virtual prototype can simulate and analyze indirect performance indicators such as the working efficiency, execution accuracy, structural strength, vibration conditions, and wear and tear of the equipment, and then evaluate the overall working effectiveness, performance reliability, and structural durability of the equipment.

[0089] Figure 3 A schematic block diagram of the simulation virtual prototype in the embodiment of the present invention.

[0090] As Figure 3 shown, step S2 includes the following sub-steps:

[0091] Step S2.1, establish a "machine - liquid - control" co-simulation model that describes the action execution of the equipment itself, and use model interchange technology to unify the simulation models in multiple fields into a feedback control model. This "machine - liquid - control" co-simulation model can analyze performance indicators such as the transmission efficiency and action execution accuracy of the equipment.

[0092] In this embodiment, the simulation models in multiple fields include simulation models in the fields of mechanical power and hydraulic transmission.

[0093] Specifically, the simulation virtual prototype uses joint simulation technologies such as model interchange and co-simulation to combine the simulation models in multiple fields obtained in the previous step into a whole. By establishing a transmission path for models and data, multiple simulation software can carry out co-simulation collaboratively, so as to more accurately simulate the entire process of equipment operation and analyze the performance indicators during equipment operation.

[0094] Step S2.2, establish a "mechanism - environment" co-simulation model that describes the real-time interaction between the equipment and the environment, perform "machine - liquid - control" coupled modeling on the mechanism, perform finite element or discrete element modeling on the environment, and use co-simulation technology to couple the two to form a preliminary simulation virtual prototype.

[0095] In this embodiment, the mechanism is the mechanical structure of the equipment, which refers to a three-dimensional geometric model that describes the mechanical structure of the equipment in the simulation. Specifically, in the simulation, the geometric model of the equipment represents the equipment, and the particle bed represents the environment.

[0096] In the actual working condition, it is the equipment that acts on the environment (such as a roadheader cutting a rock wall), so co-simulation is required to couple the equipment model based on "machine - liquid - control" coupled modeling and the environment model based on "finite element or discrete element" modeling.

[0097] Specifically, the simulation virtual prototype uses model interchange and co-simulation technologies to combine simulation models in multiple fields into a whole. The simulation virtual prototype includes an "environment-mechanism" coupled simulation model to simulate the interaction process between the equipment and the environment. The mechanism includes a "mechanical-hydraulic-control" coupled simulation model to simulate the real process of the hydraulic drive mechanism movement. Joint simulation technologies such as model interchange and co-simulation are used for model and data transfer during coupled simulation. After clarifying the performance indicators and influencing factors to be analyzed, the simulation virtual prototype can support batch simulation analysis based on experimental design.

[0098] Step S2.3: Based on the design parameters of the equipment, the usage environment, and the design requirements in the working environment, adjust the preliminary simulation virtual prototype to obtain an adjusted simulation virtual prototype.

[0099] Step S2.4: Conduct model verification on the adjusted simulation virtual prototype. After passing the verification, clarify the indirect performance indicators of the equipment that need to be obtained through simulation analysis. The indirect performance indicators of the equipment are used to measure the working efficiency, performance robustness, and structural durability of the equipment, such as the working efficiency, execution accuracy, structural strength, wear and tear, etc. of the equipment.

[0100] Step S2.5: Take the indirect performance indicators of the equipment as the simulation target, and set the influencing factors of the indirect performance indicators of the equipment as the simulation input conditions. Conduct Latin hypercube experimental design on the simulation input conditions to form a simulation experimental design table.

[0101] In this embodiment, a roadheader is used as an example. After modeling the feedback module, hydraulic module, and power module respectively using software such as SIMULINK, AMESIM, and ADAMS, the hydraulic module and the power model are both encapsulated into FMU modules using model interchange technology and used as an element in the feedback control system. Running the main control software SIMULINK can perform the "machine-hydraulic-control" coupled simulation of the equipment. At the same time, the discrete medium of the coal seam faced by the equipment is modeled using the discrete element software EDEM, and a co-simulation path is established between the environment model and the equipment model using co-simulation technology to jointly form a simulation virtual prototype for the roadheader equipment.

[0102] Step S3: Construct a digital twin: Visualize the appearance design and internal physical model of the equipment in the simulation virtual prototype, and integrate the real-time data collected by sensors to initially construct a digital twin. By continuously integrating real-time data, dynamically update the state of the digital twin.

[0103] Figure 4 It is a schematic block diagram for initially establishing the digital twin in the embodiment of the present invention.

[0104] In this embodiment, the digital twin is a digital model that describes the working state of the equipment by using visualization technology in the virtual engine Unity3D, based on the multi-domain simulation virtual prototype formed in step S2 and the sensors originally installed at the necessary parts of the equipment.

[0105] As Figure 4 shown, step S3 includes the following sub-steps:

[0106] Step S3.1: Import the 3D geometric model and hydraulic schematic diagram of the equipment in the simulation virtual prototype verified in step S2 into the Unity3D platform. Based on the scene construction function of Unity3D, create the external form, internal structure, and composition of each functional module of the equipment, and visualize the created equipment in the 3D scene through Mesh rendering to obtain a preliminary digital twin.

[0107] Specifically, based on the 3D geometric model, hydraulic transmission model, etc. of the simulation virtual prototype, integrate the interface of the real-time data of the sensor, and build a preliminary digital twin model in Unity3D through visualization technology. The model can reflect the appearance and internal composition of the equipment, and based on kinematic modeling, directly observe information such as the position, running posture, and load of the equipment in real time.

[0108] Step S3.2: Combine the dynamic simulation in the simulation virtual prototype with the preliminary digital twin, and use C# script to integrate the structural dynamics model of the equipment into the physical engine Rigidbody of Unity3D to obtain a digital twin with dynamic movement, so that the digital twin can dynamically reflect the movement mechanism of the equipment;

[0109] Step S3.3: On the digital twin with dynamic movement, due to the limitation of the extreme environment, the sensor is set at the part of the equipment that does not contact the environment. Receive the real-time data of the equipment's sensor in real time through the Socket communication of Unity3D, and map the real-time data of the sensor to each corresponding component of the digital twin with dynamic movement to obtain a digital twin integrated with real-time sensing data.

[0110] In this embodiment, the sensor data is observable data such as pressure, vibration, and displacement at the sensor installation part.

[0111] Specifically, combine the formed simulation virtual prototype, import the external appearance design and internal physical model of the equipment into Unity3D, and perform rendering and simulation to simulate the movement and physical behavior of the equipment. On the other hand, develop an interface for real-time receiving of environmental sensor data, obtain information such as pressure and displacement, analyze and solve them, combine the sensing data with the equipment model on the Unity3D platform, update the equipment state in real time, and display the direct observation information of the equipment in a graphical form.

[0112] Step S3.4: Update the state of the digital twin with real-time sensor data and display it using the real-time rendering function of Unity3D, ensuring that the digital twin can display the appearance design, internal structure, and motion mechanism of the equipment, and update the directly observable data such as the position, speed, and vibration of the equipment in real time.

[0113] Step S4: Batch simulation and adaptive construction of the surrogate model: Based on the set input conditions, the simulation virtual prototype conducts batch simulations to generate simulation results, and trains a surrogate model using the adaptive construction strategy with the simulation results, outputting the trained surrogate model.

[0114] Figure 5 It is a schematic block diagram of the adaptive construction strategy of the surrogate model in the embodiments of the present invention.

[0115] For various performance indicators of the equipment, the surrogate model is trained and the hyperparameters are optimized based on the dataset of batch simulations using the adaptive construction strategy. The surrogate model is provided with batch training data by the simulation virtual prototype, avoiding time-consuming and costly experimental analyses on equipment in extreme environments, clarifying various performance indicators to be evaluated for the equipment and their respective influencing factors. In the simulation virtual prototype, the performance indicators are used as analysis results, and the influencing factors are used as input conditions. Batch simulation analysis is carried out based on the experimental design of the input conditions, and the results of batch simulations of the virtual prototype are used for training for subsequent training of the surrogate model.

[0116] As Figure 5 shown, Step S4 includes the following sub-steps:

[0117] Step S4.1: Based on the simulation experimental design table formed in Step S2, conduct batch simulations in the simulation virtual prototype, export and organize the simulation results to obtain a simulation dataset.

[0118] In this embodiment, since multiple performance indicators with different characteristics need to be evaluated and different models are used to evaluate different performances, additional model comparison work is required during each training. Therefore, the present invention adopts a combined strategy to directly construct a highly versatile combined surrogate model. Four surrogate models with good performance in evaluating conventional data are selected to form the base learners.

[0119] Step S4.2: Construct an adaptive surrogate model to establish a robust mapping relationship between the multi-performance indicators of the equipment and their influencing factors; the surrogate model includes four base learners: support vector regression (SVR), radial basis neural network (RBF), random forest (RF), and K-nearest neighbor regression (KNN).

[0120] Step S4.3: On the simulation data sets corresponding to each performance metric, train the four base learners, and taking the accuracy of model evaluation as the goal, use the Particle Swarm Optimization (PSO) algorithm to optimize the hyperparameters of the four base learners. Among them, the hyperparameters of SVR are the penalty parameter c and the allowable error limit ε, the hyperparameters of RBF are the radial basis influence range γ and the radial basis standard deviation σ, the hyperparameters of RF are the number of random trees n and the maximum depth m, and the hyperparameters of KNN are the k value and the weight w, to obtain the optimized base learners.

[0121] The surrogate model has an adaptive construction strategy. The optimal types and hyperparameters of the surrogate models required to evaluate each performance metric are different. For each performance metric, four base learners, namely Support Vector Regression (SVR), Radial Basis Function Neural Network (RBF), Random Forest (RF), and K-Nearest Neighbor Regression (KNN), are constructed. Each learner has different advantages and will be flexibly selected according to the characteristics of the performance metric in the follow-up; at the same time, for different working conditions, the Particle Swarm Optimization (PSO) algorithm is introduced to optimize and adjust the hyperparameters of each learner during training, so as to improve the adaptability and accuracy of the model in different situations and ensure that the surrogate model can adapt to changing working conditions and extreme environments. During this process, the training parameters obtained by the same base learner are different when targeting different performance metrics.

[0122] Step S4.4: For each performance metric to be evaluated, four surrogate models are obtained based on the optimized base learners with hyperparameters.

[0123] Specifically, when the same surrogate model evaluates different performance metrics, the training parameters it takes are also different.

[0124] Step S5: Ensemble learning and improvement of the surrogate model: Adopt an ensemble learning strategy to integrate multiple trained surrogate models optimized with hyperparameters to form an ensemble learner that can perform offline evaluation of multiple performances of the equipment under different working conditions.

[0125] Figure 6 It is a schematic block diagram of the self-enhancing ensemble mechanism of the surrogate model in the embodiment of the present invention.

[0126] This embodiment proposes an ensemble method for surrogate models based on the Gradient Boost strategy. With different ensemble strategies for surrogate models, the four optimized surrogate models (SVR, RBF, RF, KNN) are combined into a new ensemble learner. This ensemble learner can combine the advantages of the four base learners to achieve more accurate evaluation for specific performance metrics. Adopting the Gradient Boost strategy, gradually improve the evaluation ability of the model, and adaptively adjust the training parameters of the base learners and the weights of the learner combination to meet diverse performance evaluation requirements.

[0127] As Figure 6 shown, step S5 includes the following sub-steps:

[0128] Step S5.1: For each performance metric, randomly and initially select a base learner from the trained surrogate models, use the base learner to preliminarily evaluate the performance, and calculate the residual between the evaluation result and the actual value;

[0129] Step S5.2: Based on the residual, adjust the training parameters of the remaining base learners, and retrain the remaining base learners. The training objective is to minimize the squared error of the residual;

[0130] Step S5.3: Combine the initially selected base learner and the remaining adjusted base learners in a weighted manner, and optimize the overall prediction result by weighting. The weight of each base learner depends on its improvement degree of the residual;

[0131] Step S5.4: Iteratively execute steps S5.1 to S5.3. Each round of training focuses on the residual generated in the previous round of training and improves the residual. Finally, an ensemble learner is formed. The ensemble learner is formed by adopting a step-by-step error correction strategy, which can improve the overall robustness evaluation ability of the model and prevent overfitting. In practical applications, it is reflected in the ability to handle complex environmental changes and sensor data noise, and provide stable and high-precision evaluation outputs.

[0132] This embodiment proposes a self-enhanced surrogate model integration mechanism. For each performance metric to be evaluated, a gradient boosting (Gradient Boost) ensemble learning strategy is adopted for the four surrogate models (SVR, RBF, RF, KNN) optimized for hyperparameters. This mechanism is optimized and iterated based on the residual, gradually adjusts the training parameters of the base learners, and finally forms an ensemble learner. Compared with a single base learner, this ensemble learner can significantly improve the goodness of fit and accuracy of performance evaluation. Specifically, the gradient boosting strategy corrects the training parameters of each base learner by focusing on the part predicted wrongly by the previous base learner, and continuously improves the overall prediction accuracy of the combined model. Gradient boosting integrates the outputs of multiple base learners by weighting, improving the evaluation accuracy of the model when facing complex data.

[0133] In this embodiment, a virtual prototype under the simulation platform is formed, specifically including: based on the actual working conditions and analysis requirements of the equipment, a multi-domain simulation model capable of sharing model and data circulation is built, which together form a virtual prototype for offline analysis of various performances of the equipment; a digital twin under the data platform is formed, combining the model of the virtual prototype and the sensing data collected in real time by the equipment, and a data-driven equipment digital twin is initially constructed, which can calculate and display the direct observation information of the equipment in real time; an offline performance evaluation method is formed. An adaptive combined surrogate model based on ensemble learning is proposed and trained through the results of batch simulations of the multi-domain simulation virtual prototype. Under the working conditions of different extreme environments, the performance indicators of the equipment are robustly evaluated; a real-time performance evaluation mechanism based on the twin platform is formed. The sensing data and the surrogate model are integrated in the digital twin. The surrogate model can use the real-time collected sensing data to carry out real-time performance evaluation, promoting the healthy operation of major equipment.

[0134] Step S6: Form a performance evaluation mechanism for the digital twin, combining the ensemble learner with the real-time sensor data in the digital twin to form a real-time evaluation method for various indirect performances of the equipment.

[0135] In this embodiment, the ensemble learner is the result of ensemble learning of four base learners and is used as a surrogate model.

[0136] Figure 7 It is a schematic block diagram of the dual-drive evaluation mechanism formed by the surrogate model and the sensing data in the embodiment of the present invention.

[0137] As Figure 7 shown, step S6 includes the following sub-steps:

[0138] Step S6.1: Package the ensemble learner formed in step S5 using C# code and ensure that it can be called in the code of the digital twin.

[0139] Step S6.2: Based on the sensor data interface established by the digital twin, receive the real-time sensor data, such as pressure, vibration and displacement. The digital twin automatically identifies and adjusts the timeliness of the data and uses it as the input of the ensemble learner.

[0140] Step S6.3: Deeply process and extract features from the sensor data through the ensemble learner, and real-time evaluate various indirect performance indicators of the equipment, including working efficiency, structural strength and durability. During the real-time evaluation process, the surrogate model dynamically updates its evaluation results according to the data changes under different working conditions to ensure that the digital twin can timely reflect the operating state of the equipment.

[0141] Step S6.4, through visualization technology, the evaluation results of the digital twin integrated with the surrogate model on multiple performance indicators of the equipment are displayed in real time, helping the operator understand the current performance of the equipment and preparing for their operation decision-making.

[0142] A performance evaluation mechanism is constructed in the digital twin. The real-time performance evaluation mechanism is to add an adaptive combined surrogate model based on ensemble learning to the digital twin platform to form a real-time multi-performance evaluation method for equipment driven by both an "offline model - real-time data". Specifically, the surrogate model quickly evaluates multiple performance indicators of the equipment based on the input working conditions offline, while the sensor data reflects the current working state and environmental changes of the equipment in real time. An interface for information flow and interaction is constructed between the surrogate model and the sensor data. The surrogate model receives the data collected by the sensor in real time, conducts synchronous evaluation of multiple performances of the equipment, and realizes real-time interaction between the surrogate model and the sensor data by developing corresponding data interfaces. The sensor drives the surrogate model to update the performance evaluation results, and the evaluation results are integrated into the digital twin platform through visualization technology, finally forming a real-time digital twin performance evaluation mechanism driven by both "model - data".

[0143] So far, the digital twin has constructed a dual-drive mechanism of "model - data". Based on the modeling and simulation of the real physical entity and the twin construction, the twin can provide real-time and accurate multi-performance evaluation for the equipment.

[0144] Embodiment 2:

[0145] The present invention also provides a digital twin construction and multi-performance real-time evaluation system for equipment. The digital twin construction and multi-performance real-time evaluation system for equipment can be implemented by executing the process steps of the digital twin construction and multi-performance real-time evaluation method for equipment. That is, those skilled in the art can understand the digital twin construction and multi-performance real-time evaluation method for equipment as the preferred implementation manner of the digital twin construction and multi-performance real-time evaluation system for equipment.

[0146] Specifically, the digital twin construction and multi-performance real-time evaluation system for equipment includes:

[0147] Module M1, building multiple domain simulation models: Decompose the functional modules of the equipment according to the actual working conditions, and build multiple domain simulation models that can analyze the working mechanism and indirect performance indicators of the equipment;

[0148] Module M2, constructing a simulation virtual prototype: Integrate multiple domain simulation models using various co-simulation technologies to construct a simulation virtual prototype that can simulate the entire working process of the equipment and conduct batch simulations;

[0149] Module M3, constructing a digital twin: Visualize the appearance design and internal physical model of the equipment in the simulation virtual prototype, integrate the real-time data collected by sensors, initially construct a digital twin, and dynamically update the state of the digital twin by continuously integrating real-time data;

[0150] Module M4, batch simulation and adaptive construction of surrogate models: Based on the set input conditions, the simulation virtual prototype conducts batch simulations, generates simulation results, and trains a surrogate model using an adaptive construction strategy with the simulation results, and outputs the trained surrogate model;

[0151] Module M5, integrated learning and improvement of surrogate models: Integrate multiple trained surrogate models optimized by hyperparameters using an integrated learning strategy to form an integrated learner that can perform offline evaluations of multiple performances of the equipment under different working conditions;

[0152] Module M6, forming a performance evaluation mechanism for the digital twin, combining the integrated learner with the real-time data of the sensors in the digital twin to form a real-time evaluation method for various indirect performances of the equipment.

[0153] Those skilled in the art know that in addition to implementing the system and its various devices, modules, and units provided by the present invention in the form of pure computer-readable program code, the method steps can be logically programmed to enable the system and its various devices, modules, and units provided by the present invention to be implemented in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers to achieve the same functions. Therefore, the system and its various devices, modules, and units provided by the present invention can be regarded as a hardware component, and the devices, modules, and units included therein for implementing various functions can also be regarded as the structure within the hardware component; the devices, modules, and units for implementing various functions can also be regarded as either software modules for implementing the method or the structure within the hardware component.

[0154] The specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the above specific embodiments, and those skilled in the art can make various changes or modifications within the scope of the claims, which does not affect the essence of the present invention. Without conflict, the embodiments of the present application and the features in the embodiments can be combined arbitrarily with each other.

Claims

1. A method for constructing a digital twin of a device and real-time evaluation of multiple performances, characterized in that, Including the following steps: Step S1, building multiple domain simulation models: decompose the functions of the equipment according to the actual working conditions, and build multiple domain simulation models that can analyze the working mechanism and indirect performance indicators of the equipment; Step S2, constructing a simulation virtual prototype: integrate the multiple domain simulation models by using a variety of co-simulation technologies to construct a simulation virtual prototype that can simulate the entire working process of the equipment and perform batch simulations; Step S3, constructing a digital twin: visualize the appearance design and internal physical model of the equipment in the simulation virtual prototype, and integrate the real-time data collected by sensors to initially construct a digital twin, and dynamically update the state of the digital twin by continuously integrating real-time data; Step S4, batch simulation and adaptive construction of the surrogate model: based on the set input conditions, the simulation virtual prototype conducts batch simulations to generate simulation results, and trains a surrogate model using an adaptive construction strategy with the simulation results, and outputs the trained surrogate model; Step S5, integrated learning and improvement of the surrogate model: integrate multiple trained surrogate models optimized by hyperparameters using an integrated learning strategy to form an integrated learner that can perform offline evaluation of multiple performances of the equipment under different working conditions; Step S6, forming a performance evaluation mechanism for the digital twin, combining the integrated learner with the real-time data of the sensors in the digital twin to form a real-time evaluation method for various indirect performances of the equipment.

2. The digital twin construction and multi-performance real-time evaluation method of a kind of equipment according to claim 1, characterized in that, The step S1 includes the following sub-steps: Step S1.1, in combination with the environmental working conditions and actual conditions, conduct a detailed analysis of the working principle and operating mechanism of the equipment to form a three-dimensional geometric model diagram, a hydraulic transmission schematic diagram, and a working mechanism specification; Step S1.2, according to the function analysis of the equipment, decompose the equipment modularly, divide the equipment into multiple different domains, so as to select professional software in each domain respectively for simulation modeling; Step S1.3, according to the characteristics of each module domain, select an appropriate simulation method and establish a corresponding simulation model, and the simulation model reflects the working mechanism and various performance indicators of the equipment in the domain; Step S1.4, based on the design parameters of the equipment, the use environment, and the design requirements in the working environment, adjust the multiple domain simulation models to ensure that indirect performance indicators that cannot be directly obtained by sensors can be analyzed through the multiple domain simulation models; Step S1.5, verify the multiple domain simulation models according to experimental data or theoretical analysis.

3. The digital twin construction and multi-performance real-time evaluation method of a piece of equipment according to claim 1, characterized in that, The step S2 includes the following sub-steps: Step S2.1, establish a "machine-hydraulic-control" co-simulation model that describes the action execution of the equipment itself, and use the model interchange technology to unify the multiple domain simulation models into a feedback control model; Step S2.2, establish a "mechanism-environment" co-simulation model that describes the real-time interaction between the equipment and the environment, perform "machine-hydraulic-control" coupling modeling on the mechanism, perform finite element or discrete element modeling on the environment, and use co-simulation technology to couple the two to form a preliminary simulation virtual prototype; Step S2.3: Based on the design parameters of the equipment, the usage environment, and the design requirements in the working environment, adjust the preliminary simulation virtual prototype to obtain an adjusted simulation virtual prototype. Step S2.4: Conduct model verification on the adjusted simulation virtual prototype. After passing the verification, clarify the indirect performance indicators of the equipment that need to be obtained through simulation analysis. The indirect performance indicators of the equipment are used to measure the working efficiency, performance robustness, and structural durability of the equipment. Step S2.5: Use the indirect performance indicators of the equipment as the simulation target, and set the influencing factors of the indirect performance indicators of the equipment as the simulation input conditions. Conduct Latin hypercube experimental design on the simulation input conditions to form a simulation experimental design table.

4. A digital twin construction and multi-performance real-time evaluation method for an equipment according to claim 1, characterized in that, In step S3, the digital twin is a digital model that describes the working state of the equipment constructed by using visualization technology in the virtual engine Unity3D based on the multi-domain simulation virtual prototype formed in step S2 and the sensors originally installed at the necessary parts of the equipment.

5. The digital twin construction and multi-performance real-time evaluation method of a piece of equipment according to claim 4, characterized in that, Step S3 includes the following sub-steps: Step S3.1: Import the three-dimensional geometric model and hydraulic schematic diagram of the equipment in the simulation virtual prototype verified in step S2 into the Unity3D platform. Based on the scene construction function of Unity3D, create the external form, internal structure, and composition of each functional module of the equipment, and visualize the created equipment in the three-dimensional scene through Mesh rendering to obtain a preliminary digital twin. Step S3.2: Combine the dynamic simulation in the simulation virtual prototype with the preliminary digital twin. Use C# script to integrate the structural dynamics model of the equipment into the physical engine Rigidbody of Unity3D to obtain a digital twin with dynamic motion, enabling the digital twin to dynamically reflect the motion mechanism of the equipment. Step S3.3: On the digital twin with dynamic motion, receive the real-time sensor data of the equipment in real time through the Socket communication of Unity3D, and map the real-time sensor data to the corresponding components of the digital twin with dynamic motion to obtain a digital twin integrated with real-time sensing data. Step S3.4: Use the real-time sensor data to update the state of the digital twin and display it using the real-time rendering function of Unity3D.

6. A digital twin construction and multi-performance real-time evaluation method for an equipment according to claim 1, characterized in that, In step S4, for various performance indicators of the equipment, the surrogate model is trained and the hyperparameters are optimized based on the batch simulation dataset using an adaptive construction strategy. The surrogate model is provided with batch training data by the simulation virtual prototype, avoiding time-consuming and costly experimental analysis of equipment in extreme environments. Clarify various performance indicators of the equipment to be evaluated and their respective influencing factors. In the simulation virtual prototype, use the performance indicators as the analysis results and the influencing factors as the input conditions, conduct batch simulation analysis based on the experimental design of the input conditions, and use the results of the batch simulation of the virtual prototype for training, which is used for the subsequent training of the surrogate model.

7. A method for constructing a digital twin of an equipment and real-time multi-performance evaluation according to claim 6, characterized in that, Step S4 includes the following sub-steps: Step S4.1: Based on the simulation test design table formed in Step S2, conduct batch simulations in the simulation virtual prototype, export and organize the simulation results to obtain a simulation data set. Step S4.2: Construct an adaptive surrogate model to establish a robust mapping relationship between multiple performance indicators of the equipment and their influencing factors; the surrogate model includes four basic learners: support vector regression, radial basis neural network, random forest, and K-nearest neighbor regression. Step S4.3: On the simulation data set corresponding to each performance indicator, train the four basic learners, and use the particle swarm optimization algorithm to optimize the hyperparameters of the four basic learners with the accuracy of model evaluation as the goal. Among them, the hyperparameters of the SVR are the penalty parameter c and the tolerance error limit ε, the hyperparameters of the RBF are the radial basis influence range γ and the radial basis standard deviation σ, the hyperparameters of the RF are the number of random trees n and the maximum depth m, and the hyperparameters of the KNN are the k value and the weight w, to obtain the optimized basic learners. Step S4.4: Based on the optimized basic learners, obtain the trained surrogate model.

8. A digital twin construction and multi-performance real-time evaluation method for an equipment according to claim 1, characterized in that The said Step S5 includes the following sub-steps: Step S5.1: For each performance indicator, initially select a basic learner from the trained surrogate model, use the basic learner to conduct a preliminary evaluation of the performance, and calculate the residual between the evaluation result and the actual value. Step S5.2: Based on the residual, adjust the training parameters of the remaining basic learners, and retrain the remaining basic learners. The training goal is to minimize the squared error of the residual. Step S5.3: Combine the initially selected basic learner and the remaining adjusted basic learners through weighting to optimize the overall prediction result by weighting. The weight of each basic learner depends on its improvement degree of the residual. Step S5.4: Iteratively execute Step S5.1 to Step S5.

3. Each round of training focuses on the residual generated in the previous round of training and improves the residual, and finally forms an ensemble learner.

9. The digital twin construction and multi-performance real-time evaluation method of a kind of equipment according to claim 1, characterized in that, The said Step S6 includes the following sub-steps: Step S6.1: Package the ensemble learner formed in Step S5 using C# code and ensure that it can be called in the code of the digital twin. Step S6.2: Based on the sensor data interface established by the digital twin, receive real-time sensor data, and the digital twin performs timeliness identification and adjustment on the data as the input of the ensemble learner. Step S6.3: Deeply process and extract features from the sensor data through the ensemble learner to evaluate multiple indirect performance indicators of the equipment in real time, including working efficiency, structural strength, and durability. Step S6.4: Through visualization technology, display the evaluation results of the digital twin on the multiple performance indicators of the equipment in real time.

10. A digital twin construction and multi-performance real-time evaluation system for an equipment, characterized in that Including: Module M1: Build multiple domain simulation models: Decompose the functions of the equipment according to the actual working conditions, and build multiple domain simulation models that can analyze the working mechanism and indirect performance indicators of the equipment. Module M2, constructing a simulation virtual prototype: integrating the simulation models of the multiple fields by using a variety of co-simulation technologies to construct a simulation virtual prototype that can simulate the entire working process of the equipment and conduct batch simulations; Module M3, constructing a digital twin: visualizing the appearance design and internal physical model of the equipment in the simulation virtual prototype, integrating the real-time data collected by sensors, initially constructing a digital twin, and dynamically updating the state of the digital twin by continuously integrating the real-time data; Module M4, adaptive construction of batch simulations and surrogate models: based on the set input conditions, the simulation virtual prototype conducts batch simulations, generates simulation results, and trains a surrogate model using an adaptive construction strategy with the simulation results, and outputs the trained surrogate model; Module M5, integrated learning and improvement of the surrogate model: integrating multiple trained surrogate models optimized by hyperparameters using an integrated learning strategy to form an integrated learner that can perform offline evaluations of multiple performances of the equipment under different working conditions; Module M6, forming a performance evaluation mechanism for the digital twin, combining the integrated learner with the real-time data of the sensors in the digital twin to form a real-time evaluation method for multiple indirect performances of the equipment.

Images