Development of a Modular Virtual Calibration Platform for Power Machinery

Abstract

The hardware-in-the-loop (HIL)-based virtual calibration technology for power machinery significantly enhances the efficiency and flexibility of electronic controller calibration, reducing dependency on physical experiments and lowering development costs. This paper presents an independently developed modular virtual calibration platform, designed based on the ISO/IEC 42010 architecture approach and the V-Model development process. The platform adopts a modular layered architecture comprising the physical layer, communication layer, model layer, software layer, and application layer. Each layer is independently developed and seamlessly integrated, ensuring excellent scalability and maintainability. Through performance validation conducted under steady-state and transient operating conditions of the diesel/natural gas dual-fuel engine electronic control unit (ECU), experimental results demonstrated that the platform possesses high computational accuracy, rapid response capability, and stable operation across different scenarios. The experiments validated that the platform can effectively replicate the real behavior of the controlled units in a virtual environment, confirming its adaptability and reliability. The platform offers an efficient and reliable solution for power machinery controller calibration.

1. Introduction

With the complexity of power machinery systems continuing to grow, traditional controller calibration methods face challenges such as prolonged durations, high resource demands, and difficulty meeting real-time requirements, particularly under multivariable conditions [1,2]. Against this backdrop, virtual calibration technology has emerged as a promising alternative. By optimizing and validating controller functionality in a simulation environment, this technology reduces reliance on physical experiments, enhances development efficiency, and lowers costs [3,4].

In recent years, HIL-based virtual calibration technology has been widely studied and applied in various fields [5,6]. Initially utilized in engine applications, Lee et al. [7] proposed a virtual calibration method integrating engine, after-treatment, and vehicle models on a dSPACE platform to address real-driving emissions (RDEs) regulations. By employing online optimization algorithms and multi-objective control strategies, the method improved controller performance under dynamic conditions. Tziolas et al. [8] further developed a high-precision transient emission prediction tool leveraging real-time interactions among powertrain components via a dSPACE platform. This tool demonstrated the potential of virtual calibration in accurately simulating combustion dynamics and improving emission prediction under transient conditions.

With the growing adoption of electrification technologies, HIL simulation has extended into the realms of electric and hybrid vehicles. Lajunen et al. [9] developed a real-time co-simulation and testing platform encompassing battery management systems (BMSs), motor control units (MCUs), and vehicle control units (VCUs) to validate optimization designs and fault diagnosis schemes for electric drive systems. Hentunen et al. [10] proposed a general HIL validation environment for heavy-duty hybrid systems, utilizing the Simulink and dSPACE platform to optimize energy management strategies. In the fuel cell vehicle domain, Gauchia and Sanz [11] constructed a fuel cell hybrid vehicle simulation platform on the dSPACE platform, analyzing energy distribution and system dynamics to improve stability and energy management efficiency. Moreover, Guo et al. [12] developed a predictive energy management strategy for fuel cell plug-in hybrid vehicles (FC PHEVs) on the Simulink and dSPACE platforms, achieving significant improvements in real-time energy management coordination and fuel economy.

Although HIL platforms have made significant progress in the automotive domain, recent studies [13,14,15] indicate that advancing technology and the increasing complexity of control systems are imposing greater performance demands on HIL systems. These demands include integrating various hardware and software modules, supporting complex control strategies, and adapting to dynamic multivariable scenarios. However, most existing HIL platforms adopt a closed architecture, leading to numerous limitations in hardware compatibility, module scalability, and adaptability. For instance, such platforms often require substantial costs and intricate adjustment processes to accommodate new requirements, making it challenging for them to respond quickly to evolving development tasks. The ever-increasing complexity of control demands in the power machinery further exacerbates this issue, positioning open architectures as a key trend for addressing these challenges.

To tackle these challenges, this study proposes an independently developed modular virtual calibration platform designed to meet the combined requirements of real-time performance, accuracy, and flexibility for power machinery control systems. By incorporating the ISO/IEC 42010 [16] architecture approach and the V-Model development process, the platform adopts a layered architecture with independently developed modules integrated via standardized interfaces. This open design significantly enhances the platform’s scalability and system compatibility, enabling superior adaptability across diverse scenarios. Compared to existing simulation platforms, the proposed platform not only maintains advantages in real-time performance and computational accuracy but also reduces system upgrade complexity and improves hardware-software reusability through modular design. Consequently, this approach effectively addresses limitations in the flexibility and scalability of current platforms, providing novel insights and technical support for the practice of virtual calibration in power machinery.

The remainder of this paper is organized as follows. Section 2 introduces the methodologies employed, including the ISO/IEC 42010 architecture approach and the V-Model development process. Section 3 elaborates on the platform’s modular architecture design, detailing the functions and implementations of the physical layer, communication layer, model layer, software layer, and application layer. Section 4 presents the platform performance validation results. Section 5 discusses the limitations and shortcomings of the validation results. Finally, this study is concluded in Section 6.

2. Methods

2.1. V-Model Development Process

The V-Model development process is a systematic development methodology. Its core concept is to ensure the high quality and stability of the system through a closed-loop structure of development and verification. Compared with traditional linear development processes, the V-process emphasizes the close integration of development and testing. By introducing targeted verification steps at each development stage, it not only detects and corrects issues in the early stage but also significantly reduces rework, hence improving development efficiency. Moreover, the V-process ensures controlled progress through clear phase definitions, helping to reduce potential failure risks in complex system development. This approach is particularly suitable for complex systems requiring high reliability and stability [17,18].

Based on the core concept of the V-Model development process, the development process of the virtual calibration system in this study is divided into five stages, which are requirement analysis, architecture design, system implementation, module testing, and functional verification, as shown in Figure 1. The development process begins with the requirement analysis phase, where the system’s objectives and functional requirements are defined. This is followed by the architecture design phase, in which the system is resolved into multiple functional modules, and their interactions are elaborately defined. In the system implementation phase, the designs are realized, and the system is integrated. The module testing phase focuses on verifying the functionality of each individual module. Finally, the functional verification phase assesses the system as a whole through comprehensive performance evaluation and testing under real-world application conditions.

The V-Model development process is applied throughout the entire development cycle, providing clear guidance during the requirement analysis and architecture design stages, while ensuring logical and complete development efforts during system implementation and testing. By introducing targeted verification steps at each stage, the designed virtual calibration system is proven to have perfect abilities in real-time simulation accuracy, system adaptability, and modular integration. This approach effectively enhances development efficiency, mitigates potential risks, and achieves comprehensive the optimization of performance, reliability, and scalability.

2.2. Architecture Design Method

With the increasing complexity of modern engineering and software systems, traditional architecture design methods often rely on experience-driven approaches, lacking systematic guidance. This can lead to issues such as requirement omissions, design conflicts, and inadequate adaptability of architectures [19,20]. The ISO/IEC/IEEE 42010:2011—Systems and Software Engineering—Architecture Description, provides a systematic methodology for designing and describing the architecture of complex systems. This standard is widely applied in embedded systems, industrial automation, and information systems, offering theoretical support for the development of complex systems that integrate software and hardware [21,22].

• Stakeholders are any individuals or groups directly or indirectly affected by the system during development. ISO/IEC 42010 emphasizes that system design should consider stakeholders’ perspectives and clearly define their needs and expectations. For the virtual calibration platform, the main stakeholders include developers, users, and maintainers.
• Concerns refer to specific aspects of the system that stakeholders focus on. According to ISO/IEC 42010, concerns reflect stakeholders’ specific areas of interest. In the virtual calibration platform, concerns can be categorized into functional and non-functional types. Functional concerns include key requirements such as real-time simulation capabilities, signal processing accuracy, and automated testing functions. Among non-functional concerns, maintainability and scalability are particularly important, requiring the system to minimize fault repair time and easily adapt to the introduction of new functions or modules.
• ISO/IEC 42010 uses architectural perspectives and views to resolve system design requirements into easily understandable and analyzable components, ensuring that architectural design addresses key development aspects. By integrating the V-Model development process with the need to consider system functionality, modular design, and verification requirements, this study analyzes and describes the architecture from both structural and functional perspectives.

The structural perspective focuses on the physical components and logical layering of the system, aiming to improve scalability and maintainability through modular design. As illustrated in Figure 2a, the virtual calibration platform developed in this study adopts a five-layer architecture, which includes the physical layer, communication layer, software layer, model layer, and application layer. The physical layer provides foundational hardware support, while the communication layer facilitates information exchange between modules. The software layer links the hardware to simulation models, and the model layer creates the virtual simulation environment. Finally, the application layer offers the user interface. These layers are interconnected through standardized interfaces, allowing for independent module development and seamless integration.

The functional perspective focuses on the overall functionality of the system and its verification processes, ensuring that the platform meets the requirements for complex simulation and calibration tasks. As shown in Figure 2b, this study implements a multi-level verification process, which ranges from basic functional testing to comprehensive performance evaluation. This process includes module function verification, software model verification, unit collaboration verification, system integration verification, and platform function verification. Through this verification process, the platform’s functionality and reliability in the field of power machinery virtual calibration are thoroughly ensured.

3. Modular System Architecture Design

3.1. Physical Layer

The physical layer forms the foundation of the virtual calibration platform and consists of hardware devices that provide an environment for physical signal acquisition, processing, and generation. Guided by modular design principles, the physical layer is divided into seven functional modules, as shown in

Table 1. Overview of physical layer modules.

Module Name	Module Overview
Real-Time Simulation and Computing Platform	Built on the National Instruments (NIs) Peripheral Component Interconnect eXtensions for Instrumentation (PXI) system with Controller, data acquisition (DAQ) module, field-programmable gate array (FPGA), and Controller Area Network (CAN) cards, the platform offers high-performance computing for real-time simulation, signal acquisition, and multi-channel communication, meeting the requirements of high-precision dynamic testing.
Signal Interface and Conditioning Module	Facilitates signal conversion, conditioning, and isolation, enabling the high-precision processing of voltage, current, resistance, and digital signals to ensure stability and compatibility.
Main Control and Management System	Functions as the system’s central hub, integrating data analysis, control operations, and resource management while maintaining support for flexible configurations, automated testing, and user-friendly operation.
Power Management Module	Equipped with PDU to deliver stable and reliable power output for both strong and weak currents, meeting the system’s power requirements.
Programmable DC Power Supply	TDK-Lambda GEN50-30 high-precision power supply module allows flexible parameter configuration and simulates various power supply conditions for complex power system testing.
Signal Distribution and Adaptation Module	Features Breakout Box devices to enable flexible signal input and output, accommodating diverse testing connections.
Load Drawer	Provides realistic dynamic loads, including components such as injectors, solenoid valves, and signal lights, to support precise calibration and response verification of controllers.

, including a real-time simulation and computing platform, signal interface and conditioning module, main control and management system, power management module, programmable direct current (DC) power supply, signal distribution and adaptation module, and load drawer. Each module is designed to operate independently while working collaboratively, greatly enhancing the system’s scalability and maintainability. This modular design also facilitates easy functionality expansion and hardware upgrades.

The functionalities of each module are as follows. The real-time simulation and computation platform is responsible for core system computation by running simulation models and achieving real-time data simulation and acquisition through various cards. The signal interface and conditioning module integrates operational amplifier circuits, current acquisition elements, and resistive components to achieve signal conversion and conditioning, meeting the system’s requirements for signal accuracy and compatibility. The main control and management system, with the software layer, performs system control, data processing, and analysis while offering operational support and automated testing capabilities. The power management module, managed via power distribution units (PDUs), provides a safe and stable power supply for other system modules. The programmable DC power source, controlled and monitored through signal transmission with the cards, delivers a 0–50 V DC output to meet the power demands of control units under complex conditions. The signal distribution and adaptive module flexibly enable external signal access and output, while the load drawer accommodates actual physical loads. This modular design not only optimizes the system structure but also significantly enhances its flexibility and reliability. Detailed descriptions of each module are presented in Table 1.

3.1.1. Real-Time Simulation and Computing Platform

The real-time simulation and computing platform are the core modules of the physical layer, providing the virtual calibration platform with high-performance computational capabilities as well as real-time data acquisition and simulation. Built on the NI PXI platform, it integrates the Controller, DAQ, FPGA, and CAN cards to meet the requirements of signal acquisition, simulation computation, and communication.

As the central processing unit of the system, the controller is responsible for running simulation models, executing real-time computations, and managing data. The controller is equipped with an 8-core processor with a base frequency of 3.9 GHz, delivering robust computational capabilities. It ensures fast response times, seamless coordination among modules, efficient execution of simulation tasks, and precise data flow management. DAQ cards, supporting real-time acquisition and generation of analog and digital signals, provide multi-channel input/output capabilities, including 52 analog channels and 148 industrial-grade digital channels, with maximum sampling rates of 1 MS/s for analog input. FPGA cards focus on high-frequency signal generation and processing. Leveraging parallel computing capabilities, they ensure high-precision real-time responses, with a clock frequency of 40 MS/s. The CAN communication cards simulate in-vehicle CAN bus communication environments, supporting up to six CAN channels. These technical specifications collectively ensure the system meets the stringent requirements of real-time performance, computational precision, and adaptability.

3.1.2. Signal Interface and Conditioning Module

The signal interface and conditioning module is an essential component of the virtual calibration platform, enabling the conversion of external physical signals into standardized signals that the system can process, as well as transforming simulation signals into the formats required by external devices. This module is equipped with flexible interface designs that support various input and output signal types, ensuring compatibility across a wide range of testing scenarios. High precision and stability are achieved through advanced amplification, filtering, and isolation processes, meeting the rigorous quality demands of complex testing environments. Additionally, the module provides robust signal isolation and protection capabilities, effectively shielding the system from external electrical interference or device failures, thereby ensuring reliable signal processing throughout the virtual calibration workflow.

As illustrated in Figure 3 the module consists of four functional submodules. The voltage conditioning submodule regulates external analog voltage signals to match the system’s standard voltage range through operational amplifier circuits, supporting a maximum voltage of 60 V. The current acquisition submodule captures and converts analog current signals into analog voltage signals using current measurement circuits and amplifiers, providing precise input for real-time systems with a transmission range of ±50 A. The resistance simulation submodule utilizes CAN messages to control internal relays, enabling series and parallel combinations of resistance elements to generate required resistance values, with a maximum simulation resistance of 2 MΩ. Lastly, the digital signal conditioning submodule configures operating modes, reference voltages, and input thresholds through CAN messages, facilitating adjustments to digital channel operation, including level conversion, filtering, and noise reduction, thereby delivering stable and accurate digital signals for the virtual calibration system. Together, these submodules comprehensively address the platform’s signal processing requirements, significantly improving signal quality and overall system reliability.

3.2. Communication Layer

The communication layer plays a key role in the virtual calibration platform, primarily responsible for the exchange of information and data flow between the internal modules of the system, providing stable support for the platform’s overall collaboration. It achieves real-time communication between different hardware module devices through efficient signal transmission channels and precise data synchronization mechanisms, ensuring seamless cooperation among different functional modules to effectively execute complex calibration and simulation tasks.

To meet the functional requirements and data transmission characteristics of various modules in the virtual calibration platform, the communication layer employs multiple connection methods, including bus transmission, power supply, and signal line connections. Bus transmission, primarily realized through PXI, CAN, and Ethernet buses, uses specific protocols to ensure not only high-speed and stable communication but also low-latency data exchange between upper systems and lower systems, as well as between cards and controllers. Power supply, through strong and weak electrical connections, provides stable and reliable power support to control units and other hardware modules, ensuring normal operation under various conditions. Signal line connections manage the high-precision transmission of analog and digital signals, ensuring the accuracy and consistency of signal exchanges between modules during real-time simulation.

Through the effective coordination of these transmission methods, the communication layer enhances overall system coordination and provides a robust foundation for dynamic simulations under complex scenarios. The specific communication methods between modules are illustrated in Figure 4.

3.3. Software Layer

The software layer is a central component of the virtual calibration platform, acting as a bridge between the physical hardware and simulation models to ensure the efficient operation and functionality of the system. As a critical element, the software layer manages both hardware and software resources, processes high-frequency signals, and provides robust support for the system’s stability and scalability. It consists of two main software components: LabVIEW 2023Q1 and VeriStand 2023Q2, each performing specific functions while working collaboratively.

In the virtual calibration platform, LabVIEW handles the core task of high-frequency signal processing, primarily by building FPGA models. This enables the processing of complex signals with extremely high time resolution, providing foundational support for the simulation of power machinery systems. The programmability of FPGA endows the platform with high flexibility, allowing the modular design to achieve versatile signal generation and processing functions. It not only supports high-precision simulation of complex signals but also quickly adapts to various testing requirements, offering robust flexibility and scalability for the virtual calibration platform. Its main functions include simulating crankshaft and camshaft position signals, acquiring fuel injection and ignition drive signals, and processing high-frequency protocol signals such as Pulse-Width Modulation (PWM) and Single-Edge Nibble Transmission (SENT). These functions are implemented through parallel development and precise scheduling, ensuring simulation accuracy and stability for power machinery controllers under complex conditions. The FPGA model architecture of LabVIEW is shown in Figure 5.

VeriStand is primarily responsible for managing hardware and software resources, real-time monitoring, and data visualization, providing a unified operational framework to support simulation tasks. Communication between VeriStand and the controller is established via Ethernet, while hardware interaction with DAQ, CAN, FPGA, and other cards is achieved through the PXI bus, ensuring the efficient and stable transmission of hardware signals. For software resource management, .lvbitx files generated by LabVIEW, .vsmodel files generated by Simulink, and .fmu files generated by AVL CRUISE M™ [23] are deployed into the controller via VeriStand, where they are subsequently transferred and loaded into the corresponding hardware modules. This process ensures that simulation models are seamlessly integrated with the physical hardware resources through VeriStand’s unified management framework. Additionally, VeriStand features a flexible graphical interface for real-time signal monitoring and data visualization, making system monitoring and diagnostics more intuitive. The signal communication architecture of VeriStand is given in Figure 6.

Through the close collaboration of LabVIEW and VeriStand, the software layer achieves efficient management of hardware and software resources, along with data processing and monitoring, laying a solid foundation for the stable operation of the entire virtual calibration platform.

3.4. Model Layer

The model layer is an important component of the virtual calibration system, primarily simulating the dynamic behavior and operating characteristics of controlled objects through simulation technologies to provide digital support for system development and optimization. By accurately replicating the operational states of real systems in a virtual environment, the model layer enables users to conduct detailed studies of complex physical systems under laboratory conditions, thereby reducing the time and costs associated with physical testing. In the current study, the model layer comprises a CRUISE M™-based plant model and a Simulink-based signal interaction model.

The signal interaction model is a key component of the model layer, responsible for signal processing and interactions among modules. To ensure accuracy and reliability, the signal flow process is divided into eight functional modules. First, the interface model maps interfaces to VeriStand, ensuring seamless signal interaction between the model layer and the software layer. The database model handles bidirectional conversion between electrical signals and physical signals, ensuring signal accuracy during simulations. The CAN communication model simulates CAN bus communication, enabling real-time data exchange and control between different hardware modules. The FPGA host model works in coordination with the LabVIEW FPGA model to compute and process complex signals, supporting real-time simulation of high-frequency signals. The power-on logic control model incorporates fail-safe mechanisms to monitor and control the power supply process, enhancing system safety and reliability. The dynamometer model simulates the load consumption of power and calculates rotational speed, providing realistic load characteristics for simulations. The injector model calculates injection quantity based on the collected injection duration, ensuring injection precision. The rail pressure model calculates and adjusts rail pressure in real time based on speed, ensuring fuel supply stability and efficiency. Through the collaboration of these modules, the signal interaction model establishes an effective bridge between the model layer and other layers, providing solid support for the efficient operation of the virtual calibration platform. The specific architecture is shown in Figure 7.

3.5. Application Layer

The application layer serves as the interactive interface of the virtual calibration platform, aiming to combine the complex simulation functions of the underlying layers with operational requirements while simplifying the operational process to enhance system usability and efficiency. Through an integrated design, the application layer presents key functional modules in an intuitive manner, providing a unified operational environment. With an all-in-one interface, the application layer enables the centralized management of tasks such as model loading, simulation execution, parameter adjustment, and signal monitoring, thereby significantly reducing operational complexity. Additionally, the application layer features excellent scalability, allowing flexible adjustments to functional layouts and configurations according to specific application scenarios to meet the development and verification needs of various calibration tasks.

The application layer is constructed using the AVL PUMA Open 2™ [24] automated control system, which provides comprehensive support for complex power machinery calibration tasks with its robust automation capabilities and flexible control modes. The PUMA system’s efficient automation functions support batch execution of test conditions and automated data processing, simplifying the operational process during calibration, significantly reducing the need for manual intervention, and improving work efficiency. In terms of control modes, PUMA offers various flexible control options to meet the testing needs of different power machinery. Common control modes include speed–torque mode, speed–throttle mode, and torque–throttle mode, which enable precise engine control by regulating parameters such as speed, torque, and throttle position, ensuring consistency and repeatability during testing. By organically integrating automation, data processing, and flexible control modes, the PUMA system ensures efficiency and precision in power machinery calibration tasks, becoming an indispensable part of the virtual calibration platform and providing solid technical support for its stable operation.

3.6. System Integration

System integration brings together the various modules of the virtual calibration platform, creating a cohesive and highly efficient system. The integration process involves key steps such as the physical connection of hardware modules, the establishment of communication networks, and the loading and allocation of software resources. Through standardized interfaces, stable and reliable data transmission and signal interaction between modules are achieved, ensuring smooth system operation and optimized overall performance. As a result, the integrated platform demonstrates excellent module compatibility and scalability, favoring independently designed modules with deep collaborative functioning under complex operating conditions. The integration principles and prototype are displayed in Figure 8.

4. Test and Results

4.1. Test Method

The testing process was conducted following a comprehensive system validation approach. This approach consisted of five key steps: module functionality validation, software model validation, unit collaboration validation, system integration validation, and platform performance validation. The first four steps focused on verifying the accuracy of individual module functions, the integrity of software models, and the collaborative operation between modules. These steps laid a solid foundation for the final stage of performance validation, which is aimed at evaluating the platform’s computational capabilities, real-time performance, precision, and stability under practical operating conditions.

In the platform performance validation, a diesel/natural gas dual-fuel engine ECU was selected as the test object. A plant model was constructed using CRUISE M™ and engine external characteristic data. This model, combined with a MATLAB/Simulink-based signal interaction model and a LabVIEW-based FPGA model, formed a complete virtual simulation environment for the engine. The testing process was automated, and data acquisition was performed using PUMA Open 2™, with input data consisting of target engine speed and torque. Real-time signal interaction with ECU was facilitated through the independently developed virtual calibration platform. Hardware resources and software models were centrally managed and monitored in real time by using VeriStand. Final data analysis and processing were conducted via PUMA Open 2™. The testing was divided into two parts: steady-state testing and transient testing. Steady-state testing, based on the engine’s steady-state characteristics, primarily assessed the platform’s computational accuracy, system stability, and adaptability under a wide range of constant operating conditions. Transient testing, conducted following the WHTC cycle, evaluated the platform’s responsiveness and real-time performance under dynamic and rapidly changing conditions, verifying its capability of handling swift transitions between operating states. Details of the tools and equipment used are provided in

Table 2. Tools and equipment used in testing.

Category	Tool/Equipment	Version Information	Function
Software	AVL PUMA Open 2™	2.0	Automated testing tool for real-time data acquisition and analysis.
	VeriStand	2023Q2	Management of hardware resources and software models, providing real-time data monitoring.
	LabVIEW	2023Q1	Development of FPGA models to support high-frequency signal processing and complex simulations.
	MATLAB/Simulink	R2020a	Development of signal interaction models for real-time processing and signal transmission.
	AVL CRUISE M™	R2023.2	Construction of plant models and provision of engine environment simulation.
Hardware	Virtual Calibration Platform	Independently developed	Facilitates real-time signal interaction with the ECU and supports efficient simulation tasks.
	ECU	Diesel/Natural Gas Dual-fuel Engine ECU	Provided by a partner company to validate the electronic control requirements of the engine.
Models	Plant Model	Based on CRUISE M™	Semi-empirical physical model supporting high-precision simulations under multiple operating conditions.
	Signal Interaction Model	Based on MATLAB/Simulink	Enables multi-module signal processing and interaction, supporting real-time simulation requirements.
	FPGA Model	Based on LabVIEW	Specializes in high-frequency signal simulation and processing, supporting complex signal implementation.
Data	Engine External Characteristics Data	-	Provided by a partner company for calibration and validation of the model.

.

4.2. Steady-State Testing

In the steady-state testing, the performance of a diesel/natural gas dual-fuel engine ECU was evaluated under a broad range of steady-state operating conditions. The engine speed ranged from 700 rpm to 1900 rpm, while the torque varied from the full-load characteristic curve to the minimum torque value. Real-time signal interaction between the virtual calibration platform and the ECU was established to generate simulation data covering various speed and torque combinations, enabling a comprehensive evaluation of the simulation results for computational accuracy and stability.

The results are presented in two formats: graphical plots and a table. Figure 9a illustrates the fuel flow rate distribution across different engine speed and torque combinations, while Figure 9b depicts the gas flow rate distribution. All data within each group were normalized to their respective maximum values. The experimental data (left) were limited to full-load operating points, whereas the simulation data (right) are displayed as two-dimensional contour maps, showing variations across a wide range of speed and torque combinations. To further quantify the consistency between simulation and experimental results,

Table 3. Absolute errors in fuel and gas flow rates.

Engine Speed (rpm)	Fuel Flow Rate Absolute Error (×10⁻3 kg/s)	Gas Flow Rate Absolute Error (×10⁻3 kg/s)
700	0.15	0.02
800	0.20	0.03
900	0.06	0.08
1000	0.21	0.10
1100	0.61	0.09
1200	0.10	0.09
1300	0.06	0.10
1400	0.13	0.11
1500	0.19	0.11
1600	1.20	0.16
1700	2.24	0.19
1800	1.42	0.23
1900	2.26	0.23

lists the absolute errors of fuel and gas flow rates at full-load conditions for different engine speeds.

The graphical and tabular data indicate that the simulation results show small deviations from the experimental data under full-load conditions, with errors remaining within a reasonable range. This demonstrates the high accuracy and reliability of the platform in simulating fuel and gas flow rates. Furthermore, the simulation data extend the coverage of operating conditions to include partial and low-load scenarios, highlighting the platform’s adaptability across a wide range of steady-state conditions. These findings validate the virtual calibration platform’s simulation capabilities under steady-state conditions, providing strong support for steady-state performance evaluation in power machinery controller calibration.

4.3. Transient Testing

The transient testing was conducted based on the World Harmonized Transient Cycle (WHTC) using an automated testing process. A total of 18,000 transient operating points were evaluated, with each point lasting 0.1 s, resulting in a total testing duration of 1800 s. Engine speed and torque were used as input variables, and real-time signal interaction and data acquisition were performed via the virtual calibration platform to comprehensively assess the platform’s response speed and real-time performance under rapidly changing operating conditions.

The simulation results are presented in Figure 10. Figure 10a,b illustrate the comparison between the target and simulated engine speed, as well as the target and simulated engine torque over time, clearly showing the responsiveness of the simulation results to target variations. Figure 10c,d display the normalized trends of fuel flow rate and gas flow rate over time, reflecting the platform’s fuel characteristics under complex dynamic conditions.

Figure 10a,b demonstrate that the simulation results can quickly respond to changes in the target data, maintaining good tracking accuracy across the majority of operating points. This highlights the platform’s high responsiveness and real-time performance under dynamic conditions. Figure 10c,d show trends in fuel flow rate and gas flow rate over time, indicating that the platform can effectively handle complex dynamic inputs, process dynamic signals reliably, and efficiently perform data acquisition and processing tasks. These results confirm that the virtual calibration platform exhibits excellent real-time simulation capability and stability under transient conditions, providing strong support for dynamic performance validation in power machinery controller calibration.

5. Discussion

This study developed a virtual calibration platform specifically designed for power machinery. By employing a modular design approach, the system was divided into distinct layers, including the physical layer, communication layer, software layer, model layer, and application layer. This layered architecture enhanced the platform’s scalability, maintainability, and adaptability while facilitating efficient integration and independent module operation. The platform’s performance was validated based on steady-state and transient testing using a diesel/natural gas dual-fuel engine ECU. These tests evaluated its capabilities in real-time data processing, the accurate simulation of engine operating characteristics, and stable operation under dynamic conditions.

The experimental results confirmed the platform’s effectiveness and reliability. For instance, under steady-state conditions, the absolute errors of fuel flow rate and gas flow rate were consistently below 2.5 × 10⁻³ kg/s across full-load points, demonstrating high computational accuracy and adaptability. Similarly, during transient testing, the platform exhibited a rapid response to target changes, maintaining a relatively good consistency between simulated and target values. However, we still observed differences between simulated and experimental data. These differences are due to limitations in experimental data and modeling capabilities, resulting in the virtual simulation environment not being 100% reflective of the real operating environment. Despite some differences, the platform effectively reflected engine operating characteristics, validating its applicability for real-time controller calibration tasks. Additionally, this study also has some limitations in the type of controller validated, and future studies may involve a broader range of controllers to further evaluate the adaptability and applicability of the platform in different systems.

Despite these limitations, the results of this study provide strong evidence of the platform’s effectiveness and reliability, showcasing its potential for diverse applications in controller calibration. Future research could expand the validation scope, explore additional operating conditions and controller types, and incorporate higher precision experimental data to further improve the platform’s comprehensiveness and accuracy.

6. Conclusions

This study proposed a virtual calibration platform for power machinery, capable of accurately modeling controlled objects and simulating their real-world performance in a virtual environment. The platform adopted a modular design and layered architecture, effectively improving system maintainability and scalability. Experimental validation demonstrated the platform’s excellent stability and precision across a variety of operating conditions. Under steady-state testing, the absolute error of the simulation results of the fuel flow and gas flow of the platform is less than 2.5 × 10⁻³ kg/s, while under the transient condition of 0.1 s, the platform also demonstrated a relatively good rapid response to target changes. These results confirm that the platform meets the high requirements for accuracy and real-time performance in calibrating power machinery controllers. In the future, as more operating conditions and controller types are validated, the platform’s application scope will further expand. It is expected to demonstrate even greater potential in supporting controller calibration under more complex operating scenarios, offering a robust solution for power machinery development and testing.